September for Bittensor proved the money was real. October is where it gets interesting, because this month the protocol changed the rules while that money kept flowing in.

Two governance upgrades shipped 12 days apart. Null Consensus broke open the 256-miner ceiling that has capped every subnet since mainnet. eTAO gave subnet owners the ability to choose how emissions split between owners, validators, and miners, instead of the protocol forcing the same 18/41/41 ratio on everyone. Both were structural changes to a system that already had revenue running through it.

And the revenue didn’t pause while the protocol retooled. SubConnect raised its annualized estimate to $50M+, a 40% jump from September. Three Wall Street firms filed separate TAO ETF products in the same month. Affine shipped the first public AI video studio built on Bittensor infrastructure.

The protocol is being rebuilt under load, and the load is increasing.

Five stories. Here is your compass for October.

How to read this compass

I’m organizing these updates by direction: North, East, South, West to give you a repeatable mental model for navigating Bittensor’s complexity.

North: Foundation 🏛️ (governance, protocol, core team, security events) The structural layer. Changes here determine who controls the network and how decisions get made.

East: Capital 💹 (markets, volume, liquidity, integrations) The economic layer. This is where price meets liquidity and institutional conviction shows up in the data.

South: Technology ⚙️ (shipped features, benchmarks, infrastructure, ecosystem health) The execution layer. What actually works, what got built, and where’s the technical proof.

West: Adoption 📡 (narrative, mainstream validation, external voices) The perception layer. How the outside world sees Bittensor, from industry titans to normie investors.

This month, North delivered some of the most impactful protocol changes in Bittensor’s recent history. East brought the money. South shipped a product you can actually use. West was quiet, and that’s fine. Not every month needs a narrative moment. Sometimes the story is that the plumbing works.

🏛️ Null Consensus: subnets break the 256-player ceiling

On October 7, Bittensor shipped Subtensor Runtime 475. The headline: Null Consensus, an alternative to Yuma Consensus that lets a subnet support up to 2,500 registered UID slots instead of Yuma’s 256.

Under Yuma, every validator scores every miner. It’s distributed, democratic, and slow. Under Null, one validator sets the weights. It’s faster, scales to nearly 10x the participants, but concentrates trust in a single scorer.

Subnet owners now choose between the two. Yuma for distributed validation with 256 slots. Null for scale with 2,500 slots and one validator calling the shots.

Why it matters: Under Yuma, a subnet has 256 UID slots shared by miners and validators, and multiple stake-weighted validators score miners through consensus. For compute marketplaces that want to register and reward many more participants, that can be restrictive.

Null raises capacity to 2,500 UIDs per mechanism, but gives scoring power to a single validator: the UID with the highest stake weight. That validator alone sets the weights that determine miner rewards. This removes Yuma’s consensus and bond-based defenses, creating a concentration risk: if the elected validator is malicious, mistaken, or unavailable, miner rewards can be distorted, or with no valid weights, fall back to an equal split across all registered UIDs.

In short, Null trades distributed, resilient scoring for much larger subnet capacity and simpler, high-throughput reward allocation.

The takeaway: Watch which subnets switch to Null Consensus. Compute and inference subnets with high miner counts are the natural adopters. Quality-sensitive subnets (text evaluation, fact-checking, AI safety) will likely stay on Yuma. The subnets that pick wrong, choosing Null when they need distributed scoring, or staying on Yuma when they need scale, will lose ground.

🏛️ eTAO: subnets choose their own emission splits

Twelve days after Null Consensus, on October 19, Bittensor activated eTAO (Runtime v481). Until then, every subnet split alpha emissions the same way: 18% owner, 41% validators (and their stakers), 41% miners. One size fits all.

Under eTAO, each subnet owner chooses how to divide newly emitted alpha among owners, validators, and miners. The protocol sets guardrails (every subnet must still pay the network and pay its miners), but within those limits, owners set their own ratios.

Early adopters are already diverging. Vanta removed its owner cut entirely, going to zero. Almanac increased miners’ share. Three subnets allocated 33% to owners. The chain recalculates each subnet’s baseline share every block.

Why it matters: This is the second half of the tokenomics overhaul that started with the emissions refactor in May. The old 18/41/41 split treated a compute marketplace and a text-scoring subnet identically, despite radically different cost structures. A compute subnet burning thousands in GPU costs needs a different emission profile than a fact-checking subnet running on consumer hardware.

eTAO lets subnets tune their economics like a mixing desk. More to miners to attract supply. Less to owners to signal alignment. More to validators to secure the network. The market will now determine whether a subnet’s emission split is well-designed or extractive.

And Vanta going to zero owner cut is a subnet voluntarily giving up its founder share. The thesis that subnets will evolve toward Bitcoin’s model (no founder allocation, no pre-mine, emissions purely reflecting work value) just got its first real-world data point.

The takeaway: Track eTAO split configurations on taostats. Subnets that cut owners too aggressively may struggle to fund operations. Subnets that take too much may bleed miners. The right split depends on the subnet’s cost structure, and the market will reward the ones that get it right.

💹 Three TAO ETF filings: Wall Street lines up

October saw three separate U.S. ETF filings for TAO:

Grayscale filed Amendment No. 2 to its S-1 on October 23, detailing plans to convert the Grayscale Bittensor Trust into the Grayscale Bittensor ETF on NYSE Arca under ticker GTAO. It would hold spot TAO.

Bitwise filed for a TAO Strategy ETF.

REX-Osprey filed for a TAO + Staking ETF, combining spot exposure with staking yield.

Why it matters: Three Wall Street firms filing TAO products in the same month is the strongest institutional signal Bittensor has received. Grayscale’s amendment is the most advanced, past the initial filing and into detailed listing mechanics. But REX-Osprey’s staking component is the interesting one. It would let ETF holders earn validator yield without operating a validator. That bridges the gap between passive exposure and active participation, something no other AI crypto project offers.

ETF approval timelines run in months, not weeks. But the filings themselves change the conversation. Financial advisors, family offices, and institutional desks that can’t custody TAO directly now have a pipeline. The question is whether the SEC greenlights a spot TAO ETF before the field gets crowded.

The takeaway: Grayscale has the first-mover position. Next to watch is SEC comment letters or deficiency notices on the S-1 filings. Those signal whether the regulator is engaging seriously or stalling. And as for REX-Osprey’s staking structure, if the SEC allows staking yield inside an ETF wrapper, it creates a regulated path for institutional investors to earn Bittensor validator returns.

💹 SubConnect revenue hits $50M+: the trend line holds

On October 2, SubConnect updated its Bittensor Revenue Index. The annualized subnet revenue estimate jumped from $32 to $35M (September) to $50M+ in one month. That’s a 40% month-over-month increase. Lium, Targon, and Engy drove most of the growth.

In September, I wrote that the Revenue Index was a baseline. October turned it into a trend line. RVCrypto’s $100M end-of-2026 target now looks less aspirational and more like a trajectory.

Why it matters: A 40% MoM jump means one of two things: revenue is genuinely accelerating, or the September estimates were conservative. Either way, the gap between $50M and $100M is no longer the entire investment thesis. It’s a growth question, not an existence question.

Also worth noting is that 14+ subnets are running active token buyback programs: real revenue purchasing alpha tokens on the open market. But generating revenue is step one. Building a sustainable business with margins, customer retention, and competitive moats is something else entirely.

The takeaway: The next volume of the Revenue Index is the real test. If it shows continued growth, the flywheel narrative gets data behind it. If it flattens, the bull case weakens. Until then, use the index as a screening tool. The subnets generating real revenue with active buybacks are the ones worth watching. The ones with no revenue and no buyback mechanism are running on emissions alone, and emissions are a shrinking resource post-halving.

⚙️ Affine ships Bittensor’s first public AI video studio

Affine (SN120) launched Affine Video, pairing a public MiniMax-H3 studio with a Bittensor competition to improve sparse attention in video generation. The goal is narrower than “decentralized video generation”: reduce inference compute by improving which attention blocks the model computes, without changing the underlying generator.

Miners train a selector that decides which attention blocks to keep. If it works, Affine gets lower cost per clip; if it fails, users get faster but less coherent video. The base model, inference stack, and generation schedule stay fixed, so this is a focused optimization competition rather than an attempt to train a frontier video model.

Why it matters: this is one of the first consumer-facing AI video products tied to Bittensor infrastructure. It does not yet compete with Runway or Pika on product depth, but it tests whether decentralized incentive mechanisms can improve a real inference bottleneck rather than merely produce benchmark scores.

Video is an unforgiving testbed, but even if sparse attention can save compute, dropped attention can show up as broken motion, inconsistent objects, or weaker prompt adherence. The relevant comparison for now is Affine’s sparse-attention output versus full-attention MiniMax-H3 at the same prompt and settings.

The takeaway: watch measured quality retention and cost reduction, not launch hype. If Affine can show near-parity quality at materially lower cost, it becomes evidence that Bittensor subnets can improve real AI products. If quality collapses as sparsity rises, it remains an interesting mechanism in search of a product.

Final thoughts

North (Foundation) ran this month. Null Consensus and eTAO together represent some of the most significant protocol changes since dTAO. Null changes who participates. eTAO changes how they’re paid. The protocol is in a phase of rapid iteration, shipping faster than most ecosystems plan, let alone execute.

But three runtime upgrades in one month (v475, v481, v484) plus the v482 patch is an extraordinary release cadence. It creates opportunity and risk. Every upgrade changes the rules and subnet operators are adapting in real time.

East delivered the capital story. SubConnect proved the revenue trend is real. Three ETF filings brought Wall Street to the door. The type of investor who starts paying attention when there’s an S-1 filing is different from the type who hangs out on the Subnet Summer Telegram group.

South (Technology) shipped something tangible. Affine Video is a product a non-technical person can understand: an AI video generator that runs on decentralized compute. You don’t need to know what dTAO means to grasp why that’s interesting.

And West was quiet. No narrative moment this month. But honestly, October didn’t need one. The protocol did the talking.

The through-line for all five stories: Bittensor spent October proving it can change the rules and grow at the same time. The protocol team shipped. The revenue grew. Wall Street filed. A product launched. The question for November is simpler: can the subnets keep up with the protocol?

Some can. Some can’t. That divergence is the story for the rest of 2026.

Until next time.

Subscribe now

Cheers,

Brian

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

Ask anyone who has ever managed infrastructure in corporate tech about their cloud bill, and you’ll get a very specific kind of quiet sigh.

Every single bill follows the exact same script: you size your infrastructure for the worst-case scenario, hit peak capacity for two chaotic days during the month, and then watch in horror as the finance team approves invoice after invoice for hardware that sits entirely idle for the remaining 28 days.

That burned corporate budget isn’t just an inefficiency, it’s the exact problem Lium (Subnet 51) was built to solve.

On paper, the tension is simple: renters need guaranteed, cost-effective GPU power right now, while providers refuse to keep hardware powered on without guaranteed revenue. Traditional cloud giants solve this with heavy-handed central planning, rigid contracts, and massive markups. Lium takes a market-driven approach powered by Bittensor emissions.

Its core mechanism comes down to a “two paychecks” design: paying providers when their GPU compute is actively being rented, and paying them under strict conditions when it sits idle, ready, and available. Understanding how those two paychecks balance out is the key to answering the only question that matters: is Subnet 51 building a sustainable computing business, or just subsidizing idle silicon with tokens?

What Lium Actually Sells

People constantly blur two entirely different forces when evaluating Lium.

The marketplace creates real-world customer demand. The subnet mechanism structures and subsidizes the supply.

Collapsing those two concepts into a single narrative leads straight to flawed analysis. In this ecosystem, it’s easy to mistake a successful incentive mechanism for genuine product-market fit. Seeing high subnet activity doesn’t necessarily mean customers are beating down the door: it often just means the token rewards are doing their job.

For any provider asking the practical question “Should I actually list my GPUs on Lium?” evaluating the opportunity requires dissecting both income streams independently.

Paycheck One: Rental Revenue

A renter books GPU time. Lium bills the renter in USD. Lium distributes the provider’s share daily to their coldkey after a settlement delay. Lium advertises a 95% provider share of rental revenue, paid daily on-chain at 17:00 UTC. Exact terms and bonuses depend on their current rewards configuration, so check the official docs and payout details before committing hardware.

This is the income stream most directly tied to customer value. It reflects actual demand for compute: if nobody rents your GPU, this paycheck is zero.

Paycheck Two: The Emission Layer

Every Bittensor tempo, SN51 receives a share of network emissions. Validators then decide how to allocate that share among providers. Lium’s design splits that allocation into three buckets: rented GPUs, eligible idle GPUs, and burn.

Rented pool: A fixed 13% of the provider-facing emission allocation goes only to currently rented nodes, scored by GPU model and count. This rewards utilization directly.

Unrented pool: Up to 87% can go to eligible idle nodes that meet Lium’s price, disk, validation, and model requirements. This is the “inventory readiness” payment.

Burn pool: Whatever is not allocated to eligible idle providers is burned. Those tokens are removed from supply. No provider receives them.

The unrented pool is not inherently wasteful. It compensates providers for maintaining capacity before demand arrives. Think of it like paying a taxi driver to stay at the rank during a quiet afternoon, so they’re there when the evening rush hits.

But it becomes economically questionable if idle rewards stay large while utilization, pricing discipline, or customer demand don’t improve. You’re paying for readiness that never gets used.

Why Idle Pay Exists

The obvious objection: “Why should idle GPUs earn anything?”

GPU marketplaces have a cold-start problem. They need supply before they can attract demand. Providers won’t list hardware if income arrives only after a rental occurs. The risk is too high. Idle pay reduces that risk and encourages a deeper, more responsive supply pool.

But it also creates moral hazard. Providers may focus on eligibility and emissions rather than competing on rentability. If you can earn emission rewards by keeping a node online that meets minimum requirements, the incentive to compete on price or performance weakens.

This is a design trade-off. Lium is paying providers to be ready for the next customer. The enduring question, and this applies to all subnets, is whether they can reduce that subsidy over time as real demand grows. If Lium’s unrented pool stays near 87% while billed revenue plateaus, the system is subsidising supply that demand doesn’t want.

Is Customer Demand Real?

Let’s look at the evidence.

Lium’s most recent monthly review, published on 10 October, reports $1.4M in GPU rental spend, up 45% month-over-month, from 1,379 paying customers, up 16%. Rentals more than doubled to 17,497, while 68.7% of launches were initiated programmatically by code or AI agents. Lium also reported 32% fewer rental failures, 32% more first-time renters, and 15% more providers.

Since then, the demand signal has become more immediate. On 22 October, Lium reported that 91% of its listed GPU supply was rented, across a fleet it valued at more than $30M and including 225 B300s. Its public network dashboard currently shows a similar picture: 1,614 of 1,799 listed GPUs rented, or roughly 90% of supply, with 83% utilization over the trailing 30 days.

Agent-initiated demand is the kind of usage that compounds. Agents don’t browse marketplaces. They call APIs, book capacity, run workloads, and release. If such a large proportion of your rentals are happening without human hands, you have early evidence of repeatable compute consumption.

OpenTensor also reported that Lium used marketplace revenue to buy back and burn nearly $2.7M of SN51 alpha across five buybacks from April to September, removing roughly 171,000 alpha tokens from supply.

Third-party estimates back up the general direction, even if they don’t tell a single uniform story.

• SubConnect (October 2) put total annualized Bittensor subnet revenue above $50M, attributing roughly a third of that pie to Lium.

• SubnetRank estimates Lium’s external revenue at $8M–$10M annualized, placing it squarely at the top of its 24-subnet tracking cohort.

• Own Your Mind’s integrity board reported September billing alone at $1.295M, which annualizes out to roughly $15.5M.

These are undoubtedly encouraging signals for the subnet’s trajectory, but let’s be disciplined about what we’re looking at here: this isn’t one clean, audited dataset. What we have is a mix of team-reported billing, third-party projections, and buyback disclosures, each spanning different timeframes and relying on different methodologies.

The strongest new evidence is that Lium’s demand and supply metrics now point in the same direction: billed revenue is rising, agent-initiated usage is high, and most listed GPUs are rented. The open question is is whether Lium can sustain that utilization as it adds supply.

Does the Incentive Design Work?

Each income stream proves something different.

Lium has more evidence of external demand than most emission-led subnets. The monthly billing and high agent-initiated rentals are material. The buybacks show revenue flowing back to token holders.

The open question is the ratio between customer-funded revenue and emission-supported supply income. If a provider earns more from sitting idle than from being rented, the incentive system and the marketplace are pulling in different directions. If rental income dominates, the emission layer is doing its job as a bridge.

High utilization makes that question more tractable. When 90% or more of listed GPUs are rented, idle-pay emissions are supporting a smaller share of the fleet at any given moment. But utilization is volatile: Lium’s own dashboard distinguishes a roughly 90% current utilization rate from an 83% trailing-30-day rate. That gap is exactly why the rented-versus-idle emission split still needs to be published consistently.

Lium does publish aggregate network statistics through an unauthenticated public stats feed, including billed revenue, GPU-hours, rentals, listed versus rented GPUs, and 30-day utilization. That is a real improvement in transparency. What is still missing is a standardised monthly reconciliation of emission allocation across rented, unrented, and burn pools, alongside provider payouts and buyback value for the same period.

VITALS Scorecard

I ran Lium through my VITALS framework using a three-level scale: Pass, Partial, Fail. Unsurprisngly, given its place in the top tier of Bittensor subnets, it has scored extremely well. We’re not uncovering a diamond in the rough here.

Value signal: Pass. “Rent GPUs by the second; GPU owners earn when their machines are idle.” Anyone can repeat the proposition without needing to understand TAO, validators, or mining. Lium positions itself as a GPU rental marketplace with per-second renter billing and provider income from both rentals and idle supply.

Interface: Pass. Lium explicitly offers a web dashboard, CLI, Python SDK, REST API, and agent skill; the SDK can list nodes, bring up a pod, run a command, and tear it down; pods are billed per second with no minimum.

Traction: Pass. Lium clears the threshold easily: SN51’s emission share is roughly 14.5–15%, well above 0.20%, and the 30-day trend is positive. It also publishes multiple impressive usage metrics.

Access: Pass. Lium has a documented MCP integration for Claude, ChatGPT, Codex, and other MCP clients, plus a Datura-published Agent Skill for Claude Code, Cursor, and Codex. These are real distribution surfaces for agent developers, though not yet deep platform partnerships or verified customer pilots.

Liquidity/Stake: Pass. Supply-side confidence is visible: Lium has active validator participation, public validator and engineering updates, and a clear mechanism through which validators assess GPU hardware and performance to allocate incentives. Providers also receive daily on-chain payouts, giving the supply side a recurring, transparent reward loop.

Sustainability: Partial. Pricing is transparent and competitive, however, this is a pricing-and-revenue-model pass rather than proof of full unit-economics sustainability: emissions still support provider economics, the burn allocation is dynamic, and the durability of rental-funded buybacks remains to be demonstrated over time.

Score: 5 Passes, 1 Partial. The marketplace has a clear value proposition, an unusually strong agent-native interface, meaningful early traction, emerging distribution through agent tools, and visible supply-side confidence. The remaining question is sustainability: rental pricing is competitive and revenue is real, but emissions still materially support provider economics, so Lium has not yet proven it can stand on rental revenue alone.

My Recommendations

1. Tie more emission weight to measured utilization.

Gradually shift allocation from static eligibility toward realised GPU-hours, repeat renters, price competitiveness, and successful job completion. The goal is to make idle pay responsive to evidence of demand. If a provider’s node has been idle for 30 days with zero rentals, the emission weight should reflect that. Right now, the unrented pool rewards eligibility. It should reward readiness that converts.

The new utilization disclosure makes this recommendation more actionable: Lium already tracks current and 30-day utilization publicly. The next step is to connect that same utilization signal to provider-level emission allocation in a transparent, auditable way.

2. Publish a canonical monthly marketplace dashboard.

Include billed revenue, active renters, rentals, GPU-hours, utilization rate, emission allocation across rented/unrented/burn, provider payouts, and burn value, all for the same reporting period. This would make the “two paychecks” comparison reproducible. Right now, we’re stitching together team-reported figures, third-party estimates, and buyback disclosures from different sources. A single, standardised dashboard would settle the ratio question.

The Real Test

Lium has designed an incentive system that pays for both demand and readiness. The encouraging part: customer revenue is real and growing with more evidence of external demand than most Bittensor subnets can point to.

The unresolved part: whether emissions are a temporary bridge to a functioning market, or a permanent cost of keeping supply online.

The next phase of Lium’s story will be decided by which of those two paychecks grows faster.

Subscribe now

Until next time.

Cheers,

Brian

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

Every 12 seconds, Bittensor mints a block of TAO and drops it into a system most people have never mapped end to end.

They see the price ticker and the subnet rankings. What they miss is the plumbing between those two things.

I’ve been involved in Bittensor for three years now, and I still meet smart investors who can explain alpha tokens but can’t tell you what actually happens to a single block of TAO after it’s created. They know the ingredients, but they can’t describe the recipe.

My wife can explain Bittensor to a friend using the mall metaphor, but ask her how the payment actually flows from the mall’s cash office to the store employees, and she’s stuck. That gap is where people fall off.

So let’s follow one block. From the moment it’s minted to the moment it lands in someone’s wallet.

The mental model that makes everything click: Bittensor runs on two clocks:

The first ticks every 12 seconds (one block, one mint, one injection).

The second ticks every tempo, roughly 72 minutes, when accumulated alpha gets distributed to participants.

The block clock is for injection, the tempo clock is for distribution. Hold those two in your head and the rest follows.

Step 1: Mint

The block subsidy is where everything starts.

Every 12 seconds, the network creates new TAO. Before the first halving in December 2025, that was roughly 1 TAO per block, about 7,200 per day. Post-halving, it’s 0.5 TAO per block, 3,600 per day. I covered the halving mechanics in The First Bittensor Halving: The End of Easy Mode.

Think of this as the network’s budget for the next 12 seconds. How much new value can be allocated to subnets right now? That single number is the top of the funnel. Every split, every filter, every distribution downstream starts from here.

The halving schedule (copied from Bitcoin, 4-year cycles) sets the long-run supply path. The per-block cadence sets the heartbeat for liquidity injections. Every 12 seconds, the network decides where to spend its allowance.

Subscribe now

Step 2: Demand

Now the freshly minted TAO needs to be divided among 128 subnets. How does the network decide who gets what?

It starts with demand. Each subnet’s demand share is driven by its Exponential Moving Average (EMA) alpha price relative to the sum across all emission-enabled subnets. The EMA smooths out single-block volatility so emissions don’t whipsaw on transient price spikes. A subnet that gets wash-traded for one block doesn’t suddenly redirect the network’s budget.

Then there’s miner burn. Miners receive a fixed share of each subnet’s alpha as payment for their work (we’ll break down the exact split in Step 5). But validators can vote to withhold some or all of that share from miners. The withheld alpha gets recycled instead of paid out. They do this when the work isn’t worth the full payout, when someone’s gaming the system, or sometimes collaboratively with the subnet owner to reduce sell pressure (miners are the most likely to dump alpha for TAO to pay bills).

The network treats this burn rate as a direct discount on demand. The formula: demand is proportional to EMA price multiplied by (1 minus burn rate). A subnet burning 100% of miner emissions has zero effective demand, no matter how high its alpha price is. A subnet burning 50% has its demand halved. A subnet paying miners in full faces no discount.

The network is saying: if you’re not compensating the people doing the work, your price signal doesn’t count as much. You can have a high alpha price and still get penalized.

In effect, demand is market-validated pull. Sustained price strength plus honest miner payouts means higher effective demand. You can’t fake it with a single block of volume. You can’t game it by withholding miner pay. The signal has to be durable, and it has to be honest.

Step 3: Filter

This is where the v440 Emission Gate comes in. If you read my three-level breakdown, you’ll recognize the mechanism.

The gate takes those burn-adjusted demand shares and passes them through a smooth cutoff around a quantile threshold (theta θ). Subnets with weak demand get suppressed. The emission they would have received gets freed up and reallocated to higher-signal subnets.

The math: final emission weight for each subnet equals its demand share multiplied by the gate function, normalized across all subnets. Low-signal subnets contribute their freed emission to the pool. High-signal subnets pick up the difference.

Why a gate instead of raw proportional sharing? It prevents fragmentation. Without the gate, emissions spread thinly across every subnet that exists, including the ones nobody is using. The gate concentrates liquidity where the market is actually signaling strength.

Demand gets you into contention, now the gate decides how much of the prize pool you actually receive.

Step 4: The liquidity boundary

Each subnet gets its slice of TAO for this block. The gate already decided how big that slice is. Now the question is: where does that TAO go?

It goes into the subnet’s liquidity pool, paired with newly minted alpha. But there’s a limit. Every subnet has a setting called root proportion that acts like a valve on how much fresh TAO can flow into the pool per block. If the slice fits under the valve, all of it becomes pool liquidity. Simple.

If the slice is bigger than the valve allows, the excess doesn’t go to waste. The protocol takes the leftover TAO and uses it to buy alpha directly from the subnet’s pool. Think of it as the overflow pouring into a separate container: the protocol’s own wallet. That purchased alpha stays there. It’s not recycled. It’s not given to miners or validators. The protocol owns it permanently.

Why does this matter? Two reasons.

First, root proportion is widely misunderstood. A lot of people think it decides how much TAO each subnet gets. It doesn’t. The allocation already happened upstream at the gate. Root proportion only controls how much of that allocation becomes pool liquidity versus protocol inventory. It’s a flow valve, not an allocator.

Second, those chain buys quietly accumulate. Every block where a subnet’s TAO exceeds its root proportion cap, the protocol buys a little more alpha. Over time, the protocol builds up a position in subnet tokens across the network. That position becomes relevant if subnets ever get dissolved or their alpha gets redistributed. It’s a slow, per-block accumulation that most people aren’t watching.

Root proportion sets the maximum fresh liquidity per block. Anything beyond that becomes protocol-owned alpha.

Step 5: Distribution

We’ve followed TAO from mint to demand to gate to injection. Now we hit the second clock.

Over the tempo (roughly 72 minutes), each subnet accumulates alpha from two sources: the liquidity injections from Step 4, and the chain buys. Both feed the same pot. At epoch close, that accumulated alpha gets distributed to three groups of people who actually make the subnet work.

18% goes to the subnet owner. This is the creator's cut. The person or team who built the subnet, defined its incentive mechanism, and set the scoring rules. They get a fixed share of every alpha distribution for as long as the subnet is alive. Think of it as a royalty on the work the subnet produces.

The community has debated whether 18% is too high, whether owners should get closer to 1% and be forced to mine their own subnet to earn more. Until now, that debate was academic as the split was welded shut at the protocol level. That's changing. eTAO (runtime v480), shipping imminently, gives subnet owners a dial to adjust their own split within guardrails. The 18/41/41 defaults stay, but owners can now tune the ratio.

41% goes to miners. These are the workers. They run the compute. They produce the intelligence outputs the subnet exists to generate. On an inference subnet, they run the models. On a data storage subnet, they host the files. On a trading subnet, they execute the strategies. The 41% is their pay for doing the actual work.

41% goes to validators and their stakers. Validators are the scorers. They evaluate what miners produce, score the quality, and assign weights. If miners are the workers, validators are the quality control team. The stakers behind them, people who’ve delegated TAO to a validator, get a cut of the validator’s share. That delegation is how passive TAO holders participate in subnet economics without running any hardware.

The split encodes a deliberate balance. Nobody dominates.

But the pot's size isn't fixed. And with eTAO arriving, the split won't be either.

Everything in Steps 1 through 4 determines how much alpha accumulates in that pot before distribution:

• How much TAO was minted (the halving schedule).

• How much demand the subnet generated (EMA price and miner burn).

• Whether the subnet cleared the gate (or got suppressed).

• How much TAO became pool liquidity versus chain buys (root proportion).

By the time the tempo closes and the alpha gets divided, the amount has already been decided. The split just decides who receives what's already there.

That's why a subnet above the gate can sustain all three groups, while a subnet deep in the tail is splitting crumbs.

If you're staking TAO to a validator, your returns depend on two things: how big the pot is (everything upstream, which you can't control) and how much of the validator share your validator earns relative to others (Yuma Consensus scoring, which you can influence by choosing carefully). Once eTAO lands, add a third: whether the subnet owner has adjusted their split, because that directly affects your validator's cut.

What this means for you

For subnet founders: Emissions reward durable price demand and honest miner payouts. Gaming burn or leaning on root proportion for share is outdated. The gate punishes weak demand. The burn formula punishes dishonest payouts. The only sustainable play is building something the market actually wants to stake into. If your subnet can’t answer “what problem, for whom, measured how” with specificity, it won’t survive the gate no matter how much hopium the Discord is pumping.

For validators and miners: The gate amplifies consistent, high-signal participation. Volatility is smoothed (by the EMA) but not rewarded. Steady demand beats sporadic spikes. If you’re on a subnet sitting near the bar, modest increases in demand produce disproportionately larger increases in emission once the gate applies. That’s the highest-ROI position on the network right now.

For investors and stakers: Chain buys create protocol-owned alpha that stays on the protocol’s balance sheet. That matters for long-term supply and any future dissolution or redistribution scenario. When you stake TAO into a subnet’s pool, you’re participating in a system where excess emission becomes permanent protocol inventory. Understand where the TAO goes that doesn’t go into your pool.

For anyone evaluating Bittensor as a platform: The emission stack is a two-clock system (block injection, tempo distribution) with a demand, filter, cap pipeline. That’s an unusual architecture, as most networks have one clock and one distribution rule. Bittensor separates the injection rhythm from the distribution rhythm, and inserts a market-driven filter (the gate) between demand and allocation. If you’re evaluating subnets as businesses, this is the incentive architecture you’re operating inside.

The takeaway

You can explain Bittensor to someone in 30 seconds with the mall metaphor. But if they ask “how does the money actually move?” you need this map.

Five steps. Two clocks. One block of TAO, tracked from mint to wallet.

Each step does work that a simpler design would skip, and the network is better for it. The block clock handles liquidity. The tempo clock handles distribution. Everything in between, demand, the gate, the cap, is the market deciding who earns what.

If you’re holding alpha or watching specific subnets, that two-clock model is where you start.

Subscribe now

Until next time.

Cheers,

Brian

If this map helped you see how the money actually moves, three places to keep going.

5 terms that unlock 80% of Bittensor conversations is the vocabulary survival guide. Five words that show up in every Discord thread, governance proposal, and subnet debate, translated into plain English. If you can't tell a subnet from a validator yet, start there.

The Story of TAO: Bitcoin's DNA, Bittensor's Heartbeat is the big picture. Where TAO came from, how it survived crashing to $30 and rocketing to $757, and why a mechanism called dTAO turned every holder into a venture capitalist in the world's most open AI marketplace. If this piece taught you the plumbing, that one teaches you the story behind it.

If you're further along and trying to explain what you just learned to someone else, I Fixed My Bittensor Pitch walks through what happened when I tried five different pitches on five different people and what actually worked.

Spoiler: the version I practiced in the shower died on contact with a real human at a bar.

All three are free. No paywall. The goal is simple: make this ecosystem legible to people who don't code.

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

I started September with a simple question in my notes file: “Does any of this actually make money?”

That question has followed me through every issue. Believers always said the money would come. Critics always said prove it. Both had a point.

September is when someone finally showed the receipts.

SubConnect's first Bittensor Revenue Index, published in late August, reframed the entire ecosystem conversation. The report estimated $28 to $35 million in annual recurring revenue across 24 subnets, with 14 running active token buyback programs. The numbers are estimates, not audits, but the shift matters.

The rest of the month built on that frame. Yuma crossed $200 million in staked assets. Bittensor hosted its first native conference with a new protocol announcement. A robotics subnet shipped real-world validation tools. And accessibility tools finally made the ecosystem navigable beyond Discord.

Five stories. Here is your compass.

How to read this compass

I’m organizing these updates by direction: North, East, South, West to give you a repeatable mental model for navigating Bittensor’s complexity.

North: Foundation 🏛️ (governance, protocol, core team, security events) The structural layer. Changes here determine who controls the network and how decisions get made.

East: Capital 💹 (markets, volume, liquidity, integrations) The economic layer. This is where price meets liquidity and institutional conviction shows up in the data.

South: Technology ⚙️ (shipped features, benchmarks, infrastructure, ecosystem health) The execution layer. What actually works, what got built, and where’s the technical proof.

West: Adoption 📡 (narrative, mainstream validation, external voices) The perception layer. How the outside world sees Bittensor, from industry titans to normie investors.

This month, East dominated the signal with two stories that redefined how people talk about Bittensor’s economics. North delivered the ecosystem’s cultural milestone. South shipped something you can touch. And West finally showed signs of life after months of silence.

Subscribe now

🏛️ Gamma tokens: Bittensor’s infrastructure becomes currency

Const took the stage at Exploit Summit on September 28 in Montréal and dropped a significant protocol announcement. He called them gamma tokens.

Exploit Summit was Bittensor’s first native conference, a two-day event at on September 28 and 29 that brought over 100 builders together. The conference itself was a milestone, but the thing everyone left talking about was gamma tokens.

Gamma tokens are a canonical standard for usage credits on Bittensor, coming in early 2027. Think prepaid gift cards for subnet infrastructure, except they’re fungible, transferable tokens that live on-chain instead of behind a login screen.

Right now, if you want to rent compute on Lium, run inference on Targon, or store data on Hippius, you go to each subnet’s website, create an account, and pay through their specific interface. Gamma tokens collapse all of that into a single standard. You buy a gamma token for Lium, hold it in your wallet, transfer it to someone else, or use it to rent compute. And because they’re bridgeable, an AI agent on Base or Ethereum could autonomously buy gamma tokens, rent a GPU, run inference, and return results without a human touching a keyboard.

Const named the issuers: Chutes, Engy, GM, Lium, Targon, Hippius, and Affine. Those are the infrastructure providers. Every other subnet becomes a customer.

The internal play is where it gets interesting.

Subnets could redirect portions of their emissions to purchase gamma tokens from other subnets.

Const framed this against a specific number: 80% of capital raised by San Francisco AI startups goes to compute. On Bittensor, subnets would fund that compute internally. The money stays in the ecosystem.

He also announced canonical bridges to measure external demand. If someone on another chain buys a subnet’s usage credits, that flow becomes a metric for directing emissions. Subnets that attract real customers from outside Bittensor get rewarded.

Why it matters: Bittensor has been building infrastructure in isolation for years. Compute, storage, inference, training. Each piece works. But the pieces haven’t been composable. Gamma tokens make that infrastructure layer interoperable and programmatically purchasable. An AI agent doesn’t need a GitHub account or a Discord invite. It needs a wallet and gamma tokens. The gap between a subnet that produces benchmarks and a subnet that produces revenue just got easier to close.

The takeaway: Watch for the gamma token specification. The subnets that become issuers (Lium, Targon, Engy, Chutes, Hippius, Affine) are selling picks and shovels during a gold rush. Every other subnet becomes a potential customer. If you’re evaluating which subnets to back, the infrastructure providers that issue gamma tokens are the ones every other subnet depends on.

💹 SubConnect’s Bittensor Revenue Index: the first real census

SubConnect published the first Bittensor Revenue Index (Volume I) in late August, and by September it had become the defining frame for ecosystem discussion. The report estimates $28 to $35 million in annual recurring revenue across 24 subnets that are generating commercial income. 14 of those subnets are running active token buyback programs, using real revenue to purchase their own alpha tokens on the open market.

The report is an estimate, not an audit. The revenue figures are modeled based on publicly available data, API usage metrics, and self-reported numbers. SubConnect acknowledged the limitations, but the framing is what shifted the conversation.

RVCrypto, who collaborated on the research, said he expects Bittensor subnets to collectively reach $100 million in revenue.

Why it matters: SubConnect’s Index is the first serious attempt to ground the external revenue debate in numbers. $28 to $35 million in ARR is modest compared to a single mid-tier SaaS company, but it’s real money flowing through decentralized AI infrastructure. And 14 subnets running buybacks means that revenue is creating direct demand for subnet tokens, which is the economic flywheel the dTAO model was designed to produce.

The deeper question is this: can these subnets become real businesses? Generating revenue is step one. Building a sustainable business with margins, customer retention, and competitive moats is something else entirely.

The takeaway: Use the Revenue Index as a screening tool: the subnets generating real revenue with active buyback programs are the ones worth watching closely. The ones with no revenue and no buyback mechanism are running on emissions alone, and emissions are a shrinking resource post-halving. The next volume of the Index will matter more than this one, because it will show whether revenue is growing or flatlining.

💹 Yuma crosses $200 million in staked assets

On September 14, DCG subsidiary Yuma reported more than $200 million in TAO and subnet tokens staked to its Bittensor validator. That makes Yuma the largest validator on the network by staked value. The company also disclosed that its accelerator has supported 20 subnets.

Business Insider covered the milestone. Yuma’s own year-end roundup framed it as progress toward growing the Bittensor ecosystem through infrastructure investment and subnet acceleration.

Why it matters: $200 million in staked assets from a DCG subsidiary is the most concrete institutional conviction signal in Bittensor’s history. This isn’t a fund allocation or a token purchase for trading. It’s capital committed to securing the network and backing specific subnets through Yuma’s accelerator. When Yuma stakes TAO to its validator and directs yield toward subnet tokens, it is making a portfolio decision about which subnets deserve institutional backing. That shapes emissions, liquidity, and which subnets survive the competitive pressure of the dTAO market.

The counterpoint is that $200 million staked with a single validator also concentrates influence. Bittensor’s governance is designed to be decentralized, and having one entity control that much stake raises questions about influence. The V450 root basket concentration cap (another protocol upgrade) is partly a response to this kind of concentration.

The takeaway: Follow Yuma’s accelerator portfolio. The 20 subnets they’ve backed are the ones with institutional backing, mentorship, and capital. Some will succeed and some won’t, but Yuma’s selections tell you which subnets serious money is betting on. Just remember that institutional backing is a signal, not a guarantee. Quasar (SN24) had institutional backing too, and it collapsed in August when an audit showed the model was plagiarized.

⚙️ OpenRoboto (SN80): the robotics subnet breaks out

OpenRoboto had the kind of month that makes you pay attention to a subnet you’d been scrolling past.

On September 22, the team launched OpenRoboto Shift, a new initiative to collect first-person robotics data on Bittensor. Three days later, they partnered with Axis Robotics to release the Open Axis Benchmark, a standardized evaluation framework for robot models. They also brought an xArm 6 robotic arm online to validate winning simulation models on physical hardware, and they switched their community post-training base model to LingBot-VLA 2.0.

Why it matters: Bittensor has needed a successful robotics subnet for a long time. Most subnets compete on text, code, or compute because those are the easiest things to verify on-chain. Robotics is harder. You can’t just check the output of a language model and score it. You need to evaluate whether a robot actually completes a physical task, which means bridging from simulation to real hardware.

OpenRoboto is building that bridge. The xArm 6 validation loop means winning models from the subnet’s simulations get tested on a real robot arm. The Open Axis Benchmark gives the ecosystem a shared scoring system. And the Shift data collection initiative addresses the biggest bottleneck in robotics AI: getting enough diverse, real-world training data.

This is the kind of subnet that makes Bittensor tangible for people who don’t care about tokenomics. You can show someone a robot arm executing a task that a decentralized network trained, and they understand it without needing to know what dTAO means.

The takeaway: Watch the Open Axis Benchmark adoption. If other robotics projects start using it as a standard, SN80 becomes infrastructure for the entire robotics AI space, not just a Bittensor subnet. That is the kind of positioning that attracts researchers and funding from outside the crypto bubble.

📡 Ecosystem tooling: the front door finally opens

September saw a wave of tools launch that make Bittensor accessible to people who don’t read code:

TaoFlows expanded its Bittensor terminal on September 3 with subnet swaps, watchlists, on-chain alerts, and a portfolio dashboard. You can now track your subnet positions, get notified when something moves, and swap between subnet tokens without navigating multiple interfaces.

TaoBubbles launched a visual subnet bubble map on September 17, showing the entire Bittensor subnet economy at a glance. Each subnet is a bubble sized by market metrics, giving you a visual snapshot of where value is concentrated.

SubnetRadar added a new Code column that scores public GitHub repository activity across subnets. You can see which subnets are actually committing code and which ones have gone quiet.

Why it matters: Bittensor’s biggest problem isn’t technology, it’s accessibility. The ecosystem has been built by coders, for coders. If you don’t know how to navigate taostats, parse GitHub commits, or understand emission mechanics, you’re locked out. These tools are the front door. TaoFlows’ portfolio dashboard is UX a retail investor expects. TaoBubbles’ visual map is what you can share with someone who has never heard of Bittensor and they’ll understand it in 10 seconds. SubnetRadar’s Code column separates serious builders from squatters.

The takeaway: If you’ve been waiting for Bittensor to become navigable before diving in, these tools are your entry point. Start with TaoBubbles to see the landscape. Use SubnetRadar to check which subnets are actually building. Track your positions with TaoFlows. You don’t need to be a coder to participate anymore.

Final thoughts

East (Capital) ran this month. The SubConnect Revenue Index and Yuma’s $200M milestone together framed Bittensor as an emerging market with real revenue and institutional capital. That reframing matters more than any single price movement or protocol upgrade, because it changes the type of investor who starts paying attention.

South (Technology) continued its steady march. OpenRoboto shipped more in one month than most subnets ship in a quarter, and they did it in the hardest category on Bittensor: physical-world robotics.

North (Foundation) delivered a major protocol upgade and a cultural milestone. Exploit Summit gives Bittensor a physical presence and a shared calendar event that the ecosystem can rally around.

And West (Adoption), after months of silence, finally showed up. The tooling wave is early-stage and rough around the edges, but it’s real. People who don’t read code can now navigate this ecosystem. That has been the missing piece since I started writing this newsletter.

The SubConnect Index says $28 to $35 million in ARR. RVCrypto says the target is $100 million. The gap between those two numbers is the entire investment thesis for Bittensor subnets right now. If revenue grows, the flywheel spins faster. Buybacks increase. Token prices reflect fundamentals. More builders enter. More revenue.

If it doesn’t grow, the emissions-only subnets start dying off, and the ecosystem contracts to the ones with real customers.

I think it grows. The tools are getting better, the capital is deepening, and the products are shipping. But I’ve been wrong before, and I’ll be wrong again. The Revenue Index gives us a baseline. Volume II will tell us whether we’re climbing or flatlining.

Subscribe now

Until next time.

Cheers,

Brian

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

What if you could copy Bitcoin’s monetary policy (the exact code that created a trillion-dollar asset), and redirect it away from ledger security toward AI work?

Here's what that would look like:

21 million tokens. 4-year halvings. An emission schedule copied line-by-line from Bitcoin's playbook. But instead of paying miners to validate transactions, you pay over 100 specialized AI networks to run inference, fact-check outputs, train models, solve specialized problems.

Think of each AI network (subnet in Bittensor terms) as a specialized startup competing for a share of the daily TAO payout. They've got their own teams, their own miners, their own tokens. And if they don't deliver utility? They get deregistered from the network. It's the most ambitious fork of Bitcoin's monetary ideas I've found.

This is the story of where TAO came from, how it survived crashing to $30 and rocketing to $757, and why a mechanism called dTAO turned every holder into an involuntary venture capitalist in the world's most open AI marketplace.

What TAO Actually Is

So if the Bitcoin parallel covers the supply side, what does TAO actually do?

Every 12 seconds, the network mints 0.5 TAO. That TAO gets distributed to miners and validators across every active subnet as payment for work. A subnet running AI inference gets a share. A subnet fact-checking outputs gets a share. A subnet training models gets a share. The work is real. The payment is TAO.

But what makes TAO different from a mining reward is it’s also the only asset that can route capital between subnets.

If you want to support a subnet, you stake TAO into its pool. If you want to move your capital from one subnet to another, you withdraw TAO from one pool and deposit it into another. TAO is the on-ramp, the off-ramp, and the currency in between.

Think of it as the reserve currency of a decentralized AI nation. The subnets are the businesses. The miners and validators are the workers. And the emission schedule, the thing that prints new TAO every 12 seconds, is the central bank. But this central bank lets the market decide where the money goes.

Origin Story: The Pseudonym, the Chain, and the No-ICO Launch

Bittensor was founded in 2019 by Jacob Robert Steeves and Ala Shaabana. Steeves was a machine learning engineer who’d worked at Google. Shaabana was a computer scientist. They conceived the idea of an AI blockchain and started building.

In a nod to Bitcoin's Satoshi origins, the whitepaper was published under a pseudonym: Yuma Rao.

Its intentionally the same playbook: anonymous author, open-source code, no IPO, no pre-mine, no token sale.

The blockchain itself is built on Substrate, the same framework that powers Polkadot. But Bittensor is an independent Layer 1 chain, not a Polkadot parachain. It doesn't rely on another chain for consensus, security, or governance. Bittensor controls its own upgrade path, its own emissions, and its own destiny

The mainnet launched on January 3, 2021. Every TAO from genesis was earned through network participation. Miners ran nodes, contributed compute, and got paid in TAO. There was no ICO allocation to dump, no founder cliff vesting schedule or any pre-sale tokens sitting on exchanges waiting to be unloaded.

Early institutional capital did show up. Polychain Capital, one of the biggest crypto-native VC funds, was accumulating TAO before the public network launch. They saw a network with Bitcoin’s monetary discipline applied to AI instead of ledger security.

In 2026, both founders stepped down from the Opentensor Foundation. Steeves in February, Shaabana in March. They left to build subnets on the network they created. The founders became competitors within their own ecosystem (although Steeves is still centrally involved in protocol upgrades for now).

Tokenomics 101: The Bitcoin Blueprint

TAO’s tokenomics are deliberately modeled on Bitcoin. If you understand Bitcoin’s supply schedule, you understand TAO’s. The differences are in what the token actually does.

21 million max supply. Identical to Bitcoin.

Emission rate: Initially 1 TAO per block, roughly every 12 seconds. That worked out to about 7,200 new TAO per day, distributed to miners and validators across all active subnets.

4-year halving schedule: When circulating supply hit 10.5 million, daily emissions cut in half. The first halving happened in December 2025, dropping from 7,200 to 3,600 TAO per day. Network inflation went from roughly 25% to 12.5% annually.

Distribution: No founder allocation in the traditional sense. Miners and validators earn TAO by doing work. As of September 2026, roughly half of issued TAO is staked to Root (the base layer of the network). A further share is staked across individual subnets or held in subnet liquidity pools, while the remainder is unstaked or held in other wallet and exchange balances.

Circulating supply right now: roughly 11.3 million TAO of the 21 million cap. We’re past the first halving. Scarcity is tightening.

I wrote a full deep dive on the halving mechanics in The First Bittensor Halving: The End of Easy Mode.

Subscribe now

Three Eras of TAO’s Price History

TAO’s price history breaks into three distinct eras. Each one was driven by a completely different force.

Era 1: Genesis (2021 to 2023) — The Dark Ages

The mainnet launched January 3, 2021. And then…. nothing happened. At least nothing you could see on a price chart.

TAO had no exchange listings for most of this period. It traded over-the-counter or through obscure channels. There was no price discovery and no narrative to play off. The token was essentially invisible to the broader market while the network was live and miners were earning TAO for real work.

First exchange listings arrived in March 2023. TAO opened at around $91.

Then it crashed.

All-time low: roughly $30 in May 2023. This was a classic post-listing dump. The market found Bittensor, got excited for a minute, then the early holders and miners took profits. TAO went from $91 to $30 in two months. Most people who’d just discovered it wrote it off. Probably another altcoin after all.

Era 2: The AI Narrative Boom (2023 to 2024) — From $30 to $757

The AI narrative exploded in late 2023. ChatGPT had been live for a year, every tech company was pivoting to AI, and crypto needed an AI story. Most AI tokens were whitepapers with a logo. Bittensor was one of the few projects with actual infrastructure: a live network with real miners and working subnets.

TAO became the AI crypto trade.

All-time high: approximately $757 in March 2024. A roughly 25x move from the May 2023 low of $30.

Institutional capital hit the big time during this phase. Grayscale launched a Bittensor trust. Bitwise filed for a TAO ETF. DCG created an index fund. The trade was crowded and the narrative was hot.

The rally didn’t hold. By late 2024, the broader crypto market cooled and TAO gave back significant gains. The AI narrative alone couldn’t sustain a $750+ price. The market needed proof that the subnets could actually generate revenue, not just emissions.

Era 3: The Mechanism Era (2025 to Present) — dTAO Changes Everything

February 2025 is the line in the sand. dTAO launched, and TAO’s market structure fundamentally changed. Price was no longer just about TAO scarcity. It became about which subnets deserved capital.

December 2025: the first halving. Emissions cut in half. TAO’s price dipped about 20% amid miner repositioning and market jitters, then stabilized. I covered this in my 2025 End of Year Report Card.

March 2026: TAO rallied to $377 with $900 million in daily trading volume on consecutive days. That volume was the strongest institutional signal we’d seen. The price was ripping while Bitcoin was flat. Something had changed.

July 2026: the V440 Emission Gate shipped. Subnets now have to clear a demand threshold before they receive full emissions. No more free lunch for idle subnets. I did a full three-level breakdown of how this works in Bittensor 101: V440 - The Emission Gate.

September 2026 (today): TAO trades at roughly $300. About 67% below its all-time high. Market cap around $3.5 billion. Circulating supply near 11.3 million of 21 million.

The price doesn’t reflect what’s happened to the mechanism. The mechanism has never been stronger, but the market hasn’t caught up. Or the market doesn’t believe yet.

The dTAO Pivot: What Changed in February 2025

Before February 2025, Bittensor had a centralization problem. A small group of Root validators decided which subnets got emissions. They were the gatekeepers. Politics ruled. If you wanted your subnet to survive, you needed the validators on your side. The system worked, but it concentrated power in too few hands.

dTAO was the solution. The pitch was simple: stop letting a handful of validators play God with emissions. Let the market decide. Turn every TAO holder into a venture capitalist.

How it works: TAO is the reserve currency. Each of the 100+ subnets issues its own alpha token: subnet equity, tradeable against TAO in a liquidity pool. You stake TAO into a subnet’s alpha pool, and that pool acts as an automated market maker.

The more TAO staked into a subnet, the higher its alpha price, and the larger its share of daily TAO emissions. If a subnet delivers real utility, users stake TAO to support it. Alpha price goes up. Emissions flow. If a subnet fails to attract staking, alpha price drops, emissions dry up, and eventually the subnet gets deregistered. The market, not a committee, decides which subnets survive.

Think of alpha tokens the way you’d think about venture investments. TAO is the bank account. Alpha tokens are the bets. Each one has its own halving schedule — roughly every 2 to 2.5 years, separate from TAO’s 4-year cycle — so subnet-level scarcity tightens on its own timeline, independent of the broader network. Some will compound into real value, but most end up deregistered.

dTAO changed the staking dynamics fundamentally. TAO staked to Root (Subnet 0) is now only counted at 18% of its nominal value in validator weight, while alpha stakes are fully weighted. Every TAO holder became an active allocator whether they wanted to be or not.

I put skin in the game. I picked three subnets with real utility, and moved meaningful TAO into their alpha pools. My thesis was simple: the market rewarding quality.

Six months later, I was down 30%, and my alpha trading phase was over before it started.

The subnets were still working, but the capital had moved elsewhere.

What I learned: these markets reward allocation speed, not patience. You can’t hold alpha like regular equity. The full autopsy is in One Year of dTAO: The Market That Took Over Bittensor. But the mechanism was doing exactly what it was designed to do: forcing capital toward whatever’s working today, not what you think should work tomorrow.

The Product Take: TAO Learned to Measure the Right Things

Here’s the question that matters: how do you know if the work being done on your platform is actually good?

Bitcoin answered this clearly. Proof-of-work. You either produce a valid hash or you don’t. The network verifies it in milliseconds with no committee or judgement calls.

Bittensor inherited Bitcoin’s monetary discipline but a harder version of the verification problem. AI work isn’t a hash. You can’t score a weather prediction model or a fact-checking agent with a single function. The output is messy, subjective, and constantly evolving.

The design philosophy that makes this work is simple: define what to verify and how to reward it, then leave the method undefined. Anyone in the world can show up and solve the problem however they can. By not prescribing the method, you expand the global state of potential solutions.

For TAO as a currency, that evolution matters more than any single subnet’s balance sheet. A reserve currency is only as credible as the system that routes it. When the routing mechanism goes from committee politics to market signals to demand thresholds, the currency backing it becomes more trustworthy because the system filters out the noise on its own.

Jacob Steeves compared it to NVIDIA. GPUs were a new type of computing. It took roughly 30 years for them to become the foundational layer for AI. Bittensor, he argued, is inventing another new type of computing (incentive computing) and the trajectory is just as long. The network started with one incentive mechanism in 2021. Now there are 128, with a new one appearing roughly every four days. When GPUs were first invented, they had a couple of applications. Now they’re in every industry on earth.

Which is how you should read the revenue numbers. The SubConnect Bittensor Revenue Index counts roughly 24 revenue-generating subnets today, estimating ecosystem ARR at $28M–$35M with a plausible path to $100M+ by end of 2026. These are subnets earning dollars from real customers, not just recycling TAO emissions.

Bittensor is moving from emissions-funded experimentation to customer-funded businesses. When 24 subnets are generating real external ARR and the projection is 35–40 by year-end, the question becomes “when does revenue become the dominant signal?”

Where TAO Goes From Here

The Bull Case:

The mechanism has never been stronger. dTAO solved allocation. The halving solved supply. The Emission Gate solved quality. Institutional volume is real. TAO bridged to Base through Chainlink, opening the subnet economy to one of the deepest DeFi liquidity pools in crypto. And 24 subnets are generating real external revenue with a credible path to $100M+ by end of 2026.

The Bear Case:

TAO trades at $300, 67% below its all-time high. The revenue is real but early and emissions still subsidize the majority of subnet operating costs. The complexity problem also hasn’t gone away. Ask 10 people to describe what TAO does and you’ll get 17 different answers. The market either hasn’t caught up, or it doesn’t believe yet.

Why I’m Still Here

I’ve been through all three eras. I bought the narrative. I staked into alpha tokens and lost money. I’m still here, not just from faith, but because the product keeps shipping. Each era fixed a real problem the previous one couldn’t. The mechanism today is unrecognizable from the one I started with, and every change has made the system more honest.

What I’m watching

The trajectory, not a single moment. Can the number of revenue-generating subnets scale and can canonical bridges (already existing on EVM, Solana, Base) start measuring external demand as an emissions metric, so the mechanism rewards subnets pulling in real buyers from outside Bittensor? Can the complexity gap narrow before patience runs out?

The Bigger Stakes

Steeves framed the stakes in a way that connected back to Bitcoin. Bitcoin was money outside the state. Bittensor is trying to build intelligence outside the state: AI that doesn’t sit behind a lab’s closed doors. If you build a data center, someone can seize it. If you build a mechanism, it doesn’t have an address.

This is the trajectory TAO is on. A currency backing a new type of computing that defines the outcome, leaves the method open, and lets the world come solve the problem. The mechanism keeps getting more honest and the revenue keeps getting more real.

Every subnet is a business competing for capital. Some are shipping revenue. One just shipped a frontier model. Most won’t survive. The mechanism will kill them without a committee vote.

That’s what you bought when you bought TAO. Not just exposure to AI. Exposure to a sorting mechanism that might actually work.

Subscribe now

If this piece made the picture clearer and you want to keep going, two places to start. 5 terms that unlock 80% of Bittensor conversations is the vocabulary survival guide. It gives you the five words that show up in every Discord thread, governance proposal, and subnet debate, translated into plain English. If you can’t tell a subnet from a validator yet, start there.

If you’re further along and trying to explain Bittensor to someone else, I Fixed My Bittensor Pitch walks through what happened when I tried five different pitches on five different people and what actually worked. Spoiler: the version I practiced in the shower died on contact with a real human at a bar.

Both are free. No paywall. The goal is simple: make this ecosystem legible to people who don’t code.

Until next time.

Cheers,

Brian

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

Lately I’ve been spending less time on price charts, emission schedules, and whale wallet movements. I’ve wanted to spend more time looking at what matters: Bittensor subnets as businesses to be analyzed.

The subnets that survive don’t always start with the right product, they start with the instinct to change it. In the startup world, we call this a pivot. Eric Ries wrote the book on it (The Lean Startup). Marty Cagan built a career teaching product teams how to evaluate whether they’re building the right thing before they run out of runway.

Bittensor subnets are now facing the same pressure. The V440 Emission Gate, which shipped in August, made holding a low-demand slot financially painful. Subnets below the demand bar saw their emissions collapse. The 94 subnets sitting under the bar went from collecting 38.4% of total emission to 12.5%. The top 8 went from 32.8% to 52.7%.

This is an emphatic forcing function. And forcing functions produce pivots.

In this vein, three subnets caught my attention recently, so I’m going to read each one through a Product lens. Two frameworks doing the heavy lifting: Ries’s pivot taxonomy and Cagan’s four risks of product discovery. Plus my own VITALS diagnostic, which I built specifically for subnet evaluation.

How do these stand up as viable businesses? Let’s start with the one that got it right.

Score (SN44): The Textbook Evolution

Score launched as a computer vision subnet for European football clubs. Max Sebti’s team built vision models that could analyze football games for smaller clubs at a fraction of what expensive proprietary systems charged.

Then they expanded. Football became manufacturing. Manufacturing became agriculture. Then, agriculture became gas station security cameras across Europe as they landed contracts with one of the continent’s largest gas station conglomerates.

But they kept going. In August, they launched Score Studio, a platform where developers can build vision apps through “vision vibe coding,” using distilled models shrunk from 3.4GB to roughly 19MB that run on CPUs, not GPUs. They also announced Native Vision AI plugins for vibe coding tools, starting with Cursor AI.

And then, on September 1st, Score posted an announcement: “We Will Build Worlds.”

Score is taking Subnet 44 from computer vision into world models. Asking customers to describe a problem and have miners build a production vision system for it.

The evolution chain moves from computer vision into Score Studio and now world models. Each one a broader application of the same core capability: making cameras understand what they see, on cheap hardware, at the edge.

In Ries’s framework, Score executed a zoom-out pivot. The original product (football analytics) became one application of a broader capability (computer vision on edge devices). That capability became a platform (Score Studio). And now the platform is becoming infrastructure for world models: systems that don’t just detect what’s in a frame, but predict what happens next.

This is the clearest signal of a pivot: each expansion was validated by revenue before the team moved to the next one. Football proved the tech worked, manufacturing proved it generalized, the gas station contracts proved enterprises would pay. Then the platform play came after all of that, not before.

Strong product teams tackle four risks before building: value (will customers buy it), usability (can they figure it out), feasibility (can we build it), and business viability (does the economics work). Score has answered all four with evidence, not theory:

• Value: Enterprise clients are paying. Real contracts, real revenue.

• Usability: Vision vibe coding, downloadable codecs and Cursor AI plugins. They’ve systematically removed friction for developers.

• Feasibility: Models running on CPUs at 50MB is a genuine technical moat, and they shipped it.

• Business viability: Multiple revenue streams. Yuma backing. CPU inference means better unit economics than GPU-dependent competitors.

Yuma’s Evan Malanga said it plainly: the growth Score’s team accomplished in 12 months is “insane,” going from “something for football clubs in the UK” to manufacturing, fuel, and agriculture.

Running Score through my VITALS diagnostic, they score 6 out of 6. This is the benchmark.

Score never panicked because they never needed to. They kept finding new cameras to make intelligent. The pivot was gradual, revenue-validated at each step, and the core capability never changed.

Subscribe now

Zipcode → Instant (SN46): The Forced Pivot

Subnet 46 started life as RESI, a real estate data lake. Founder Seby Rubino forked Subnet 13 to build a residential property database, scraping Zillow-style data. The team evolved into Zipcode, expanding to real estate price prediction and credit tooling.

Seby had a philosophy: subnets are not one-trick ponies. He framed subnets as processes. “Think about it like a prompt,” he said. “I would like to know the price of this house. Here is an inspection report.”

In early August 2026, the team pivoted the entire subnet to Instant: lightning-fast private inference. The first customer signed was DittoBench (Ditto AI, Subnet 118). The pivot coincided directly with the V440 Emission Gate upgrade.

In Ries’s taxonomy, this is a customer need pivot combined with a technology pivot. The team still serves developers who need AI compute, but the specific problem shifted from “price this house” to “run my model fast and privately.”

Is the pivot too drastic? On the surface, yes. Real estate to generic inference is a 180-degree turn. The accumulated learning from building a property data layer doesn’t obviously transfer to fast inference.

But the structural pressure made it rational. Real estate price prediction is a niche market with a small TAM and long sales cycles. Under V440’s emission rules, the combination of niche and slow revenue means starvation. The team looked at the math and made a call: the slot is worth more as inference infrastructure, which has a massive TAM and immediate demand.

The validated learning that carries forward here is execution capability. Seby’s team proved they could fork a subnet, build infrastructure fast, iterate through three versions, and manage validator relationships. That matters in a competitive inference market.

But here’s my concern, and it’s a huge value risk: inference is a crowded space. And here, the team can no longer rely on the domain expertise they had with property.

Chutes, Engy and Targon subnets already exist, providing inference at scale. Instant’s wedge must be speed and privacy, but fast inference alone doesn’t answer the question: faster than what? At what cost? Why can’t I use another subnet or a centralized provider?

Running Instant through VITALS at day 30 post-pivot, I get 3 out of 6. One signed customer is a proof point, not a business. Distribution beyond that partner isn’t visible yet. The inference market is competitive, and without a sharper differentiation, Instant risks entering a crowded space with no moat.

Vocence → UMI (SN78): The Ambitious One

Vocence (Subnet 78) was built by Koyuki Nakamori, who also builds Perturb (Subnet 26). Vocence was a voice-intelligence subnet: text-to-speech, voice cloning, custom voice design, music generation. The website pitched it as “Decentralized Voice AI.” A partnership with BabelBit (Subnet 59, real-time speech translation) was a natural fit.

Then Vocence pivoted to UMI (Universal Motion Intelligence), expanding from voice into vision, motion, expression, and language. The first commercial product in development is bitsign: real-time sign language translation between deaf and hearing people. The SN78 alpha token rallied sharply in the days after the announcement.

In Ries’s framework, this is a zoom-out pivot combined with a platform pivot. Voice synthesis was the product. Now voice is one feature inside a multimodal intelligence platform. And Vocence went from being an application (generate speech) to positioning as a platform (orchestrate specialized subnets into a unified product).

This is the most intellectually interesting pivot of the three. Here’s why.

Voice AI to Universal Motion Intelligence sounds like a massive leap. But look closer. The underlying customer problem stayed the same: real-time human communication translation. BabelBit translates speech-to-speech in real time. bitsign translates sign-to-speech in real time. This is the same product category (communication translation), but a different modality (motion instead of audio).

But feasibility risk is the existential threat here. Orchestrating multiple subnets (vision for sign capture, voice for speech output, language for translation) into a single real-time product is genuinely hard engineering. bitsign needs to actually work, not just sound good in a pitch.

Running UMI through VITALS at this early stage, I get 3 out of 6. The value signal is muddied: “Universal Motion Intelligence” is vague. “Real-time sign language translation” is concrete. They need to lead with bitsign, not UMI, as the platform story comes after the product works.

On the risk assessment:

• Value: Strong if they lead with bitsign. Real-time sign language translation is a genuine unmet need. Weak if they lead with UMI as a concept.

• Usability: Two-way translation between signing and speech is a hard UX problem. A camera capturing signs and a speaker outputting speech needs to feel seamless.

• Feasibility: High risk. Multi-subnet orchestration in real-time is the hardest engineering challenge of the three pivots.

• Business viability: Unclear. Accessibility markets can be enterprise-grade (government, healthcare, education compliance) but sales cycles are long.

The lesson: the best pivots keep the customer problem constant and change the solution. Vocence’s pivot works on paper because the user need (communication translation) didn’t change. But a coherent thesis isn’t a working product. Ship a working bitsign demo in 90 days and this could be the most promising pivot on the network. Don’t ship it, and UMI is a vision statement with nothing inside.

What This Tells Us About Bittensor Subnets as Products

Gradual evolution beats dramatic pivots when the core capability is strong. Score never needed a 180 because its core capability (distilled vision models on cheap hardware) kept finding new markets. Each expansion was backed by revenue. This is the Lean Startup Build-Measure-Learn loop executed properly.

Pivots forced by structural pressure are rational but inherit new market risk. Zipcode’s pivot to Instant was the right call given V440. But entering a crowded inference market without a sharp differentiation is the classic zoom-out trap: bigger market, no advantage.

The customer problem should stay constant when the solution changes. Vocence’s pivot from voice to multimodal communication works because the user need (real-time translation between humans) didn’t change. Compare this to Zipcode, where the customer problem changed entirely (property valuation to inference compute). That’s harder to pull off because you’re starting from zero on customer relationships.

The three case studies here are early examples. But they reveal a pattern that will accelerate:

1. Subnets with proven revenue and expanding markets (Score) don’t need to pivot. They just keep growing.

2. Subnets with niche markets that can’t clear the gate (Zipcode/real estate) are forced to pivot or die. The pivot direction matters more than the pivot speed.

3. Subnets with strong technical capability but unclear product-market fit (Vocence) have the option to zoom out into a broader platform play, but the execution risk goes up exponentially with the scope of the ambition.

From Runway to Revenue

The V440 gate is essentially Bittensor’s version of a startup running out of runway. When the money runs out, you either find product-market fit or you pivot. The gate just made the runway visible.

But emissions buy time, not a business.

Canonical Labs published research in May showing that Bittensor’s inference subnets run at 22:1 to 40:1 emissions-to-revenue ratios. For every dollar of real revenue, the network subsidizes $22 to $40 in token emissions. That’s an accelerator model that will end at some point. The goal is to graduate from subsidy to self-sustaining revenue.

All three subnets here are responding to the right pressure. But responding to pressure isn’t the same as building a business.

We need to start evaluating Bittensor subnets the way we’d evaluate any startup:

1. Product-market fit.

2. Unit economics.

3. Revenue that covers costs.

4. A moat that compounds over time.

The incentive layer is the most powerful thing Bittensor built. It pulls talent and compute into the network that no centralized company can match. But an incentive layer that subsidizes work nobody pays for is just a cost center.

For Bittensor to prosper long-term, the incentive layer needs to become productive. Emissions should fund the transition from experiment to revenue. When they’re funding experiments indefinitely, the ecosystem becomes a perpetual incubator.

Score demonstrates what the path looks like when you get it right. Most subnets are still early in that journey. V440 made the economics urgent, but ultimately, the question every subnet has to answer is the same one every startup faces: can you build something customers will pay for before you run out of runway?

Subscribe now

Until next time.

Cheers,

Brian

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

An audit had just found that a subnet called Quasar (SN24) was running a model 98% identical to an existing open-source model, and its token had crashed 75%.

Two compute subnets, Chutes and Engy, were publicly fighting over inference pricing with benchmark results being published on X.

A falling-out between a subnet builder and his former collaborator had escalated into a public Cold Key custody dispute.

My notes file had “conflict” written as the guaranteed headline in my head.

Root Reborn went live on August 3, the same week all of this was kicking off. The biggest structural upgrade since dTAO. Then TAO bridged to Base on August 20. Then Const announced a 110-billion-parameter permissionless training run on August 24. By the time I sat down to write this, the conflict stories felt a little smaller.

The month had plenty of drama, and two of my five stories are genuinely about subnets going at each other. But the bigger signal turned out to be about who was shipping and at what scale.

Foundation (North) and Technology (South) both had strong months. South edged ahead because the technology shipped at frontier scale while North delivered a culmination rather than a new direction. The baton is passing, so let me show you what happened.

How to read this compass

I’m organizing these updates by direction: North, East, South, West to give you a repeatable mental model for navigating Bittensor’s complexity.

North: Foundation 🏛️ (governance, protocol, core team, security events) The structural layer. Changes here determine who controls the network and how decisions get made.

East: Capital 💹 (markets, volume, liquidity, integrations) The economic layer. This is where price meets liquidity and institutional conviction shows up in the data.

South: Technology ⚙️ (shipped features, benchmarks, infrastructure, ecosystem health) The execution layer. What actually works, what got built and where’s the technical proof.

West: Adoption 📡 (narrative, mainstream validation, external voices) The perception layer. This is how the outside world sees Bittensor, from industry titans to normie investors.

Each month, some directions dominate the signal. Others stay quiet. This month, North and South both moved with real force, East made a single important play, and West stayed silent. Let’s review what happened.

🏛️ Root Reborn goes live

On August 3, Bittensor shipped v4.4.1. The Root Reborn upgrade.

I covered the proposal in June when Const unveiled it, and I tracked the debate through July when Yuma published their moral hazard warning. The core problem was simple: root stakers automatically liquidated subnet alpha tokens, creating persistent structural sell pressure across the entire subnet market. Every time emissions flowed, someone somewhere was dumping.

Root Reborn replaces that mechanism with validator-linked subnet baskets. Instead of blindly converting root yield to TAO and selling, validators now direct their yield into subnet tokens they believe are productive. Messari confirmed it as “a major redesign of how root subnet inflation is managed.”

Why it matters: For the entire post-dTAO era, subnet tokens have operated under an artificial price ceiling. Root yield auto-dumped, suppressing the native economies of every productive subnet regardless of how well they performed. Removing that ceiling changes the economic incentives for validators from passive gatekeepers to active ecosystem asset managers. The sell pressure that has dragged subnet tokens down since launch is structurally gone.

The takeaway: Watch subnet alpha token prices over the next 4 to 8 weeks. If Root Reborn works as designed, the tokens of productive subnets should start reflecting their actual fundamentals instead of fighting a constant drag. The subnets that benefit first are the ones validators already trust.

Subscribe now

🏛️ The Quasar rug (SN24)

In early August, SN24 (Quasar) was hit with public allegations that its model and training claims were plagiarised, and that the subnet was being run by people with a history of scams. A reproducible audit notebook showed the model was overwhelmingly inherited weights with only a tiny fraction of new training.

The token crashed around 75% in a few hours as the story spread. Community channels lit up with over 1,500 messages in roughly 25 hours during the peak.

The timing made it worse. On July 31, just days before the audit dropped, the Quasar team appeared on Novelty Search (the weekly Bittensor community call hosted by Const / OpenTensor) claiming they’d trained a 120B SOTA model with 5M context. They positioned it as a breakthrough: decentralized long-context models that could compete with centralized labs.

The governance response was swift following the discovery. By the following week, the old team was officially ousted, the owner was banned from the server, mining was paused, and a cold-key swap was announced under new ownership. On-chain, Quasar’s emissions showed the largest single drop in the v440 gate reshuffle, down roughly 86% to 251k.

Why it matters: The system did what it was supposed to do. An open audit exposed a copied model, and the token price reflected that reality within hours. The real problem runs deeper: Quasar’s legitimacy came from who backed it, not what it built. Investors trusted DSV’s capital and Const’s reputation instead of verifying the technical claims. When a subnet carrying those endorsements collapses under scrutiny, it spreads reputational damage to every other project in their portfolio.

The takeaway: Decentralized AI only has one real advantage over a lab press release: someone else can open the weights. If the checkpoint, step count, lineage, and miner contribution cannot survive that check, the subnet is fundraising in disguise.

⚙️ Teutonic-II: 110B permissionless training

On August 24, Const announced Teutonic-II-110B: a frontier-scale, fully permissionless training run of a 110-billion-parameter LLM on Bittensor Subnet 3.

This is a 10x scale increase from their previous run. In March, Templar made history with a 72B decentralized training run that I covered in my March issue. Teutonic-II is 50% larger than that, and it’s running on a subnet that holders took back through the Conviction mechanism I covered in May. SN3 was one of the three subnets Covenant abandoned in April. The community reclaimed it, onboarded it through Conviction, and now it’s running the largest decentralized pretraining run in history.

Why it matters: This is the execution layer firing at full capacity. In July, I wrote about Chutes training a 20B model for under $10 an hour and asked whether the cost advantage over centralized cloud was still theoretical. Teutonic-II answers that question at a completely different scale. A 110-billion-parameter model trained over the internet, permissionlessly, on a subnet reclaimed by its community through governance. The Conviction mechanism, the Emission Gate, the Root Reborn upgrade, all of it was building toward this kind of proof.

The takeaway: Watch the training progress on SN3 over the coming months. If Teutonic-II completes successfully, it changes the conversation about what decentralized AI can actually do. The leap to frontier infrastructure narrows with every parameter count increase. 110B is getting close to the territory where people outside the Discord have to pay attention.

⚙️ Chutes (SN64) vs Engy (SN53): the inference war

Two inference subnets went to war in August, and the entire ecosystem got to watch.

Engy (SN53) launched the full 2.8 trillion-parameter Kimi K3 model at what they described as “the lowest price on the market.” They did it on gaming cards positioned near cheap energy sources, using TOPLOC cryptographic proofs to verify that miners are actually running the committed models. Engy’s pitch is verified inference at a price no one else can match.

Chutes (SN64) is the incumbent. The largest compute subnet on Bittensor by market cap, a serverless AI compute platform that lets you deploy any open-source model. In June, Chutes moved to Trusted Execution Environments and brought Blackwell B200s online. In July, Chutes trained a 20B model for under $10 an hour. Chutes has scale, infrastructure, and $43M in Q1 real AI revenue.

The competitive dynamic broke into public view when James Altucher, who runs BlueTAO.ai, published head-to-head benchmark comparisons of every AI model hosted by both subnets. He tested them across multiple use cases for his AI chat and coding application. The results were published openly for anyone to check. Coding went to Kimi K3 hosted by Engy. The full benchmark results and test methodology are available online.

The subnet founders engaged in a public back-and-forth, with supporters from both sides weighing in across Discord and X.

Then on August 25, Chutes fired back. They proposed a Conviction Lock mechanism for federated AI compute access. If adopted, compute buyers would need to lock stake to access Chutes’ federated compute network. It creates a new demand sink for SN64’s alpha token and makes it economically costly for customers to switch to a competitor like Engy on price alone.

Why it matters: This is the dTAO market working exactly as designed. Two subnets competing on real product metrics: price, model quality, verifiability. Engy’s edge is cryptographic verification and cheap consumer hardware near cheap energy. Chutes’ edge is scale, serverless architecture, and now lock-in through conviction staking. They’re competing on different axes, which is healthy for the ecosystem.

The takeaway: Track the Conviction Lock proposal on SN64. If it passes, Chutes builds a moat that price competition alone can’t breach. Engy’s response will tell you whether they compete on price, technology, or their own lock-in mechanism. The inference war is the most visible subnet competition in the ecosystem right now, and the winner becomes the default compute layer for decentralized AI.

💹 TAO arrives on Base via Chainlink CCIP

On August 20, TAO bridged to Base through Chainlink’s Cross-Chain Interoperability Protocol. ForeverMoney went live on August 21, letting native TAO move cross-chain. The integration connects Bittensor assets to Aerodrome, Base’s primary liquidity venue, and opens 16 subnets to DeFi liquidity on one of the most active chains in crypto.

Why it matters: Complex on-chain AMM pools, custom wallets, and fragmented liquidity have historically locked out traditional and retail capital from Bittensor’s subnet economy. Bridging to Base through Chainlink’s infrastructure gives TAO access to one of the deepest DeFi liquidity pools in crypto. Aerodrome is where Base’s capital sits and every cross-chain integration like this expands the pipe through which capital can flow into and out of the Bittensor ecosystem without forcing investors to learn subnet-specific mechanics.

The takeaway: Watch TAO’s liquidity depth on Aerodrome over the coming weeks. If the Base integration draws meaningful volume, other subnet alpha tokens could follow the same CCIP bridge path. The first subnets to bridge get the first wave of Base liquidity.

Final thoughts

North (Foundation) and South (Technology) both shone in August. The difference is in the delivery.

North’s stories were culminations. Root Reborn was proposed in June, debated through July, shipped in August. The Quasar audit tested mechanisms that already existed. Both are the payoff of months of structural work and read like the final pages of a chapter.

South’s stories were escalations. Teutonic-II multiplied their previous training record by 10. The inference war produced public benchmarks, strategic counter-moves, and a new governance proposal from the incumbent. This chapter is still being written, and the next page is bigger than the last.

But west (Adoption), still, stays silent. The outside world hasn’t caught up to what’s being built inside this ecosystem yet. The technology is shipping at frontier scale, the governance is mature enough to catch bad actors, and the capital infrastructure is expanding. None of that is priced in.

When West wakes up, it catches up fast. We all just need patience to stay the course.

Until next time.

Subscribe now

Cheers,

Brian

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

Every couple of issues, someone replies with a version of the same message:

“Great, now I understood about 60% of this. The other 40% I had to Google.”

Or:

“I’ve been following Bittensor for months and this is the first time X actually made sense.”

The 60% they got was the story and the metaphor. The 40% missed was always the vocabulary and the mechanism, where I name a concept and assume the reader already has a mental model for it.

They don’t. Most of the content ecosystem doesn’t either.

I see it at home with my wife. She can explain now Bittensor to a friend with the mall metaphor but if you ask her how the payment actually works she’s already stuck. That gap is where people fall off as the content jumps from metaphor straight to governance proposal with nothing in between.

I tried fixing it in pieces. The Bittensor Pitch, 5 Terms That Unlock 80% of Bittensor Conversations, Bittensor Terms Decoded (the living glossary with Textbook and Bar Talk versions). People bookmark the glossary, but in reality it’s only a dictionary. You look something up when you already know you need it, it doesn’t show why a concept matters this week.

Here's how we close that gap: a monthly deep-dive series where we tackle one concept at a time across three levels of explanation, with every topic pulled from something that actually changed subnet economics that month.

• Level 1 (Bar Talk): If you can’t explain a concept to a friend over a pint without pulling out a glossary, you don’t actually understand it. Start here to get the core mental model.

• Level 2 (Read the Signals): The mechanics and the numbers. How to understand what the live metrics actually mean for your portfolio.

• Level 3 (Follow the Incentives): The product reality and unit economics. How the math alters incentive structures and accelerates the survival of the fittest across the network.

First up is the biggest change to emissions since dTAO: the Emission Gate that shipped with the v440 upgrade.

You can already see the new meta. Subnet 46 spent most of its life as RESI (then Zipcode): real-estate price prediction and appraisal models. In early August the team pivoted the entire subnet to Instant Inference: lightning-fast private inference.

That was not a random rebrand. Once the free lunch for low-demand slots disappeared, the calculus for every subnet owner changed. Holding a slot that could not clear the demand bar became expensive in a way it never was before. Teams that moved fastest started asking: what can this slot actually earn under the new rules, and is the current product the highest-value use of it?

Instant Inference is the first high-profile answer. The pivot is so drastic it deserves a separate conversation, but it’s an early example showing the pressure is now structural. Emission is still price-based, it just has to clear a moving bar first.

V440 - The Emission Gate Explained

Level 1 (Bar Talk): The End of the Free Lunch

Imagine the Mall of Intelligence.

Every day the mall prints a fixed pile of cash - the daily payday pot - and hands it out to the stores. That pile is emission. For a long time the rule was simple: every store got a slice based on how popular its location looked, even if the shutters were down and the lights were off. A boarded-up unit still collected rent money just for existing.

Quiet stores were diluting the payday for the ones actually serving customers. The cost of opening a new store stayed high because people treated empty slots as passive income machines.

Then the mall changed the rules.

They put a gate at the entrance to the cash room. The gate looks at real foot traffic. Stores that are busy keep almost everything they earned under the old system, and they pick up a bigger share of what the quiet ones lose. Stores that are empty still get a tiny trickle, but it is so small you can’t live on it. The gate itself moves every hour or so based on where people are actually shopping, so a store that was quiet last week can climb back if customers return.

That’s the Emission Gate in one picture.

What most people will feel:

• The busy stores get richer.

• The quiet stores get poorer.

• Parking a store just in case stops making financial sense.

• Opening a brand-new store should eventually get cheaper, because the free lunch that used to prop up empty leases is gone.

If you are holding the token of a quiet store, the math just changed. The network is no longer paying you simply for occupying space.

Subscribe now

Level 2 (Read the Signals): The New Math for Subnet Survival

Since v440 went live the room has been asking the same questions on repeat:

Where is the bar right now?
Did my subnet just get gate-smacked?
Is the line still moving?
Should I be worried about deregistration?

Here’s how to answer them yourself.

The Emission Gate does not change the fact that emission is still driven by demand - the de-manipulated moving price, adjusted for miner burn. What it changes is how that demand turns into real TAO.

Every tempo (roughly 72 minutes) the network sorts every emission-enabled subnet by its current demand share. It walks down the list until the subnets above a certain point are carrying a set fraction of total demand. That crossing point is the bar (thetaθ). Everything then gets passed through a smooth threshold. The defaults that shipped put the bar roughly around rank 30-35 on the July distribution.

What that feels like in practice:

Comfortably above the bar: you keep almost all of your old proportional share, plus you pick up a slice of what the tail just lost.

Sitting right on the bar: you keep roughly half of what the old linear rule would have given you.

Below the bar: emission falls away quickly. Just under the line you might still keep 15-30% of your linear share. Deep in the tail you’re looking at 1-5%. The residual never hits absolute zero, but it is deliberately too small to make parking a slot economically rational.

The July snapshot numbers that still get quoted are useful anchors:

• The 94 subnets that sat below the bar went from collecting 38.4% of total emission to 12.5%.

• The top 8 went from 32.8% to 52.7%.

• The rank at which the network has paid out 80% of emissions moved from 64 down to 23.

• Effective subnet count collapsed from roughly 50 to roughly 22.

No subnet is hard-zeroed by the gate itself. That residual drip is intentional. It removes the incentive to game a hard cutoff while still making idle slots worthless.

For anyone holding alpha or watching specific subnets, growth near the bar is the highest-ROI move on the network right now. A modest increase in demand when you’re sitting just under or just over the line produces a disproportionately larger increase in emission once the gate is applied and the freed-up share is redistributed. Sliding further into the tail hurts more than the old linear model ever did.

A few practical checks:

1. Look at your subnets current demand share (moving price * (1 - miner burn)) relative to the live bar.
2. The bar is recalculated every tempo from the same manipulation-resistant prices that already drive emission. A single block of wash trading does almost nothing.
3. A sudden drop in emission after v440 does not automatically equal deregistration risk. Deregistration still cares about the lowest non-immune subnets when new ones register. The gate changes how much TAO a low-demand slot earns; it does not rewrite the immunity or pruning rules by itself.
4. Miner burn still matters. High burn reduces the demand signal that feeds the gate, so a high-burn subnet starts from a weaker position.

Level 3 (Follow the Incentives): Evaluating Subnet Viability Under v440

Lets strip away the analogies. If you are building on Bittensor or analyzing its tokenomics at a granular level, v440 isn’t just a fairness patch, it’s a fundamental shift in how subnet economics work and how you should model long-term viability.

Before the Emission Gate, the distribution curve for TAO was relatively forgiving. Even if a subnet was hovering at the bottom of the Root Network’s priority list, Yuma Consensus still trickled some emissions down to its miners and validators. It was a gentle slope.

v440 takes that gentle slope and introduces a cliff.

From a systems perspective, the Emission Gate is a step-function. It establishes a hard threshold (a minimum proportion of aggregate validator weight) that a subnet must achieve. If the subnet achieves 99% of that threshold, its emissions aren’t slightly reduced; they are effectively zero. Its a binary state: you are either economically viable, or you are starved of oxygen.

The Builder’s Dilemma: Subnets as Products

As someone who looks at this ecosystem through a product lens, this is where it gets fascinating. Pre-v440, you could launch a subnet as an experiment or a neat feature and let the trickle of emissions fund your iteration.

A post-v440 subnet cannot rely on technical novelty or a marginal use case alone. If we run this through an opportunity assessment, the stakes have completely changed:

1. Exactly what problem will this solve? (The Value Prop): The value must be strong enough to generate sustained demand. A specialised subnet can succeed, but only if it owns a valuable niche and has strong performance differentiation or becomes embedded in downstream products.

2. For whom do we solve that problem? (The Customer): Who creates demand for the subnet, who consumes its output, and who benefits economically from its success? This may include validators seeking reliable and valuable work to score, miners seeking profitable tasks, developers integrating the subnet’s output, enterprises consuming its service, applications using its intelligence or infrastructure, and TAO/alpha stakers allocating capital toward its future value.

3. How will we measure success? (The Metric): Success is no longer measured in gradual adoption. The critical metric is whether the subnet can establish and maintain enough genuine demand to perform strongly against the network’s dynamic emission bar.

Second-Order Effects: The Velocity of Compute

This brings us to the most critical downstream effect of the gate: miner migration. Miners are rational, profit-seeking entities. If a subnet falls below the Emission Gate and emissions drop to zero, the miners aren’t going to stick around for a turnaround story.

They will immediately repurpose their GPUs and migrate to subnets that are above the threshold.

This creates a ruthless, hyper-efficient market. Failing subnets will lose their compute power almost instantly, accelerating their death spiral. Meanwhile, surviving subnets will see a rapid influx of displaced miners, massively increasing the competition (and ideally, the quality of output) on the networks that actually have product-market fit.

The Emission Gate doesn’t just protect the TAO token; it turns Bittensor into an evolutionary pressure cooker.

What This Means for the Journey

The Emission Gate is Bittensor continuing to do what it has always done when the incentives drift: remove free options and force the network to reward real demand. The mall just got stricter about which stores get the good locations.

Holding a quiet slot is no longer a viable strategy. The residual drip still exists, but it is deliberately too small to live on. That pressure is already visible - the Zipcode teams pivot to Instant on SN46 is one early example of owners looking at the new surface and choosing a direction with a clearer path to sustained demand. More of those decisions will follow.

It also clears the path to expand beyond the current 128 subnets. With the gate in place, adding new slots becomes economically viable because unproductive subnets no longer drain meaningful emissions from the system. Only the subnets that clear the demand threshold get paid.

Where this leaves you depends on how deep you want to go:

If you’re explaining Bittensor to someone new, the gate is your clearest proof point that this network isn’t about hype: it’s about ruthless allocation to real utility.

If you’re tracking your portfolio, the bar is now your signal. Open Taostats once a week and check where your subnet’s demand share sits relative to the current gate threshold. That single habit will make every emissions discussion immediately legible.

If you’re building or evaluating subnet viability, stop treating subnets as token issuers and start auditing them as products. Run the Opportunity Assessment from Level 3 on any subnet you’re watching. Those three questions are: what problem, for whom, measured how. If a subnet can’t answer them with specificity, it won’t survive the gate no matter how much hopium the Discord is pumping.

Next month: We’ll break down another concept that moved the network using the same three-level structure. If there’s something you’re tired of half-understanding, reply and tell me. This series exists because the 40% you Google every issue adds up to the difference between casually following Bittensor and actually understanding what you hold.

Subscribe now

Until next time.

Cheers,

Brian

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

My older son spent nine months refusing to let go of the pool wall. Swimming lessons every Saturday. He’d kick, he’d blow bubbles, he’d practice the arm motions on dry land in the living room. But the moment his instructor said “let go,” and he has had three different instructors, his hand shot back to the tiles like a magnet.

Then one Saturday, he just swam. To my disbelief, with no fanfare or countdown, he let go and went. Four meters, face in the water, doggy-paddling like he’d been doing it his whole life. His instructor didn’t even cheer. He just nodded like he knew it was coming.

All that groundwork, invisible to anyone watching from the deck. And then one day, the proof. July felt like that for Bittensor.

For four months, the compass pointed North (Foundation). Governance fixes, emissions overhauls, decentralization roadmaps. The kind of progress that matters but doesn’t make headlines outside the Discord. In May, I wrote that South (Technology) was quiet, waiting for summer. In June, I asked whether the architecture could actually handle scale.

July answered with some real achievements. I’ve filtered them down to two for this piece.

Even Jensen Huang is out there writing open letters about why open source AI matters. The wind is at our backs. Bittensor made its own this month.

How to Read This Compass

I’m organizing these updates by direction: North, East, South, West to give you a repeatable mental model for navigating Bittensor’s complexity.

North: Foundation 🏛️ (governance, protocol, core team, security events) The structural layer. Changes here determine who controls the network and how decisions get made.

East: Capital 💹 (markets, volume, liquidity, integrations) The economic layer. This is where price meets liquidity and institutional conviction shows up in the data.

South: Technology ⚙️ (shipped features, benchmarks, infrastructure, ecosystem health) The execution layer. What actually works, what got built and where’s the technical proof.

West: Adoption 📡 (narrative, mainstream validation, external voices) The perception layer. This is how the outside world sees Bittensor, from industry titans to normie investors.

Each month, some directions dominate the signal. Others stay quiet. This month, three directions moved. One fired. Let’s review what happened.

🏛️ v440: The Emission Gate Opens

On July 28, Bittensor shipped v440. The “Emission Gate.” It introduces a demand threshold for subnet emissions. Idle subnets that can’t show real demand now see their base return drop to near-zero.

The protocol team modeled the gate against a July 2026 mainnet snapshot of all 128 subnets, treating every subnet as emission-enabled. The math is simple: subnets without genuine usage face a dramatic emission haircut. Productive subnets see their rewards increase.

This is the final piece of an emissions puzzle I’ve been tracking since May when the reward logic was rewritten. Then, June purged 57 dead subnets with a hard cutoff. Now v440 introduces market mechanics, letting demand naturally starve out idle subnets instead of manually blocking them.

Why It Matters: The network is now systematically engineering capital toward subnets that earn real revenue. June was about blocking bad actors. July is about making idle subnets structurally unviable over time.

The Takeaway: The v440 effects will show up in August emission data. Watch which subnets gain share and which lose it. The list of gainers is your smart money signal.

🏛️Two Subnet Owner Wallets Drained in Social Engineering Attack

On July 13, two Bittensor subnet owner wallets were compromised on the same day. ORO (SN15) lost 147,000 alpha tokens, drained and sold over roughly 10 hours. Connito AI (SN102) lost 2,500 TAO from their owner wallet.

ORO attributed the attack to a North Korean state-backed threat group. A team member clicked a fake Zoom/Teams meeting link and the attacker drained the owner wallet.

The breach was thankfully contained to the owner wallet, with validator keys, user data, and the subnet itself untouched. Both subnets stayed operational.

Three days earlier, Bittensor shipped v437 with air-gapped signing, but the team hadn’t migrated yet.

ORO published a full post-mortem on July 21, completed a coldkey swap, and moved owner keys to multisig hardware protection. Crucible Labs noted that both teams “chose transparency over silence.”

Why It Matters: Bittensor can ship air-gapped signing, miner collateral, and emission gates, but none of that stops a team member from clicking a meeting link that looks real enough in the moment. The response also matters: ORO published a detailed post-mortem within eight days, while Connito openly acknowledged the incident. Its the type of the culture you want when something breaks, write the report yourself and don’t wait for a press release.

The Takeaway: If you’re a subnet team, move your owner keys to hardware-backed multisig today. The tools now exist with v437. If you’re an investor, start asking subnet teams about their key management the same way you’d ask about revenue. “Who holds the keys and how?” is the new due diligence question.

Share

💹 Kraken Builds a Subnet Token Pipeline and Lists SN8

On July 3, Kraken announced they had completed native integration of Bittensor’s dTAO and will begin listing subnet tokens in the near future. Seven Bittensor subnet alpha tokens were added to its listings roadmap.

On July 16, Kraken listed Vanta Network (SN8) for spot trading. The first Bittensor subnet token to get a dedicated trading pair on a Tier-1 exchange.

Vanta Network operates as fintech infrastructure within Bittensor, enabling financial services across subnets. After listing, SN8 found its floor at $5.14 on July 28. This is price discovery backed by Kraken’s order book depth, a different animal than the AMM pools subnet tokens typically trade on.

Why It Matters: Retail investors can now buy subnet tokens the same way they buy SOL or ETH. The dTAO market graduates from DeFi natives to mainstream crypto traders. Every listing brings new liquidity, new price discovery, and new attention to subnets that previously lived in the shadows of TAO itself.

The Takeaway: Track which of Kraken’s 7 roadmap tokens list next. Each listing is a new price discovery event for a subnet economy. The subnets that list first get the first wave of retail liquidity, and that order matters.

💹Synth (SN50) Posts Profitable Q2, Announces $100K Buyback

Synth, the financial intelligence subnet co-founded by Mode Network, published its “Synth Reborn: Q2 Alpha Update” with two milestones: Q2 was profitable, and they’re committing $100,000 to an SN50 alpha token buyback over 8 weeks, executed weekly.

This is the subnet I spotlighted in Machine Allocators Don’t Want Your Best Guess. Hundreds of competing AI models submit 100 simulated price paths every 5 minutes. Validators score them on how close their probability cones matched reality. The output is a live API that financial traders, DeFi protocols, and AI agents can query for volatility data.

The buyback comes with what Opentensor’s ecosystem highlights called “a renewed focus on turning revenue into value capture.”

Why It Matters: Every subnet talks about value accrual. Synth is the first financial subnet to put $100K of real revenue behind that talk. If the buyback is transparent, it gives the market a measurable signal of subnet revenue health.

The Takeaway: Watch SN50’s alpha price response over the 8-week window. If the buyback stabilizes or lifts the price, other revenue-generating subnets will copy the playbook. The first one to prove the model becomes the template for the rest.

⚙️ Chutes Trains a 20B Model for Under $10 an Hour

Chutes pre-trained a 20-billion parameter AI model for under $10 per hour. Without a dedicated cluster. The achievement used Chutes’ “Parallax” approach, coordinating rented GPUs across distributed hardware to train a model that would normally require a centralized server farm.

This is the same Chutes that generated $43M in Q1 2026 real AI revenue and moved to fully Trusted Execution Environments in June. July is the payoff.

Why It Matters: If you can train a 20B model on distributed rented GPUs for under $10 an hour, the cost advantage over centralized cloud is no longer theoretical. It’s a published number you can compare against any provider’s pricing page.

The Takeaway: Watch whether other subnets replicate the Parallax approach to distributed training. If training at this cost becomes repeatable across multiple subnets, centralized cloud providers lose their last structural argument.

⚙️ Targon (SN4) Launches TargonOS: Permissionless VM Rentals with Root Access

Targon shipped TargonOS to production. The first permissionless compute network offering full virtual machine rentals, with root access from day one, hardware-attested isolation, and end-to-end encrypted workloads.

Most decentralized compute platforms offer sandboxed environments with limits. TargonOS gives you everything: root access means you can install whatever you want, configure it however you need, and run whatever workload you have. This is the type of feature set enterprises demand before migrating off AWS.

At $3.29 an hour for an H200, the cost conversation against centralized cloud gets very uncomfortable for the incumbents.

Why It Matters: Enterprises don’t deploy production workloads in sandboxes. They need full machines with verifiable isolation and encrypted data paths. TargonOS delivers that feature set at a price that makes the migration conversation with AWS very short.

The Takeaway: Watch TargonOS adoption metrics: VM rental volume, unique customers, enterprise signups. If enterprises actually rent H200s at $3.29 an hour, Bittensor compute transitions from competing to winning.

Final Thoughts

For four months, the compass pointed North. Governance fixes, emissions overhauls, security patches, critical infrastructure. The kind of work that keeps a network alive but doesn't make headlines outside the Subnet Summer Telegram group.

Then July delivered what all that foundation work was building toward. It’s coordinating intelligence at price points that make centralized alternatives sweat. The technology works and the products are shipping.

But there is still a disconnect. The price doesn’t reflect any of this yet.

TAO is trading like it’s still in survival mode and the only story that matters is emissions schedules and validator drama. But that's not where we are anymore.

I’m more excited about this ecosystem now than I was six months ago. Not because of price action, as that will eventually catch up. I’m excited because the infinite game is playing out exactly as it should: the builders who stuck through the heavy months are now shipping proof. That’s the signal I trust more than any chart.

The compass keeps turning. July proved the direction was right all along.

Subscribe now

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

You know what most job postings look like after you’ve read a thousand of them?Noise. Copy-paste corporate templates with the company name swapped out. Half the descriptions could be for any company, any role, anywhere.

I see them all the time in my 9-5. Then last month, I finally found an interesting one.

On June 19, 2026, the Opentensor Foundation posted a role on Polychain Capital’s job board. The title: Subnet Owner (Bittensor Network). Location: Toronto, Canada. Remote.

The pitch, in their words:

“Are you an AI innovator with an entrepreneurial spirit? The Opentensor Foundation is seeking visionary builders to become Subnet Owners on the Bittensor network. In this role, you won’t just join a company, you will found a new AI product venture powered by Bittensor’s decentralized compute network.”

The ask is to found a new AI product venture. And the job description makes it explicit:

“Essentially, you will act as both the lead engineer and the CEO of your decentralized AI venture, responsible for its technical direction and business growth.”

The responsibilities read like a startup founder’s wish list crossed with a machine learning engineer’s dream: design incentive mechanisms, architect tokenomics, build an AI product, manage a community of miners and validators, and drive business strategy. You get dual revenue streams: traditional product revenue and network TAO rewards. And maybe best of all, you keep ownership of your company and your IP.

This job didn’t exist three years ago. The idea that someone could “found a new AI product venture” on a decentralized network, with the chain paying their workforce, with no VC pitch deck, with no cloud infrastructure bill, that’s a category of work that Bittensor’s incentive architecture invented.

And it’s just the beginning.

The Decentralized AI job market

Silicon Valley is fond of saying AI will create new jobs that haven’t been invented yet. Usually it conveniently sidesteps the transition pain. But what if it’s directionally right and just incomplete?

What if the new jobs aren’t “prompt engineer” or “AI trainer”? Those are transitional roles. What if the real new jobs are structural, roles that exist because of a fundamentally different economic architecture?

Bittensor is quietly creating exactly that. Across 128 active subnets, people are getting paid in TAO, directly by the chain, for work that doesn’t fit any existing job category. There’s no LinkedIn profile for a miner, or a standard career path for incentive mechanism design yet.

Call them decentralized AI jobs, if you need a label, but the key is Bittensor is building economic plumbing for a job market nobody can see yet.

The Three Economic Primitives That Change Everything

These roles don’t exist because of better AI. They exist because Bittensor rewrote the rules of how work gets paid for, who owns it, and how it’s coordinated.

1. The Protocol as Employer

Your boss doesn’t decide your paycheck. The protocol does.

Bittensor distributes 3,600 TAO per day based on measurable performance. There’s no infuriating HR department involved here. Just one question: did you produce useful work?

You’re not hired. You’re permissionlessly participating in a market. On Bitcast (Subnet 93), podcasters earn TAO from advertisers while an AI verifies the work. The chain pays automatically.

2. Ownership Baked Into the Protocol

Traditional startups grant equity. Bittensor grants emissions.

Launch a subnet and you get roughly 18% of block rewards. This is protocol-level ownership with a revenue stream that flows as long as your subnet is useful.

The Subnet Owner job posting says it plainly: dual revenue streams (product and TAO), and you keep your IP.

3. Markets, Not Managers

Bittensor turns coordination into competition.

Centralized AI companies use managers to coordinate teams. Bittensor uses incentives to coordinate miners. The protocol rewards useful work with no middle management required.

When Subnet 85 (video) feeds into Subnet 62 (coding) feeds into Subnet 75 (storage), someone needs to architect those workflows. When AI agents start transacting autonomously, you need someone to manage fleets of autonomous workers. You need yet another someone to stress-test incentive mechanisms. You also need a certain someone to provide the judgment machines can’t replicate.

The work is being reorganized around incentives instead of bosses.

Share

The 7 Hidden Roles

Here are 7 roles that only exist because decentralized AI built a new kind of economy. Some of these are live right now. Others are emerging as subnets stack, agents transact, and protocol economics get more complex. (These came from pulling subnet landing pages, cross-referencing Taostats emission data, and reading community discussions about what each subnet actually needs. I listened to every founder interview I could find and 7 roles kept surfacing across all of it.)

A note before we start: You won't see "miner" on this list. Mining is critical technical work, but it predates decentralized AI. The roles below are new job categories that didn't exist until Bittensor created them, and many are accessible to non-coders.

1. The Subnet Owner (already hiring)

Picture a startup founder. Now remove the VC pitch deck, the cloud infrastructure bill, and the board of directors.

What’s left is a person who designs an incentive mechanism, builds a product on top of it, and lets the chain pay their workforce. The Opentensor Foundation is actively recruiting for this. The application is a resume and a proposal for the AI product you’d build.

In a centralized world, founding an AI company means raising capital, hiring a team, renting compute. On Bittensor, the chain subsidizes your miners, the protocol provides your infrastructure, and the incentive mechanism IS your organizational structure.

2. The Agent Operations Lead

You manage a team of 50 workers who never sleep, never tire, and never ask for a raise. They’re AI agents, and on Bittensor they’re economic actors: mining subnets, consuming services, transacting autonomously.

Someone needs to set their objectives, review their outputs, and intervene when they go sideways.

The infrastructure is landing right now. SOMA (Subnet 114) is building MCP servers so agents can reliably connect to external tools. Loosh AI (Subnet 78) launched a “Cognition Engine” to give agents memory and ethical reasoning. As these systems start operating across multiple subnets, the need for human oversight becomes real. A runaway agent can drain resources before anyone notices.

In a centralized company, AI agents are tools. On Bittensor, they’re workers. Workers need managers.

3. The Subnet Compositor

Take compute from Lium. Add storage from Hippius. Bolt on inference from Chutes. Feed in data from D-Search. You’ve just built a complete AI product without touching AWS or buying a single GPU.

A compositor chains subnets together into production pipelines. In a centralized world, one company owns the entire stack, so nobody needs this role. On Bittensor, each layer is a separate, competing market. Composing them into something coherent is a skill that literally didn’t exist before decentralized AI fractured the stack into independent, incentivized pieces.

Builders are already doing this. Lium (SN51) plus Chutes (SN64) gives you a vertically integrated AI product with zero cloud spend.

4. AI Quality Curator

Machines can generate. But they’re a terrible judge. (As an easy test, just ask ChatGPT to rate its own jokes).

Which dataset actually matters? Which model output is genuinely good? Which subnet is solving a real problem versus recycling emissions into a circle of nothing? Without human judgment, competition devolves into a pure compute race. Your taste, your domain expertise, your editorial instinct become the premium differentiator.

“Let’s use the human talent to do the stuff that’s less mundane and focus the energy in the proper space so that value is created. That’s what Bittensor is really trying to do.” (Ala Shaabana, Bittensor co-founder)

Subnets like Subnet 13 (Data Universe) scrape and curate datasets from X and YouTube for model training. Without human editorial judgment guiding the scoring, those pipelines fill with garbage fast.

In a centralized AI company, the curator is a product manager inside the org. On Bittensor, curation is open. Anyone can do it. The chain rewards you for being right. Your judgment becomes a stakeable asset.

5. Subnet Economist

128 subnets. Each one a tokenized startup with its own economy, incentive mechanism, and its own alpha token. They compete, some succeed and others get deregistered.

Someone needs to read the prospectus.

You analyze tokenomics, emission flows, alpha/TAO ratios, net TAO flow. You separate real value from recycled emissions. Capital allocation depends entirely on staker behavior across those 128 micro-economies.

Analysts are already using Taostats to track which networks generate real revenue. Chutes AI is projecting over $2.4M in annual usage fees. Targon Compute is projecting $10.4M. Those numbers may justify Alpha token demand, or other factors may play a bigger part.

In traditional markets, equity analysts cover public companies. On Bittensor, each subnet is a micro-market that needs the same scrutiny, for a community of stakers and delegators who need to know where to put their TAO.

6. Protocol Risk Manager

I wrote about this in a previous issue: Bittensor doesn’t just need machine learning experts, it desperately needs people who understand how complex systems fail and how to build resilience.

That was me speaking from the inside of 20 years of corporate risk management. Those skills transfer directly.

You stress-test incentive mechanisms, identify failure modes, build crisis response frameworks. Bittensor is a complex system with no central authority. When something goes wrong, there’s no CEO to issue a directive. The community has to spot the problem and respond.

Right now, dedicated community members informally stress-test deregistration vulnerabilities, where underperforming subnets lose immunity and get dropped. If subnet count continues to grow, the risk of cascading failures grows. Formalizing this into a dedicated role will be essential.

A centralized company has a risk department because the company is the entity at risk. On Bittensor, the network itself is the entity at risk.

7. Incentive Mechanism Designer

Every subnet runs on a custom grading algorithm. That algorithm is actually the product.

It determines who gets paid, how much, and for what. Get it wrong and your subnet fills with garbage or gets gamed. Get it right and you’ve built a self-sustaining economy.

“Bittensor abstracted the incentive layer, not the money layer. A subnet is just somebody figuring out how to best incentivize a large swath of resources to do one task very, very well.” (Jacob Steeves, Bittensor co-founder)

The Opentensor Foundation lists incentive design as a core responsibility of Subnet Owners today. Tomorrow, it becomes a specialized consulting role: designing mechanisms for other people’s subnets.

Subnet 8 (Vanta) requires a specific mathematical structure to score algorithmic trading strategies. Designing that means treating the mechanism like a product roadmap, balancing game theory, market demand, and technical feasibility.

In a centralized company, incentives are salaries, bonuses, and stock options, designed by HR. On Bittensor, the incentive mechanism is the core product. No traditional job category covers that intersection.

Can't decide? Hit reply and tell me which role you want to unpack in detail in a future issue.

The Gap Machines Can't Close

These 7 roles all sit at the intersection of three things:

1. Human judgment (taste, oversight, risk assessment, economic design)

2. Machine autonomy (agents, subnets, incentive mechanisms)

3. Decentralized infrastructure (the chain as payroll, the protocol as employer)

All three are required. Remove human judgment and you have automation. Remove machine autonomy and you have a traditional job. Remove decentralized infrastructure and you have a job at OpenAI.

The data says this is already happening

The evidence isn’t just in the job postings. It’s in the numbers:

AI created 640,000 U.S. jobs between 2023 and 2025, according to LinkedIn analysis reported by the Wall Street Journal, including new white-collar positions that didn’t exist before.

Stripe Atlas hit 100,000 incorporations, up 130% year over year. AI‑native founders and agents are building companies faster than they’re tearing down jobs.

At the top end, Bittensor subnets are already streaming tens of thousands of dollars per day to miners. These on‑chain income flows increasingly look like a native payroll layer for AI.

Major decentralized AI milestones are being hit: large-scale LLM training runs using commodity hardware, subnet composition pipelines in production, and autonomous agent marketplaces going live.

Code is digital lego. If these lego bricks get cheaper and easier to lay, you don’t use fewer builders. You build what was previously too expensive, too slow or out of reach.

Bittensor is making the bricks (compute, inference, storage, data) radically cheaper. But who builds with them? And what do we call the new jobs that emerge?

Right now, we don’t call them anything. They’re invisible. But they’re real and they’re multiplying.

Start before the job title exists

In my previous scenario planning issue, 83% of you said the “B2B Backend” scenario would arrive by 2027. That’s the future where Bittensor powers your company’s AI infrastructure without anyone in marketing knowing. It’s also the future where these invisible jobs get real job titles.

And when I think about my kids’ future careers, I do worry about whether they’ll know where to look for the jobs being created right now. The ones nobody’s named yet.

The answer in my mind is clear. It’s embracing entrepreneurship.

And in a world where AI automates more tasks every quarter, incentive mechanisms like Bittensor’s may be one of the few options left to add real value to society.

The Subnet Owner job posting is open right now. Its a fully remote role available worldwide. But you don’t have to start there. The funnel is wide:

• Start with content. Subnet 93 (Bitcast) pays creators. An AI verifies your work. The chain pays you.

• Start with data. Subnet 42 (Masa) pays for data cleaning. Basic Python is enough.

• Start with monitoring. Pick a subnet on Taostats. Track its health. Document what you see.

• Start with community. Join a subnet Discord channel. Help someone who’s stuck. Document the solution.

• Start with analysis. Read subnet white papers. Evaluate incentive mechanisms. Become the person others rely on for subnet research.

The worst thing you can do is wait until the job title exists. By the time it has a name and a LinkedIn category, the people who invented it will already be three roles ahead.

Bittensor is designed to reward measurable utility. The invisible job market doesn’t care about your resume. It cares about what you can do.

The chain is the boss. And it’s hiring.

Subscribe now

New to Bittensor? Start Here

If this is your first time reading about Bittensor and your head’s spinning a bit, I’ve written a few pieces specifically designed to get you oriented:

Can’t explain Bittensor to save your life?
Start with I Fixed My Bittensor Pitch. Five real conversations, five different angles. Find the one that works for your audience.

Drowning in jargon?
Check out 5 Terms That Unlock 80% of Bittensor Conversations. Think of it as your survival kit. This has just enough vocabulary to follow Discord without feeling lost.

Want to actually use the network?
Read The Missing Front Door. It walks you through the current state of apps and tools so you can stop researching and start experimenting.

Ready to hold TAO properly?
My 2026 Guide to Bittensor Wallets breaks down your options based on your actual needs, not what crypto Twitter thinks you should use.

The invisible job market is real. These pieces will help you see it.

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

Three people sit on one side of the table: the CFO, the head of alternative investments, and someone from legal who hasn’t said a word in 40 minutes. On the other side, their advisor from Mercer clicks to the next slide.

They’re reviewing the endowment’s 3% “emerging technology” allocation: $50 million that needs to go somewhere this quarter. Three options on screen:

1. a16z Crypto Fund VI - Standard venture exposure, 10-year lockup, minimum $5M.

2. Direct equity stake in an AI startup - High upside, high operational complexity, board seat required.

3. Yuma Total Market Fund - Something about “diversified exposure to decentralized compute infrastructure”.

The CFO interrupts. “Walk me through custody on number three.”

“Institutional-grade,” the advisor says. “Same structure as any other fund. Quarterly statements, third-party audits, full regulatory compliance. You wire dollars, you get exposure.”

“And what exactly are we getting exposure to?”

“TAO, that’s the base layer token. Plus a rebalanced basket of subnet alpha tokens. Think of it like an index fund, except instead of buying Apple and Microsoft, you’re buying decentralized compute capacity.”

The legal rep finally speaks. “Does this fit our mandate?”

“It’s infrastructure,” the advisor says. “Emerging tech, infrastructure play, diversified exposure. Checks all three boxes.”

The head of alternatives asks the only question that matters: “What’s the risk-return profile?”

Ten minutes later, they vote. Yuma gets $10 million.

Not because anyone in that room can explain what a subnet is or they believe at all in decentralization. Because the risk-return profile fits their mandate, the custody is institutional-grade, and the reporting matches their compliance requirements.

That allocation flows into dozens of subnets, including Chutes (SN64), which will process 9.1 trillion AI inference tokens this quarter. But the committee never discussed Chutes. They never discussed subnets at all. They discussed “diversified exposure to decentralized compute infrastructure.”

This is the moment Bittensor stops being a cool experiment and starts being infrastructure.

In June 2026, a subsidiary of Digital Currency Group launched the Yuma Total Market Fund. The pitch is dead simple: you give them dollars, they buy TAO and a basket of subnet alpha tokens, they rebalance quarterly. No wallets, no validators, no token mechanics. Just a quarterly statement like any mutual fund.

The technology doesn’t need to be understood anymore. It just needs to be accessible.

But before we look at how the money moves, we need to understand who is actually playing.

The Landscape: Who’s Actually Here

The institutional wave in Bittensor now contains real players with real capital deployed or vehicles live:

• Yuma (DCG-backed): Launched October 2025 with $10M from Barry Silbert’s Digital Currency Group. Runs three funds: Total Market (TAO + subnet alphas), Subnet Composite (market-cap weighted index) and Large Cap (Mag7 for subnets).

• Grayscale: GTAO trust trading OTC, spot ETF filed December 2025 (amended April 2026), targeting NYSE Arca.

• DSV Fund (Siam Kidd): First 100% Bittensor-focused hedge fund. Kidd presented “Subnets as an Asset Class” at Proof of Talk 2026.

• Others: Bitwise (parallel ETF filing), Safello/Deutsche Digital (European regulated staked TAO ETP), Stillcore Capital (US hedge fund), TAO Synergies (Nasdaq-listed, $750K into Yuma).

But nobody explains this well (or at all): When a billion-dollar allocator buys into Bittensor, where does the money actually go? What happens to the subnet developer building a decentralized weather model in their garage? What happens to the price of an alpha token?

And most importantly, what does this mean for you?

Subscribe now

Following the Money: Six Steps from Allocation to Loop

Step 1: The Allocation

An institution, let’s call it the Oaktree Endowment, decides to allocate $10 million to decentralized AI. Their advisor recommends Yuma’s Total Market Fund because it’s the only product that gives them TAO plus exposure to the subnet layer without operational complexity.

Oaktree writes a check. They don’t fumble about trying to set up a Talisman, Crucible, or TAO.com wallet. They don’t have a notion of what a validator is. They simply get a quarterly statement, just like a mutual fund.

Step 2: The Fund Buys

Yuma takes Oaktree’s $10 million and deploys it across two layers:

1. Layer 1: TAO. The base token. This gets staked (likely with large, reliable validators), and the fund earns staking yield.

2. Layer 2: Subnet Alpha Tokens. This functions like an index. Think of it like the S&P 500, but instead of Apple and Microsoft, it’s Chutes (SN64), Targon (SN4), Vidaio (SN85), and the rest.

Step 3: The Subnet Gets Funded

Let’s follow the money into Targon (SN4), the decentralized AI compute platform.

Yuma buys Targon’s alpha token. Price rises 15% in a week. Now:

• Miners expand GPU capacity - The token price rise makes mining more profitable, so miners add more hardware

• Developers show up - A funded ecosystem attracts talent. Builders who were on the fence commit full-time

• Marketing budget appears - Subnet owners can afford to attend conferences, sponsor hackathons, and land enterprise pilots

This is the initial capital injection. Money hasn’t turned into revenue yet, but the conditions for revenue are being built.

Step 4: The Fundamentals Improve

Six months later, those investments start paying off.

Targon now has 3x the compute capacity it had before Yuma bought in. The developer who joined in Step 3 shipped a confidential AI inference API that works with existing enterprise workflows.

An insurance company signs a 12-month contract for $250K. A biotech firm runs proprietary drug discovery models for $180K/quarter. By Q3 2026, Targon is generating $2M annually from paying customers.

That revenue gets routed back into the subnet’s incentive pool:

• Miners earn more TAO and alpha tokens for providing compute

• Validators earn staking yield

• Alpha token holders benefit from buy pressure as revenue gets converted back to tokens

The token has moved beyond speculation to real revenue-backed yield.

Step 5: The Rebalance

Every quarter, Yuma rebalances the fund. If Yuma’s allocation framework is market-cap weighted, then subnets that grow faster will tend to receive larger weights over time, while smaller subnets will receive less.

• Winners (subnets whose market cap grew) get more weight.

• Losers (subnets that stagnated) get less weight.

This is the same mechanism that keeps Apple and Microsoft at the top of the S&P 500. Index funds don’t decide to buy more, they’re required to. The bigger you are, the more automatic buying pressure you receive.

Say Targon starts Q2 at 10% of the portfolio ($5M in a $50M fund). By quarter-end, Targon grows 40% while the overall market grows 15%. At rebalance, Targon gets 12% of what’s now a $60M fund, $7.2M instead of $5M.

That extra $2.2M comes from selling underperforming subnets.

Step 6: The Loop Perpetuates

Oaktree’s quarterly statement arrives. TAO is up 22%, the subnet basket up 18%. Each quarter’s performance determines the next quarter’s allocation.

But this isn’t a straight line. It’s a loop that runs in both directions.

The Full Loop: Two Paths Forward

The same mechanism that creates exponential growth also creates exponential contraction. Here’s how both paths work:

What triggers this: Oaktree’s $10M becomes $11.8M. The allocation committee approves another $10M. Yuma rebalances up. Winning subnets get heavier weights. The next quarter starts with more capital than the last.

What triggers this: Oaktree’s $10M becomes $6.2M. TAO down 35%, subnet basket down 42%. If they’re mandate-driven (most institutions are), they rebalance down or exit entirely. The next quarter starts with less capital than the last.

What Determines Which Path?

Two factors:

1. Type of Capital

• Patient Capital (endowments, strategic allocators): Rides through -40% drawdowns. 10+ year mandates.

• Hot Capital (momentum funds, prop desks): Exits at -20%. Quarterly performance pressures.

2. Subnet Fundamentals

• Real Revenue: Acts as a price floor. Customers don’t care about token price.

• Speculation Only: No floor. Pure mark-to-market.

The dangerous combination: Hot capital + weak fundamentals = fastest downside spiral.

The safest combination: Patient capital + strong revenue = slowest downside, fastest recovery.

The Subnet Picker’s Dilemma

Where retail still has an edge is where the flywheel begins.

Many funds are market-cap weighted. That means they buy the biggest subnets. The subnets that are already winning.

Why? Because market-cap weighting is the only liquid metric available to a fund managing $50M+. They can’t diligence 128 subnets. They need a rule, and market cap is the rule.

Here’s what market-cap weighting ignores:

• A subnet with real revenue but low token float.

• A subnet with breakthrough tech but no marketing budget.

• A subnet that’s growing 300% month-over-month but started from a tiny base.

The Institutional Blind Spot

Subnet A: $120M market cap, $5M quarterly revenue, 10% growth. It’s in Yuma’s fund.

Subnet B: $8M market cap, $1.5M quarterly revenue, 80% growth. It’s not in Yuma’s fund yet.

Subnet A is “safe.” Subnet B is mispriced with better fundamentals, but too illiquid for a $50M fund. If you have $5K or $50K, you can buy Subnet B. You can see what institutions can’t.

Three Strategies for the Gap

Here are three frameworks for finding subnets that benefit from the institutional wave without competing on market cap.

1. The Pre-Rebalance Strategy Institutional funds rebalance quarterly. But their research teams start looking at new subnets months before.

• Identify subnets that are approaching the liquidity threshold to be included in the next rebalance. You want to buy before the fund does.

• Look for daily trading volume crossing $100K+ consistently, market cap approaching the bottom of the current institutional basket, and real product or revenue beginning to materialize.

  Example: A subnet ranked 15-20 today could enter the fund in Q4 if it crosses a liquidity threshold. Getting in at 18 is better than buying at 8 after the fund already did.

2. The “Revenue-First” Strategy Market cap is backward-looking; it reflects past attention. Revenue is forward-looking; it reflects whether anyone actually pays for the product.

• Find subnets with the highest revenue-to-market-cap ratio—the ones generating real dollars but still undervalued by attention metrics.

• Look for disclosed or estimable AI service revenue, enterprise clients or API usage growing, and a token price that hasn’t caught up to revenue growth.

  Example: A subnet processing $500K/month in API calls but valued at $10M has a very different profile than a $100M subnet with no customers. The former is an operating business disguised as a token.

3. The “Niche Moat” Strategy Institutional funds need broad exposure. They can’t buy into hyper-specific subnets with small TAMs (total addressable markets). But that doesn’t mean those subnets aren’t valuable.

• Find subnets dominating a niche that big compute providers ignore.

• Look for unique data or compute offering (weather, biologics, edge devices, specific languages), deep domain expertise that can’t be replicated by generalist AI, and revenue from customers who have no alternative.

  Example: A subnet doing decentralized drug discovery for rare diseases won’t be in Yuma’s fund. But it might have a $2M/year contract with a pharma company and zero competition.

Subnet Spotlight: Two Riding the Gap

Here are two subnets that illustrate the blind spot:

Vidaio (SN85): The Compression Leader

• Ranked 1 for video compression efficiency in benchmark tests (June 2026).

• Addresses a real market: video streaming costs are a $100B+ problem.

• Currently small enough that it’s not in major institutional baskets.

• If Vidaio starts hitting big revenue figures, it becomes an obvious rebalancing candidate. Getting in before that rebalancing is the play.

Zeus (SN68): The Weather Oracle

• Ultra-precise weather data via decentralized sensor networks.

• Serves agriculture, insurance, and logistics: industries that spend billions on weather data.

• Scientific compute niche that generalist AI can’t replicate.

• Niche moats like this don’t show up in market-cap indices. But they show up in enterprise procurement departments.

The Simple Rule

Here’s the one-sentence framework to take away:

Institutions buy market cap. You should buy fundamentals that haven’t become market cap yet.

The moment a subnet enters Yuma’s fund, the easy money is gone. The edge is in finding what gets there next.

The Deeper Question: Decentralized AI, or a New Goldman Sachs?

Bittensor’s entire value proposition is decentralized AI. But institutional capital is, by definition, concentrated. It is controlled by large allocators who deploy in size.

If a DCG subsidiary controls the largest fund, Grayscale controls the ETF, and a handful of validators control most staked TAO... are we actually decentralized? The April 2026 Covenant AI departure was partly sparked by these exact governance centralization concerns.

The reality is the Incentive design is what matters. Institutional capital creates price floors for alpha tokens and reduces sell pressure. This gives subnet teams runway to build products and generates healthy pressure to ship. Without institutions, alpha tokens are thinly traded and easily manipulated.

If the protocol can keep institutional influence on governance separate from their influence on price, it might work. The Root Reborn debate is happening exactly because the community is actively trying to figure out where validators should sit in the power structure.

Putting It Together: Your Action Checklist

If you only take three things from this issue:

1. Understand the loop. Institutional capital → subnet funding → product improvement → revenue → token support. It’s real, it’s happening, and it’s accelerating.

2. Don’t chase what’s already in the fund. Chutes and Targon are great subnets. But if you want edge, look at what’s approaching institutional thresholds, not what’s already there.

3. Use what institutions can’t Time, flexibility, and small-size access. A $50M fund can’t touch an $8M subnet. You can.

The fact that you can still buy the small sweaty startups on Bittensor while Yuma can’t is your proof the loop isn’t closed yet.

But watch the rebalance dates. The next quarter is always on the horizon and the next wave of capital is already being allocated.

Share

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

Take early AWS back in 2006. It starts as a chaotic, internal experiment: loosely structured, developer-first, and full of edge-case failures. But its real inflection point came when Amazon imposed strict service-level guarantees, standardized APIs, and operational discipline, transforming it from a tinkerer’s toolkit into infrastructure that enterprises could reliably build on.

This is analogous to what we just witnessed across the Bittensor network. June 2026 is the month the ecosystem stopped being a speculative playground and started becoming real infrastructure.

How to Read This Compass

I’m organizing these updates by direction: North, East, South, West to give you a repeatable mental model for navigating Bittensor’s complexity.

North: Foundation 🏛️ (governance, protocol, core team, security events)

The structural layer. Changes here determine who controls the network and how decisions get made.

East: Capital 💹 (markets, volume, liquidity, integrations)

The economic layer. This is where price meets liquidity and institutional conviction shows up in the data.

South: Technology ⚙️ (shipped features, benchmarks, infrastructure, ecosystem health)

The execution layer. What actually works, what got built and where’s the technical proof.

West: Adoption📡(narrative, mainstream validation, external voices)

The perception layer. This is how the outside world sees Bittensor, from industry titans to normie investors.

Each month, some directions will dominate the signal. Others will stay quiet. The goal is to see which quadrants are moving and what that means for your strategy.

All four directions on the compass moved this month. Two held their own, but the others took center stage. Lets review what happened.

Subscribe now

💹The Core Incentive Engine Gets a Radical Tune-Up

Bittensor’s economic engine just received its most aggressive overhaul since Dynamic TAO (dTAO) went live. Two massive structural changes hit the network in the exact same window.

First, the network activated a major shift in how daily token emissions are carved up across subnets, moving back to a rigorous price-based model that replaces the older flow-based framework.

The mathematical allocation now actively penalizes unproductive or withheld emissions while shifting weight toward newer, high-performing subnets over time.

Second, and far more dramatically, OpenTensor shipped a protocol update on June 22 that completely blocked token emissions to 57 inactive and exploitative subnets.

This single move redirected roughly a third of all network emissions directly into the pockets of subnets that are actually building out real code. A new “miner-burn penalty” was instituted to eliminate squatters who route rewards to themselves without delivering meaningful output, and subnet eligibility is now scrutinized on a strict weekly review cycle.

The immediate beneficiaries are exactly who you would want to win: Chutes (SN64), the powerhouse AI inference subnet that generated $43M in Q1 2026 real AI revenue, and Targon (SN4), Manifold Labs’ confidential decentralized compute platform that currently ranks as the highest-revenue subnet for enterprise hardware-enforced AI workloads.

Why It Matters

For two years, the ecosystem tolerated dead weight collecting emissions, acting as a massive drag on protocol efficiency and network credibility. Shifting emissions to high-utility subnets forces capital allocation to map directly to real-world demand and revenue generation rather than arbitrary timing games.

The Takeaway

This is Bittensor’s first genuine quality enforcement mechanism. The network is finally learning to reward builders over squatters.

🏛️Validators as Fund Managers? The Root Reborn Debate

On June 16, Bittensor co-founder Const unveiled the “Root Reborn” proposal, triggering the most intense governance debate the ecosystem has seen in months. The underlying problem is simple: right now, root stakers automatically liquidate subnet alpha tokens, which exerts persistent, structural sell pressure across the entire subnet market.

Root Reborn proposes a total rethink: replacing that automated dump mechanism with validator-directed reinvestment baskets. Instead of blindly dumping subnet tokens onto the open market, validators would actively choose which productive subnets to back and compound their yield into.

The pushback, however, was immediate and severe. On June 19, Yuma (a DCG subsidiary) published a scathing analysis warning of deep moral hazard. They argued that giving validators this level of centralized capital allocation authority introduces substantial risks to the underlying balance of power. Institutional voices like Safello CEO Frank Schuil quickly weighed in, and the debate dominated OpenTensor’s weekly Novelty Search community call.

Why It Matters

Automated liquidations create a structural drag on subnet token prices, suppressing the native economies of productive subnets. Fixing this changes the economic incentives of validators from passive gatekeepers to active ecosystem asset managers, but doing it incorrectly risks centralizing too much power in a few hands.

The Takeaway

If Root Reborn passes, it could completely alleviate the artificial price ceilings that have suppressed subnet token prices since dTAO launched. But the real win here is the maturity of the governance itself. This isn’t a founder handing down dictates from a mountaintop; it’s a sophisticated, multi-billion-dollar architectural negotiation.

🏛️The Hard Road to Full Decentralization

On June 22, co-founder Jacob Steeves (Const) published a remarkably candid decentralization roadmap that began with an admission most crypto founders would never dare utter publicly: “Bittensor is currently not a decentralized protocol in the way Bitcoin is. It can be, and it will be, but it isn’t yet.”

Specifically, Const acknowledged that Bittensor’s core economic incentive layer remains fundamentally centralized. This transparency follows a highly volatile period of governance friction, most notably the Covenant AI departure in April 2026. At the time, subnet founder Sam Dare publicly accused the core team of unilateral control and halting emissions without consensus, a dispute that triggered a painful crash in the price of TAO.

The new 18-month roadmap explicitly targets full decentralization by December 2027. Const explained that the foundation intentionally prioritized rapid development speed over absolute decentralization during the early hyper-growth phase of the network. Now, the timeline is locked in.

Why It Matters

Centralization is the ultimate bear case that traditional and institutional investors use to dismiss the network. By openly acknowledging the current centralization gaps, the core team eliminates a major narrative vulnerability and creates a transparent, multi-year benchmark for real accountability.

The Takeaway

The core team has given the community a visible yardstick to hold them accountable. Longevity requires decentralized foundations, and the network is finally drawing the blueprint.

💹Retail & Institutions Get One-Click Access

On June 25, Yuma announced the launch of the Yuma Total Market Fund, an institutional-grade investment vehicle that offers a single entry point into both TAO and the wider subnet alpha token ecosystem. As Barry Silbert framed it, the fund represents a paradigm shift: decentralized AI investing is no longer restricted to venture capital or standard hardware equities like Nvidia.

But the access expansion isn’t just for institutions anymore. This is part of a broader standardizing trend we’ve seen all month:

Retail Exchange Listings: On June 29, Kraken announced they are listing several Bittensor subnet alpha tokens directly on their centralized exchange. This allows retail traders to seamlessly access these AI-focused assets without having to navigate complex on-chain automated market maker (AMM) pools.

Regulated European Access: On June 2, Safello and Deutsche Digital Assets cross-listed the Staked TAO ETP (ticker: STAO) on Euronext Paris, expanding access to yield-bearing TAO across major European exchanges.

The Louvre Track: The Proof of Talk 2026 conference featured a dedicated Bittensor Track, putting the network directly in front of 2,500 C-level asset managers representing a staggering $18T in collective AUM.

DeFi Primitive Alpha: On June 25, Forge opened its public testnet for Bittensor-native TAO borrowing against alpha token collateral, providing the first real glimpse of capital efficiency rails for subnet investors.

Why It Matters

Complex on-chain AMM pools, custom wallets, and fragmented liquidity have historically locked out traditional and retail capital. Bringing these assets onto centralized venues and regulated ETPs radically expands the addressable market and removes the operational friction of the ecosystem.

The Takeaway

For a long time, the mission of making decentralized AI accessible to regular investors was stalled by complex user experiences. This new centralized exchange and institutional plumbing bridges that gap. Bittensor is evolving from an esoteric crypto experiment into a highly accessible alternative AI infrastructure thesis.

⚙️Chutes Levels Up

This month, Chutes completed a full move onto TEE (Trusted Execution Environments) and brought its first Blackwell B200s online, while also expanding distribution through openGPU. This is a concrete sign that the network’s AI compute stack is maturing into something highly reliable and deployable.

Why It Matters

When we look at decentralized AI through a traditional product management lens, enterprise adoption always stalls on two major product risks: feasibility (can it scale?) and business viability (is it secure and compliant?).

By bringing Blackwell B200s online and expanding through openGPU, Chutes is actively solving the feasibility problem: proving that a decentralized network can handle heavy, enterprise-grade compute demands.

But the real quiet victory here is the rollout of TEE. For the non-technical investor, TEE translates directly to “verifiable inference”: cryptographic proof that an AI model executed exactly as requested without ever exposing the underlying prompt or sensitive data. It is the exact security baseline corporate clients demand before letting proprietary business data touch an open-source network.

The Takeaway

We are watching Bittensor evolve from a fascinating technical experiment into commercial-grade infrastructure. For long-term investors, this is exactly the transition you want to see: a shift away from token speculation and toward shipping sticky, B2B-ready products that can compete directly with centralized cloud providers.

📡Centralized AI Stumbles, The Decentralized Narrative Accelerates

On June 13, the US government abruptly suspended export access to Anthropic’s AI models, sending shockwaves through global tech hubs. The sudden suspension put the core vulnerability of centralized AI on full display: an entire enterprise pipeline can be severed overnight by a single regulatory stroke.

Grayscale immediately capitalized on the moment, publishing extensive research positioning Bittensor as the premier decentralized AI alternative. Their analysts explicitly noted that demand for permissionless, open-source AI infrastructure is bound to accelerate as centralized moats face increasing geopolitical and regulatory interventions.

The market response was immediate and violent: TAO surged nearly 30% within 12 hours of the news as capital rotated heavily into the decentralized AI narrative.

Why It Matters

It demonstrates that centralized AI pipelines are inherently exposed to single points of geopolitical and regulatory failure. Enterprise clients cannot afford to have their foundational infrastructure cut off overnight due to cross-border policy shifts.

The Takeaway

This was a fundamental proof of concept. Every closed, corporate AI provider is one executive order away from restriction, and Bittensor’s globally distributed architecture is the natural insurance policy against that exact point of failure.

Final Thoughts

We’re entering the 18-month window that will define whether Bittensor becomes infrastructure or remains an experiment. The roadmap to full decentralization is public, low-signal subnets are being ejected, and capital onramps for both retail and institutional players are operational.

The experimental era just ended. Now comes the stress test: can this architecture actually handle scale?

Here’s what I want you to do: Don’t just read this and move on. Choose one development from this month; tracking alpha token price action on Kraken, scrutinizing whether the decentralization roadmap is hitting its marks, whatever resonates, and actively test your conviction. Are you in this for returns? Or because you believe Bittensor represents something bigger than profit? Because if Bittensor succeeds at the latter, the former takes care of itself.

Share

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

I rented the $300 bounce house, set up the tables in the garden, bought decorations that weren’t supposed to get wet, and waited.

1 PM arrives and exactly when nine hyperactive pre-schoolers were supposed to turn up, the sky turned the color of a bad bruise.

It poured.

Not a light drizzle. A full-blown downpour that turned our lawn into a swamp and left a bunch of sugar-fueled five-year-olds trapped in my kitchen, screaming at a decibel level that probably violated local noise ordinances.

The app wasn’t exactly lying; it just gave me a single guess instead of the full probability distribution.

Human beings crave clean, single-point answers. But the world doesn’t work in yes or no. It works in ranges and variance and fat tails. In finance, ignoring that variance isn’t just lazy. It’s how you get wiped out.

That exact gap, the distance between a comforting guess and raw probability, is where Synth (Subnet 50) steps in.

Inside the Swarm: How Synth Maps the Future

Most price-forecasting products give you a single number. “BTC hits $65k.” It’s like flinging one dart at a dartboard and hoping for the best. Synth gives you the entire dartboard, and tells you which sections are bigger.

Here’s how it works. Hundreds of miners submit 100 simulated price paths every 5 minutes. Validators score them using CRPS (Continuous Ranked Probability Score), which answers one question: how close was your probability cone to what actually happened? The output is a live API that computes liquidation probability at 6, 12, 18, and 24 hours.

As Synth’s founder James Ross puts it, modeling raw price is nearly impossible, but you can continuously improve at modeling volatility. He calls the subnet “the Olympics for Monte Carlo simulations.”

By shifting the goalposts from predicting a price to mapping a distribution (sharpened every five minutes by hundreds of models competing to be least wrong), Synth turns Bittensor’s raw compute into a pristine financial utility.

Subscribe now

Why Now? The Rise of the Machine Allocator

In early 2026, prediction markets cleared over $20B in monthly volume.

But the bigger shift is the autonomous AI agents now trading real money on-chain.

A legacy oracle excels at telling an agent what an asset costs this exact second. But to size a massive bet or manage liquidation risk over a 24-hour horizon, machine allocators need deep statistical resolution: hundreds of models competing to be the least wrong.

Multiply hundreds of miners delivering 100 paths every five minutes across an entire asset index, and the network hammers out roughly 165 million high-resolution data points a day. The old way gave us a fragile guess, but the Bittensor way gives machine buyers the deep statistical map they actually need to survive.

It’s building the foundational risk API that the next generation of machine allocators, lending protocols, and DeFi agents will plug into to survive. Call it a new category: Risk Infrastructure.

Four Questions That Actually Matter

Let’s evaluate Synth the way a product manager would: through adoption signals, not just architecture.

Q1: What problem does it solve?

Anything making financial decisions on-chain is working from data that’s bad at the thing that matters most: tail risk.

Single-point guesses are close to useless for managing liquidations, sizing positions, or pricing options. A trader who acts on “BTC to $65k” has two outcomes: roughly right and a modest return, or wrong and annihilated. There’s no middle.

A trader who acts on a full distribution: “54% bullish, 90% confidence between $61k–$66k, fat-tail scenario at $52k” can size using the Kelly Criterion. Risk management transforms from paranoid guessing into a calculated, machine-readable routine.

Right now, miner behavior looks like it could mean two things: they’re building forecasts good enough to trade on, or they’re optimizing for protocol rewards. We can’t distinguish between the two from public data yet. Until we see live trading results or independent performance audits, the demand is largely based around the protocol paying miners who produce output the protocol consumes.

Q2: For whom?

Two tiers. Quant traders and options pricers are the primary market; on-chain DeFi and AI agents are the secondary. This is a power-user product for hedge funds, options desks, perpetual DEXs, and autonomous trading agents moving millions without human oversight.

The go-to-market advantage is Mode, the Ethereum L2 that co-founded Synth and serves as a built-in first customer, wired into its AI Terminal and DeFAI agents:

Most subnets die because they build a product and then go looking for customers; Synth built the product inside a customer. But if Mode is both builder and principal consumer, much of the apparent demand is captive, not independent validation. Synth shows signs of external monetization, but no public evidence yet demonstrates meaningful demand separation from Mode.

There’s a hidden third tier Ross emphasizes: ML teams training financial models. Synth’s daily data isn’t just for querying, it’s for training. Historical data leads to overfitting; synthetic data generates the edge cases that matter in live markets. A quant team that bakes Synth’s data into its model weights isn’t switching APIs easily. That’s a stickier, harder-to-commoditize relationship than query-time consumption.

Q3: How big is the opportunity?

A narrow SAM of approximately $2.5B in alpha signals and alt data, with a broader TAM of around $44B as DeFi and autonomous systems adopt better forecasting infrastructure. But data and oracle markets are notoriously hard to monetize, and the gap between a $44B opportunity and a $6M market cap only narrows if Synth can capture value, not just create it.

The core pain is that traders, protocols, and agents need dependable forward-looking signals that fit into live decision workflows. Synth is solving that by making forecasting a repeatable product loop: users get probabilistic outputs, miners compete to improve them, and the four-week Polymarket trial in October 2025, which returned roughly 110%, provides an early signal that the workflow can create measurable value.

There’s also an expansion vector the $44B doesn’t capture: Synth as a dependency inside the Bittensor subnet graph itself.

If composability becomes real, any subnet touching financial data needs a risk input: take a trading-agent subnet querying Synth for liquidation probabilities, or a lending subnet using it for collateral scoring. If Synth becomes the risk oracle in that stack, the addressable market is every financial subnet on Bittensor. Whether that happens depends entirely on the team's execution discipline: the current plan to calculate operating costs and redirect revenue to buybacks puts all value capture in their hands.

Q4: How will we know if it’s working?

We’ll know Synth is working when external users choose it as forecast infrastructure, not just when the internal network is active.

The clearest signal is independent API consumption: usage from traders, protocols, and agents outside the Mode ecosystem, with repeat integrations over time.

The second signal is out-of-sample performance across many trials and asset classes, showing that forecasts remain useful in live conditions rather than only in backtests or one-off wins.

On the supply side, we should see sustained miner participation from teams that consistently perform well, with weak speculative strategies filtered out through weekly reporting, immunity periods, and reward persistence.

Roadmap execution matters too: adding new assets (recently equities), shorter time horizons, and eventually more use cases should expand both demand and supply without degrading forecast quality.

The Takeaway: Bull vs. Bear

The bull case: Synth is the first financial subnet selling a better question, not just a better answer. It has a working product, a built-in customer in Mode, and hidden upside in ML training-data lock-in. If subnet composability happens, Synth becomes the default risk oracle. A $6M market cap against a $44B TAM is real asymmetry.

The bear case: Value creation isn’t value capture. Deep options markets already do this well. Mode’s demand may be captive. And if miners truly had a superior edge, they’d trade on it rather than farm emissions.

Synth is the most legitimate financial subnet on Bittensor right now. But the only number that confirms product-market fit is independent API consumption. When non-Mode entities start paying for this data, it shifts from a well-engineered project to an undeniable market force. Everything until then is promising but unproven.

The Bounce House Question

Back to the birthday party: that weather app didn’t give me a distribution because I didn’t want one. I wanted the comforting answer. The clean “yes” that let me rent the bounce house and move on. The app sold me certainty, and I bought it. Then I spent an hour hosing cake frosting off my kitchen walls.

If it had shown me the real distribution, let’s say 70% mild, 20% light rain, 10% biblical downpour, I still might have rented the bounce house. But I would have set up the backup plan, maybe bought indoor some decorations. Carried the variance instead of pretending it didn’t exist.

That distance, between the answer we want and the distribution we actually need, is where Synth lives.

The real bet isn’t that miners can beat options desks. It’s that the next million financial decisions on-chain won’t be made by humans who crave clean answers, but by machines that require dirty, honest, high-resolution probability.

Now, for me, the accessibility tension is still real. The subnet that figures out how to show probability distributions to someone who doesn’t know what CRPS stands for is the one that breaks out of the quant ghetto. Making probability visual is making probability adoptable.

But every piece of infrastructure that changed the world started as something nobody understood. TCP/IP was impenetrable, while HTTP was academic. Technology that matters arrives as a necessary first, and then the builders who survive figure out the translation layer.

So I’ll leave you with this: if you could see the full probability distribution for one decision in your life: your mortgage, your portfolio, your kid’s college fund, would it change what you do? Or would you still reach for the comforting answer?

The gap between those two responses isn’t just Synth’s market. It’s the entire gap between the world we live in and the world we’re building.

Share

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research

Three are price charts. Five are on-chain analytics dashboards. Two are X feeds tracking whale wallets. One is a Google Doc where I’m manually logging trade ideas. The rest are incubating research rabbit holes.

And then there’s my wallet: a treasure-filled filing cabinet that does exactly nothing with any of this information.

But as artificial intelligence merges with decentralized finance (a movement known as DeFAI), this kind of passive wallet is becoming obsolete.

One team in the Bittensor ecosystem is trying to change this. Subnet 45 (Talisman AI) is attempting to turn the wallet from a static asset holder into an intelligent, autonomous co-pilot. Backed by Team Rizzo, the DNA Fund, and the builders of the Talisman Wallet itself, it’s one of the most ambitious subnet plays I’ve come across.

But how do we know if it’s actually solving anything?

Most Bittensor subnet coverage focuses on emissions, alpha price charts, or miner counts. Those are lagging indicators, telling you what happened, not whether something worth building is being built.

So I’m evaluating this differently using a structured way to separate products from projects. Four questions:

• What’s the business objective?

• How would we know if it worked?

• What problem does it actually solve?

• Who is it for?

Let’s break down Subnet 45 through this lens and diagnose where the strategy shines and where the product friction is starting to mount.

The Strategic Play: What Business Objective Does Talisman (SN45) Address?

For years, Talisman has been a leading multi-chain wallet, offering a sleek interface for Ethereum, Solana, Polkadot, and Bittensor. I covered it in my wallet guide here. However, in a crowded wallet market, beautiful design is no longer a sustainable moat.

The launch of Subnet 45 is a highly strategic shift to turn Talisman into the undisputed gateway for the AI-agent economy.

The DeFAI Integration Roadmap

Phase 1: Perception (Live)

Launch of Subnet 45 (Talisman AI) on Bittensor. Miners analyze social signals (initially from X/Twitter) to determine sentiment and technical relevance for specific subnets. Now available within the wallet.

Phase 2: The DeFAI Portal (In Progress)

Beta launch of the DeFAI Portal within the Talisman wallet. This introduces the “Decision Engine,” which provides users with trading signals and staking suggestions based on SN45 data.

Phase 3: Agentic Execution (Upcoming)

Launch of Policy-Based Automation. This allows users to set “intents” (e.g., “maximize yield on TAO”) where the agent executes rebalancing or staking within pre-defined risk boundaries and spend caps.

Phase 4: Autonomous Marketplace

A full Agent Marketplace and custom builder, allowing users to deploy specialized agents that manage liquidity autonomously across multi-chain environments.

To pull this off, the team is executing three core strategic moves:

1. Building a Proprietary Intelligence Moat: SN45 acts as the wallet’s decentralized “eyes and ears,” harvesting live market intelligence (social sentiment, whale movements, tokenomics). This gives Talisman an exclusive, real-time data engine you simply can’t bolt onto MetaMask or Phantom after the fact.

2. Constructing “Agent-Ready” Rails: Capitalizing on their May 2026 partnership with General Tensor (who recently acquired Backprop Finance), Talisman is co-developing secure multisig infrastructure. This provides the secure plumbing required for institutional funds and autonomous AI agents to co-exist safely on-chain.

3. Aligning Token Incentives: Transitioning from a pure software tool to a token-powered network, the SN45 Alpha token coordinates the ecosystem by rewarding miners for feed quality, acting as the currency to unlock premium wallet actions and compensating developers for new agent strategies.

   Subscribe now

Measuring Success: How Will We Know If Talisman Has Succeeded?

For Subnet 45 and the Talisman ecosystem, success isn’t just about block times or mining difficulty. The framework for their product-market fit comes down to five core metrics:

The Friction Point: What Core Problem Does It Solve?

Talisman is directly targeting three massive user frustrations:

• Information Overload: you need to constantly monitor X for hype, watch block explorers for whale moves, and track technical charts, all in different apps. By the time you’ve pieced the picture together, you’re late. And you’re emotional. SN45 aggregates these fragmented workflows into a clean, structured data stream to beat FOMO-driven late execution.

• The Dumb Wallet Limitation: Current wallets cannot execute conditional logic like, “Take profit on my TAO if social sentiment drops 20% and on-chain whales start selling.” The Decision Engine introduces “if-this-then-that” intelligence directly into self-custody.

• The AI Trust Gap: Automated trading bots usually require you to hand over your private keys: a total non-starter for responsible people. Talisman solves this via human-in-the-loop approvals. The AI handles the heavy lifting of scanning and packaging trade batches, but you retain 100% control of your keys and must provide a one-click approval before anything hits the chain.

Who is Talisman for? The Split Identity Crisis

Talisman is currently trying to design a single product for three entirely distinct target markets:

1. The DeFAI Retail Power User needs simplicity. They want an invisible co-pilot that surfaces clean, one-click staking or trading opportunities without adding cognitive clutter.

2. The Institutional Treasury needs the exact opposite. They require explicit complexity: rigid audit trails, compliance-friendly multisig infrastructure, and custom programmatic risk boundaries.

3. The Autonomous AI Agent doesn’t care about a UI at all. It needs low-latency APIs, reliability, and permissionless financial rails to execute machine-to-machine commerce.

The core product risk is that the same wallet cannot be all three things without confusing all three users. Optimize for retail, and institutions view it as a toy; optimize for institutions, and retail finds it overwhelming. Trying to be everything to everyone risks alienating the core user base. Talisman needs to decide if these AI features are an invisible backend enhancement for retail, or an advanced cockpit for power users and whales.

The Product Reality: Where Ambition Meets Friction

Shifting a product from a clean consumer wallet to a decentralized intelligence layer introduces deep architectural complexity.

Ecosystem Sprawl vs. Core Focus

To the team’s credit, they are transparent on their website that the AI agent features are “Coming Soon.” The immediate issue isn’t vaporware, it’s ecosystem sprawl.

Right now, the platform is pulling the user in four different directions:

• The core multi-chain wallet (pitched at retail users managing standard swaps).

• The Subnet 45 AI integration.

• A section dedicated to the details, status, and utility of their native community NFT collection.

• A highly thematic portal for the $SEEK token (complete with its own staking and lore).

For a product whose primary value proposition is making Web3 easier, this volume of pillars creates a steep learning curve. Managing a native NFT project, a complex token ecosystem, and an advanced AI integration within a single interface risks severe feature bloat.

As the AI features roll out, the UI will need massive consolidation to keep the AI acting as a streamlined assistant rather than another overwhelming tab.

The Multi-Token Friction

The mental model for participating in this ecosystem is further complicated by asset fragmentation. Users have to navigate both $SEEK (the ERC-20 token for governance and premium agent access) and SN45 Alpha (the Bittensor-native commodity token minted via dTAO for network utility).

The immediate user confusion is obvious: “If I believe in Talisman AI, what am I supposed to buy?” Fortunately, leadership seems to recognize this friction. In May 2026, Frank Rizzo noted they are working on a plan to streamline the ecosystem, with the ultimate goal of having the SN45 token represent both the wallet utility and the subnet. Until that consolidation materializes, however, users are left navigating a token maze.

The Takeaway

Talisman is executing on a brilliant macro strategy. Most subnets would kill for what this team has: an existing consumer product with real users, a revenue model that doesn't depend on emissions, a functioning data network, and a credible institutional partner in General Tensor.

The partnership structure, Rizzo's subnet feeding intelligence into a wallet that people already use, bypasses the academic-only trap that strangles so much of the broader Bittensor network. And there’s a consumer product ready to use the data from day one.

But the gap between the vision and the product you can touch today is real.

The Bull Case:

The first AI Agent feature ships inside the wallet, and it's actually impressive. Adoption follows. The Alpha token unlocks real utility. The burn rate accelerates beyond baseline fees. General Tensor brings institutional flow. Talisman becomes the default intelligent wallet in Web3 with AI as a core component.

The Bear Case:

The roadmap stays ahead of shipped features for too long. Meanwhile, competitors with even larger user bases bolt basic AI tools onto their wallets (simpler, worse, but available) and the first-mover advantage evaporates. Feature bloat and token confusion stall retail adoption, and while the subnet keeps earning and the data keeps flowing, the product never catches up to the promise.

Here’s what I’m watching for: The day I open Talisman and an AI agent does something visibly useful without me asking. Maybe a trigger that fires, or a sentiment signal that saves me from a bad trade. A single automated action that I couldn’t have done as easily on my own.

When that happens, this project moves from a smart speculative bet to an essential piece of Web3 infrastructure.

The bet is smart today. But the clock is ticking, and competitors aren't waiting.

Share

Which subnet should I evaluate next? Reply and let me know.

And if you found this useful, share it with someone who’s trying to make sense of Bittensor subnets beyond the price chart.

Until next time.

Cheers,

Brian

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

His pyjamas, pillow, duvet, the bedroom carpet.

And worst of all, his favorite penguin teddy took the direct hit.

But he wasn’t crying. Still half-asleep, he just stood in the hallway holding the soaked penguin in one hand, waiting for me to fix it. That moment told me something: the thing wasn’t working anymore, and there was no panic. Just an expectation that Mommy and Daddy would fix it.

May felt like that for Bittensor. April’s Covenant exit cracked the seam, but May was the month the community started stitching it back together.

And luckily for me, I’d had a replacement teddy ready for a while. Now he carries around three of them, a new penguin family.

How to Read This Compass

I’m organizing these updates by direction: North, East, South, West to give you a repeatable mental model for navigating Bittensor’s complexity.

North: Foundation 🏛️ (governance, protocol, core team, security events)

The structural layer. Changes here determine who controls the network and how decisions get made.

East: Capital 💹 (markets, volume, liquidity, integrations)

The economic layer. This is where price meets liquidity and institutional conviction shows up in the data.

South: Technology ⚙️ (shipped features, benchmarks, infrastructure, ecosystem health)

The execution layer. What actually works, what got built and where’s the technical proof.

West: Adoption📡(narrative, mainstream validation, external voices)

The perception layer. This is how the outside world sees Bittensor, from industry titans to normie investors.

Each month, some directions will dominate the signal. Others will stay quiet. The goal is to see which quadrants are moving and what that means for your strategy.

Three of the four directions on the compass moved this month, one stayed quiet and another dominated. Lets review what happened.

Subscribe now

🏛️ Conviction ships to mainnet

What Happened:

On May 13, Conviction shipped to mainnet. It’s Bittensor’s formal response to BIT-0011 and to the Covenant exit in April.

Every emission a subnet owner earns now auto-locks the moment it lands in their wallet. Conviction is calculated as stake multiplied by time, re-averaged every 30 days, and designed to make ownership more accountable.

Future plans involve converting those locked positions into transferable governance power so the community can vote underperforming owners out.

The first three subnets onboarded were the three Covenant abandoned: SN3 (Templar/Teutonic), SN39 (Basilica), and SN81 (Grail).

Why It Matters:

My April post called community miners rebuilding the abandoned subnets “ownership through commitment, not founder discretion.” Conviction is that idea, formalised at the protocol level.

What I noticed is the speed. Covenant walked out on April 9. By April 16, Steeves was on Discord explaining the mechanism. By May 13, it was code shipping to mainnet.

This is the kind of response cycle you want to see from a protocol that just took a public punch. Slow committees would have spent six months arguing. Bittensor wrote the fix, tested it, and pushed it in five weeks.

The Takeaway:

If you’re delegating, start tracking the conviction scores on the subnets you back. This is the new operator due diligence metric, and the people running the most committed teams are about to look very different from the people running the loudest ones.

For non-coders, its not quite available on Taostats yet (visible in the UI but no data), but until then you can view Total Stake Weight as proxy.

Conviction is locked alpha, and staked alpha can still be unstaked and sold, just not on a whim. But a high staked ratio means holders are parking tokens rather than keeping them sell-ready, which is the same question conviction is built to answer: who's committed, and who's liquid. So I read it as a floor, not a verdict.

🏛️ The emissions refactor changed how rewards flow

What Happened:

Around the same time Conviction shipped, Bittensor also updated its emission logic to a new model. The refactor concentrates rewards among roughly 30 productive subnets and introduces a new parameter called Net TAO flow.

The protocol now distributes newly issued TAO based on normalised flow data, tied to prior emissions and buybacks. In plain terms: the protocol now actively rewards subnets that produce useful work and starves the ones that don’t.

Why It Matters:

Conviction fixes who controls the network and now the emissions refactor fixes what the network rewards. Together they are the most significant tokenomics change since dTAO.

The economic logic is simpler now. Under the old model, low-emission subnets could still drain meaningful TAO into the market with little to show for it. Under the new model, that pressure compresses.

If you’re holding TAO, you should expect cleaner alignment between productive subnet activity and token issuance. If you’re staking into subnets, the gap between top performers and the long tail is about to widen sharply.

The Takeaway:

Emissions data over the next 90 days will be telling. Watch which subnets gain share and which lose it: the list of subnets gaining emissions share in this period could be the new smart money signal.

💹 Wormhole opens the Solana door

What Happened:

On May 5, the Solana Foundation and Wormhole Labs announced that TAO is now live on Solana via Wormhole’s Sunrise gateway. It’s a canonical bridged version of TAO trading natively on Jupiter, Meteora, Phantom, and Solflare.

Wormhole’s Native Token Transfer framework backs the bridge, which means liquidity is unified across Solana rather than fragmented across wrapped variants. Within seven days of the announcement, TAO was up 25%.

Why It Matters:

Solana wants to be the execution layer for AI agents. Bittensor is the intelligence layer for decentralised AI. Until now, those two worlds were connected through wrapped versions of TAO that fragmented liquidity and created friction. Wormhole just removed that friction.

The result is that TAO can now be used inside one of the deepest DeFi ecosystems in crypto, with the same liquidity guarantees you’d get from a native token. For the agent economy thesis, this means agents running on Solana can now hold, transfer, and stake TAO without leaving their execution environment.

The Takeaway:

Track TAO volume on Solana DEXs over the next 90 days. If volume on Solana grows faster than volume on EVM bridges, the agent economy thesis is becoming more of a reality.

📡 Sara Hooker chose Bittensor

What Happened:

On April 30, SILX AI announced that Adaption Labs is partnering with them to advance the Quasar foundation models on Subnet 24. Adaption Labs is co-founded by Sara Hooker, the former Vice President of Research at Cohere and a veteran of Google DeepMind.

Her firm raised $50 million in seed funding to advance adaptive AI research. The partnership will see Adaption Labs supply high-quality adaptive datasets to train Quasar models, with model quality and competitive SOTA performance set as priorities from day one.

Why It Matters:

Hooker has real credentials at the top of her field. She could have gone anywhere, but chose Bittensor. This is a perception-layer signal that an investor in a corporate job, listening to colleagues debate whether decentralised AI is real, can finally point to.

The Quasar thesis is that decentralised training only matters if it produces models people actually want to use. Its the standard Hooker built her career around, and it’s something that Bittensor has been failing to clear in mainstream AI press.

The Takeaway:

Watch Quasar’s HuggingFace releases over the next 12 months. If their open models start showing benchmark numbers competitive with Llama or Mistral at similar parameter counts, that would be strong evidence that decentralized training on Bittensor can produce research-grade models.

If they underdeliver on benchmarks, the claim that decentralized training produces real research becomes harder to defend, though the subnet could still be valuable in niches like long-context recall or cost efficiency.

⚙️ Quiet

May had no headline benchmarks or major shipped features outside the protocol layer. This is somewhat to be expected in a month dominated by foundation work.

Summer time is when this has potential to fire. Teutonic’s targeted trillion-parameter run, the Rayon Labs subnet trio’s continued share of emissions, and Quasar’s first model releases are the candidates. Until one of them lands, this direction is a watch-and-wait.

Share

What sits with me most from May is the cadence. Five weeks from Covenant’s walkout to Conviction shipping live code is an unusual response time.

But its classic Bittensor, the kind of velocity you only see when the people building know exactly what’s broken and have already designed the fix.

The optimist’s read is that this network is more antifragile than its critics give it credit for. The skeptic’s read is that a fast patch can paper over a deeper governance problem, and we’ll only know which it is in the next couple of years.

Which of these moves you the most? The protocol moving fast on Conviction, or a researcher of Hooker’s calibre choosing Bittensor over a closed lab?

And if I missed something important from May, what would you add? Reply to this email and let me know. I reply to every single one.

Until next time.

Cheers,

Brian

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

And whatever new streaming service she signs up for when the latest season of The Summer I Turned Pretty mercifully returns to our screens.

Every couple of months we sit down at the kitchen table and ask the “what are we actually watching?” question, and one of the streaming services gets axed.

The Claude bill sits in there with all of them. Auto-renews on the same day every month. It has never made that list and probably should have, as there are a few things that have been bugging me about it.

I can’t pick the model architecture I’m using. Anthropic picks for me, from their range. No DeepSeek for the hard reasoning problems, Qwen for spreadsheets, FLUX for images. Just Claude, in whatever version they decide to ship that month.

Every prompt I type goes to a server in San Francisco run by a company whose terms of service could quietly change next month.

And I hold TAO. I’m an investor in the largest decentralized AI network in the world, and the AI I actually use every day is closed, centralized, and locked behind one private company’s API.

There’s a stack you can build today, in about 15 minutes, that fixes all of that.

It runs on Bittensor and it’s free or as little as $10 a month for an API. You don’t need to write a line of code.

The three layers in every chat product

Every AI chat product is three layers stacked on top of each other.

The model. The brain. GPT-5.1, Opus 4.7, DeepSeek-V4, Kimi K2.6, GLM-5.1.

The inference layer. The thing that runs the model on a GPU when you hit send. AWS, OpenAI’s own data centres, or in our case, Bittensor’s network of miners.

The interface. The chat box you actually look at.

ChatGPT (or Claude, Gemini, Perplexity) bundles all three and hands you the result. You can’t change the model or the inference and the interface is non-negotiable.

The anti-ChatGPT move is to pick each layer yourself:

• Open-source model.

• Bittensor inference (Chutes).

• Whatever chat UI suits you.

Where Chutes fits

Chutes is Subnet 64 on Bittensor. It’s what turns subnet emissions into AI you can actually talk to. Open-source models (DeepSeek, Kimi, GLM, Llama) get served by a network of miner-operated GPUs around the world. Validators score the output quality.

You hit send, somewhere a miner runs your prompt and you get an answer back. The TAO you hold is, in part, paying for the GPU you just used.

Chutes is OpenAI-compatible. Anything that speaks OpenAI’s API speaks Chutes. That’s the key that makes every chat app and every browser extension in this post work without code.

A lot of other Bittensor subnets rely on Chutes for their own inference too. It’s a layer the rest of the network builds on.

When a subnet app needs to talk to a model, there’s a good chance Chutes is in the chain somewhere. As an example, Affine (SN120) fine-tunes models on hard reasoning problems, then pushes winners to Chutes where anyone can query them. It does the training while Chutes takes care of the inference. That composition is what the ecosystem is becoming.

A bit of recent history matters here also: until early 2026, Chutes ran a free tier. 200 requests a day for nothing. It was a beta in disguise, kept alive by token emissions.

In February this year, they killed it.

On March 20, they published a blog post called From Volume to Value, explaining they’d cut 45% of their token volume because it was unprofitable and pushed heavy users onto pay-as-you-go.

And on May 19, they did it again. The $3 Base tier is gone. The minimum is now $10 a month.

For some people, that was bad news. For me, it’s the opposite.

Free things on the internet don’t last: real products do. Chutes is becoming a real product and that’s the signal a serious investor should care about.

Share

Sidebar: where does my prompt actually go?

On Chutes, it lands on a miner’s GPU somewhere in the world. Validators score miner output, but that doesn’t keep your prompt private from the miner running it.

The fix is Trusted Execution Environments (TEE). Many Chutes models run inside one. Think of it like a sealed envelope that only the recipient’s hardware can open. Chutes encrypts your prompt on your end, ships it to a confidential-compute GPU, and the lock only breaks inside the chip. No one in the chain reads it first.

To use it, pick a model with “TEE” in the name. deepseek-ai/DeepSeek-V3.2-TEE, zai-org/GLM-5.1-TEE, moonshotai/Kimi-K2.6-TEE, MiniMaxAI/MiniMax-M2.1-TEE. Same API, same setup.

ChatGPT, Claude and Gemini can’t promise this. The contents of your prompt are physically unreadable by the people running the model.

One caveat: TEE doesn’t make Chutes anonymous. Your account is still linked to your usage. It just makes the contents of your prompts unreadable by the miner.

The Basic Stack: Chutes Chat

The simplest version of the stack is one URL.

Open chutes.ai/chat That’s the whole stack. No account, no API key, no setup.

Inside that chat, you can:

• Switch models mid-conversation. DeepSeek for one question, GLM for the next, no friction.

• Use Chutes Search, their built-in web search. Launched recently, this works well enough to replace ChatGPT’s search for most questions.

• Generate images with Qwen Image 2512, one of the best open-source image models around.

• Upload files. PDFs, documents, whatever you’d attach to a ChatGPT prompt.

• Save conversation history and come back to it later.

What you can’t do is build a saved system-prompt library. There is no plugin ecosystem or agents. The UI is theirs, and if they redesign it next month, you’re along for the ride.

If you’re a casual user (a few prompts a day, the occasional image, no power-user setup), stop here. Pocket the $20 a month you used to send Sam Altman.

Bull case: zero friction, real model choice, free.
Bear case: you don’t own the interface, and Chutes’ UI doesn’t yet match the polish of ChatGPT/Claude/Gemini.

Get your API key

Ready to push past Chat? Time to set up an account and grab an API key.

1. Go to chutes.ai. Make an account.

2. Pick a tier. Plus is $10 a month, Pro is $20. Each gives you up to 5x its dollar value in usage. Start with Plus, you can upgrade later in under a minute.

3. Generate your API key under the API Keys tab. It starts with cpk_.

4. Treat it like a credit card number. Don’t paste it into Discord, don’t email it to yourself and (for me) don’t screenshot it for the newsletter.

Your API key is the thing that makes every chat app on the next page work without code.

And on Chutes, this is big value. Ten dollars buys you up to $50 of inference per month. If you go over, you top up with pay-as-you-go credits.

For a few chats a day on open-source models, this cap is invisible. If you’re like me, who is not a coder but can still go pretty hard on research, you’ll never go anywhere close to those limits.

Now you need somewhere to actually use that key. A chat tool that takes your prompts in and renders the outputs back nicely. The one I use is up next, with a few other good options after.

The Advanced Stack: Chutes + TypingMind

TypingMind is where the stack gets serious.

Why I picked it: this is a one-time license, you own it.
Price tiers range from $39-$99 with regular discounts. There is no recurring fee. You still pay Chutes per-token for the actual inference, but the interface is yours for life.

It officially documents Chutes as a supported provider. They have per-model setup pages on their site.

What you get over the basic chat: plugins, agents, vision, function-calling, custom personas, a prompt library you can build over months. Switch between Chutes, OpenAI, Anthropic, and the rest from the same window if you want a multi-provider setup.

Setup is eight fields.

1. Open TypingMind. Models → Add Custom Models.

2. Name: Model name (I prepend with Chutes to be sure where its coming from)

3. API Type: OpenAI Chat Completions API

4. Endpoint URL: https://llm.chutes.ai/v1/chat/completions

5. API key: your cpk_ key.

6. Model ID: copied from the API tab on your chosen model at chutes.ai/app e.g. zai-org/GLM-5.1-TEE

7. Context length: set per the model also from the API tab.

   

8. Advanced- Custom Headers: Authorization [tab] Bearer API KEY (insert your API key here)

Hit Test. If it says the endpoint is working, just hit Add Model and you’re in.

My three starter model picks on Chutes:

• DeepSeek V3.2 TEE for general work, drafting, summarizing.

• GLM 5.1 TEE for creative writing with hardware-level privacy.

• Kimi K2.5 TEE for help with spreadsheets, formulas, or any code-flavored task.

The open-source model landscape moves fast. A model you’ve never heard of can leapfrog the top spot in a quarter. Bookmark the Open LLM Leaderboard at llm-stats.com/leaderboards/open-llm-leaderboard and cross-reference what’s running on Chutes at chutes.ai/app. That combination keeps you on the best model available without thinking about it more than once a month.

Other shapes of the same stack

The OpenAI-compatible thing means TypingMind isn’t the only option.

Good alternatives include Msty for a desktop-first local/remote workflow, Chatbox for cross-platform chat with custom endpoints, Page Assist for in-browser sidebar chat, and self-hosted options like LibreChat and Open WebUI.

Two honest Chutes trade-offs

1. The frontier-model gap is real. On the absolute hardest tasks (deep research, multi-step reasoning, agentic coding), the closed frontier models still pull ahead of the best open-source models on Chutes. For everyday work (drafting an email, summarizing an article, brainstorming, generating an image), the gap is invisible. And you’re not spending $20 a month.

2. The where-does-my-data-go question is the other one. Honest answer: through a network of GPU miners around the world, ranked by validators. If that makes you nervous, use a TEE model and the contents of your prompts are physically unreadable by the miner running them. Centralized providers don’t offer this at all.

For the absolute best model on a hard, specific job, OpenAI, Claude and Gemini still win. For everything else (which is most things), the anti-ChatGPT stack wins on cost, privacy, and the part where you’re paying the network you’ve already invested in.

When Chutes isn’t enough

Three alternatives to know in case you need them.

OpenRouter. 300+ models, single API key, free tier with rate limits. Best for exploration. Worth knowing: Chutes is one of OpenRouter’s providers, so sometimes you’re paying Chutes through OpenRouter without realizing it.

Groq. A strong choice when speed matters most, thanks to its custom LPU hardware and free developer access.

Together AI. A good alternative if you want a broad but more curated model platform and a practical fallback when you need another provider.

Bittensor isn’t your only ride. But for someone who already holds TAO, it’s the one that makes you a customer of the network you already invested in.

The stack you’ve just built

Run the three layers one more time.

The model: picked. Open-source, your choice, switched mid-conversation if you want.

The inference: Chutes. Bittensor’s network. Your TAO is paying for the GPU.

The interface: chutes.ai/chat if you want simple, or TypingMind if you want to own a serious model-agnostic chat tool.

The deeper move is the one you might not feel for a couple of weeks. You stopped being a passive token holder: now you’re a customer of the network. Your TAO has a job.

One last thing. If you decide to grab TypingMind for the advanced stack, feel free to use my link. Both of us pick up 0.5GB of TypingMind cloud storage at no extra cost. Skip it if you'd rather, the recommendation stands either way.

Subscribe now

Until next time.

Cheers,

Brian

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

It was Saturday afternoon. I was refereeing two boys fighting over who gets to turn on the tap for the garden hose.

My phone buzzed. I got distracted, and the younger one ended up soaked.

A friend, also a TAO (and other cryptos) holder, texted in our “Crypto School” group chat: “Bro, what do you think of NEAR? 1M TPS, 46M monthly users, AI agents on the consumer wallet. This is getting real traction. Should I sell some TAO and buy NEAR?”

Quick disclosure before we go further: I’ve been aware of NEAR for years, long enough to have lost a pretty penny on NEAR NFTs in 2021, hoping to ride the next Solana boom. Now I’m a proud owner of NEAR Tinker Union and Antisocial Ape Club, which is to say I own priceless lessons on wasting money.

I get one version of this question every few weeks:

Should I be in TAO or NEAR? Or Render? What about ICP? Isn’t ASI an AI thing too?

The honest answer is: those projects aren’t competing for the same slot in your portfolio. They sit at different layers of the same stack.

Most “crypto AI landscape” pieces collapse them into a single horse race, and that’s how readers end up holding three accidental bets on the same layer and zero on the others.

So after the garden clean up job, I sat down with my coffee and answered the only way I know how.

With a map and a framework.

The Four Layers of Crypto AI

The AI economy is increasingly splitting into four layers: Compute, Intelligence, Agents & Data and Execution/UX. Each layer is seeing a centralized stack dominate today while decentralized networks are emerging to compete on incentives, access, and ownership.

Read it down, not across. A bet on Render isn’t a substitute for a bet on TAO. A bet on TAO isn’t a substitute for a bet on NEAR. They could all win or lose.

Let’s walk through each layer.

Subscribe now

Compute is the rented muscle

Render started as a decentralized GPU marketplace for 3D rendering and has aggressively expanded into AI inference and training workloads. Akash runs a “decentralized cloud” model: anyone can rent out spare compute. io.net aggregates GPUs on Solana. Nosana plays a similar game.

The pitch is the same across all of them: cheaper than AWS, more resilient, no single point of failure. The reported cost savings vs. centralized clouds run in the 50–80% range depending on workload, which is real money for AI startups burning through inference budgets.

This layer is the picks and shovels of crypto AI.

Where they win: Cost, access and resilience.

Where they lose: Frontier-scale training. The hyperscalers still have tighter GPU supply relationships and the kind of integrated networking you need for the biggest runs. Decentralized compute is best positioned for marginal workloads like inference, fine-tuning, edge deployment, or rendering pipelines.

Intelligence is where Bittensor lives alone

This is the part most readers underweight when they map the space.

Bittensor is the most established crypto network explicitly trying to build decentralized markets for model production at scale.

NEAR and ICP are more execution and app-layer plays, Render rents compute, and ASI is more about agent coordination and data infrastructure.

Bittensor is different. Miners compete to produce valuable model outputs, validators score them, and subnet economics determine where emissions flow.

With dTAO, that incentive system has become more market-driven, which is why I think of it as Proof of Intelligence: AI production coordinated by an open economy of 128 subnets rather than by one company’s balance sheet.

Where it wins: Specialization, open competition and economic alignment between the people who build the AI and the people who use it.

Where it loses: Onboarding. It is lacking the kind of seamless UX that makes ChatGPT effortless for someone who’s never opened a terminal.

I’ll come back to Bittensor in more detail in the spotlight section. The point for now: it’s the only one of the projects on this map that’s playing the intelligence game.

Agents and data are the ASI Alliance bet

The Artificial Superintelligence Alliance (ASI): Fetch.ai, SingularityNET, and Ocean Protocol who all merged under the FET ticker; are the full-stack agent and data play. Fetch handles autonomous agents that can negotiate and execute tasks, SingularityNET runs an AI services marketplace and Ocean handles tokenized data exchange.

The vision is good, but the execution is more fragmented than the marketing suggests and the merger is still in the awkward phase where three projects are operating under one banner without fully feeling like one stack. Its worth watching, but not yet a thesis I’d build a portfolio around the way I would around the intelligence layer.

Where it wins: Vertical integration. Owning agents, services and data in one stack is genuinely rare in crypto AI.

Where it loses: Coherence. The merger is still bedding in and you can feel the seams.

Execution/UX Layer: Where AI Apps Actually Live

NEAR is the most practical of the bunch. It’s an “AI-ready” L1 with developer tooling that lets you build with natural-language prompts, generate contracts and ship dApps without being deep in the weeds. Its high throughput, user-friendly and offers solid grants for AI builders. NEAR is trying to be the place where AI-powered applications get used.

Internet Computer (ICP) plays a different game. It runs AI models and agents fully on-chain via canisters, with ICP burned for compute cycles. This is total verifiability and sovereignty: nothing is off-chain, and everything is trustless. That makes it heavier infrastructure, but powerful when on-chain history and execution really matter.

Where they win: On-ramps. Both projects make AI-powered dApps actually usable.

Where they lose: Neither is producing intelligence. They’re hosting the apps that consume it.

One non-crypto note worth flagging

The most important decentralized AI player on the planet doesn’t have a token at all.

Hugging Face hosts more than two million open-source models. It’s the GitHub of AI.

It’s where Llama, DeepSeek, Qwen, and most of the open-weight world get distributed. Multiple Bittensor subnets host their models there too. Anyone can pull weights, fine-tune on their own hardware, and ship a product without ever touching a crypto network.

Why does this matter for your TAO thesis? Because it’s the alternative path.

Open-source AI is closing the gap with closed-source faster than most retail crypto investors realize. Bittensor’s value at the model layer has to clear a bar set by free, open weights you can run locally. Sometimes it does, when specialization, incentives, and economic alignment matter. Sometimes it doesn’t.

Knowing where Hugging Face fits on your map is part of being honest about where TAO fits.

Bittensor Subnet Spotlights

Three live subnets worth knowing by name, because “Bittensor subnets” as a concept is too abstract for most people to act on.

Chutes (SN64). Serverless inference. You hit an API, a subnet routes to the cheapest available miner, you get a response. The closest thing to a drop-in replacement for the OpenAI API that exists in crypto AI today. If you’ve ever wanted to test what decentralized inference actually feels like, this is the one to start with.

Apex (SN1). The flagship language-model subnet with agentic workflows, tool use, and chain-of-thought reasoning built to push inference quality past what single open-source models like Llama or Mistral can deliver alone. Clean SDK and API, explicitly positioned for builders. If Chutes is the fastest way to experience decentralized inference, Apex is the fastest way to build on it.

Teutonic (formerly Templar) (SN3) and the broader training subnets. Decentralized training of large models: the kind of work that produced the 72-billion-parameter run earlier this year that outperformed older Llama baselines. The governance picture around this layer of subnets has been turbulent in 2026, but the infrastructure work is still real and still worth tracking.

These three are a starting point and by no means a complete list. The point is that “subnets” stops being abstract once you’ve used one.

Two frameworks you can use this week.

Framework one: how to evaluate any crypto AI project

Three questions, in order:

• What layer does it sit at? If you can’t place it on the map cleanly, that’s a flag.

• Does the token actually do work? Is it required for the network to function, or is it bolted on as a pretend governance token?

• Is anyone using it? Look for real revenue, usage and workloads. Not pitch decks.

If a project clears all three, it’s worth deeper diligence. If it clears two, it’s worth watching. If it clears one or zero, it’s a vibe.

Framework two: no-code entry points

You don’t need to run a validator to participate. The simplest way to use Bittensor is to query a subnet. Chutes is one of the easiest on-ramps. I’m working on an upcoming walkthrough to demonstrate just how easy this can be.

The simplest way to participate economically is to stake TAO on a subnet whose work you believe in, which under dTAO directs emissions toward that subnet. Both can be done from a wallet without a terminal.

What this looks like in real life

A freelance writer today pays Anthropic $20–200 per month for drafting and fact-checking. Tomorrow: they query Apex (SN1) directly for comparable quality at lower cost (no corporate margins). If performance slips, they simply switch subnets with no lock-in or surprise hikes.

Apex is live today with working APIs and products, the main friction is awareness.

Or, picture a small business owner building a customer service agent: ASI coordinates the logic, Apex (SN1) powers the language model and NEAR handles the interface. No single vendor owns the stack.

Your compass from here

Crypto AI isn’t one bet. It’s at least four. The projects you’ve heard lined up against each other are doing different jobs.

The smart move is to know which ones you’re exposed to and why: Compute, Intelligence, Agents and Execution are different layers with different roles to play.

Bittensor’s edge isn’t that it’s the only AI token. It’s that it’s the only one playing the model game, coordinating the production of intelligence itself, not renting the hardware or hosting the apps.

If you’ve never actually used a subnet, do it this week. Start with Chutes. Then come back and tell me if the intelligence layer still feels abstract.

Share

Until next time.

Cheers,

Brian

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

My younger son is obsessed with testing things. He drops food off his plate not because he doesn’t want it, but because he wants to know what happens.

He’s learning cause and effect: broccoli and potatoes mashed on the floor for Daddy to clean up.

The experiments never stop, even if he doesn't know to call them that yet.

April felt like that, except nobody planned this month’s experiment. One operator pulled out of three subnets, dumped $10 million in TAO, and walked away. The community left behind picked up the pieces.

This month’s issue covers the four compass directions for April. Two of them dominated, another went quiet and the last one kept building like nothing happened.

How to Read This Compass

I’m organizing these updates by direction: North, East, South, West to give you a repeatable mental model for navigating Bittensor’s complexity.

North: Foundation 🏛️ (governance, protocol, core team, security events)

The structural layer. Changes here determine who controls the network and how decisions get made.

East: Capital 💹 (markets, volume, liquidity, integrations)

The economic layer. This is where price meets liquidity and institutional conviction shows up in the data.

South: Technology ⚙️ (shipped features, benchmarks, infrastructure, ecosystem health)

The execution layer. What actually works, what got built and where’s the technical proof.

West: Adoption📡(narrative, mainstream validation, external voices)

The perception layer. This is how the outside world sees Bittensor, from industry titans to normie investors.

Each month, some directions will dominate the signal. Others will stay quiet. The goal is to see which quadrants are moving and what that means for your strategy.

Subscribe now

🏛️“Decentralization Theater”: the governance crisis and the aftermath

• What Happened:

On April 9, Covenant AI announced it was leaving Bittensor.

They shut down three subnets: SN3, SN39, and SN81, dumped 37,000 TAO worth roughly $10.2 million into the market, and publicly accused co-founder Jacob Steeves of centralized control.

The specific allegations were pointed: that Steeves unilaterally suspended subnet emissions, overrode operator authority, and used large visible token sales as a tool to enforce compliance.

Founder Sam Dare called it “decentralization theater.”

TAO dropped 25% in six hours. $650 million in market cap gone with $9.1 million in long liquidations. Trading volume spiked to $1.72 billion on April 10, more than triple the monthly average.

I covered the allegations in my piece here.

• Why It Matters:

The crypto press covered the 25% drop. They missed what followed: community miners, with no guidance from founders and no central operator involvement, rebuilt all three subnets from open-source code. The chain continued running normally.

This was widely reported as a demonstration of the protocol’s antifragility and organic decentralization in action. New teams (e.g., Teutonic stepping in for Templar) picked up the work and the subnets resumed operations.

I’ve been writing about Bittensor’s antifragility thesis for a long time. This was the first real test at scale.

The protocol’s direct response was BIT-0011: the Conviction Mechanism. Subnet founders and stakers will lock alpha tokens across 30-day intervals to earn conviction scores. The staker holding the highest score gains ownership of the subnet.

And crucially, tokens locked during the active period can’t exit (rug) until the interval closes.

It formalizes at the protocol level exactly what the community just demonstrated organically: ownership through commitment, not through founder discretion.

• The Takeaway:

Watch BIT-0011 closely. After it goes live, if you’re delegating to subnets, understanding which teams are locking tokens needs to become part of your due diligence from here.

The governance layer fractured in April. The response revealed as much as the failure.

💹Institutional capital made its moves before the crash

• What Happened:

On April 2, a full week before Covenant AI went public, Grayscale and Bitwise both filed for spot TAO ETFs on NYSE Arca. An SEC review is underway with the decision window in August 2026.

On April 7, two days before the worst day of the month, Grayscale raised its TAO weighting inside its AI-focused fund to 43%. Its largest single-asset reallocation in the fund’s history. They did that before the broader market knew what was coming.

On April 20, BitGo and Yuma launched institutional-grade custody and staking for Bittensor subnet tokens: the first regulated, secure access point for large capital to get direct exposure to subnet economics. DCG’s Barry Silbert called it a meaningful step forward for institutional participation in Bittensor.

• Why It Matters:

This is the timing detail that nags at me.

What does it mean that the biggest institutional moves of April happened before, and immediately after, the worst governance crisis in Bittensor’s recent history?

It means institutional analysts are not reading the same headlines retail is.

They’re reading the staking data, revenue figures and structural supply dynamics post-halving. What they see apparently survives a 25% single-day drop without changing the thesis.

• The Takeaway:

The August 2026 ETF decision window is now a live catalyst on your calendar. A positive signal from the SEC unlocks a whole class of capital that currently can’t touch Bittensor even if they wanted to.

⚙️The execution layer didn’t notice

• What Happened:

The network generated $43 million in revenue across 128 active subnets in Q1 2026. That number didn’t move when Covenant AI left (their subnets were research-based).

Covenant operated three of Bittensor’s highest-emission subnets. Their exit was the single largest operator departure the network had seen. And the revenue baseline held because the network doesn’t run on any one team’s output. It runs on 128 markets competing simultaneously.

Looking forward: Teutonic (formerly Templar) is targeting a 1-trillion-parameter model training run by late May.

After Covenant-72B proved in March that decentralized training at 72 billion parameters wasn’t a whitepaper concept, a trillion-parameter attempt is a real escalation. It would be the largest decentralized training run ever attempted.

• Why It Matters:

When the biggest operator in the room walks out and the revenue line doesn’t flinch, it’s worth noticing. The timing of the potential trillion-parameter run also matters, as that milestone could land inside the ETF’s highest-visibility window during the SEC review calendar. If Teutonic delivers, attention from the AI world becomes a real possibility.

The output layer is accelerating independent of the drama layer.

• The Takeaway:

The trillion-parameter run and the ETF review window overlapping is not a coincidence to ignore. Successful decentralised training at that scale is exactly the proof of utility the institutional thesis needs to go mainstream.

📡No Mainstream Breakthrough

• What Happened:

Nothing meaningful here this month.

April had no Jensen Huang moments or Jack Clark quotes. No mainstream AI press picking up the Bittensor story. The governance crisis got crypto coverage but not AI coverage.

• Why It Matters:

The Covenant crisis was legible to crypto investors who already know what a subnet is, but it wasn’t legible to the broader AI audience that would need to understand why decentralized training matters to care.

This story requires context most AI journalists don’t have yet, so the translation gap is still wide open.

• The Takeaway:

Track mainstream AI media, not just crypto media. When Bittensor breaks through to that audience (and April wasn’t the month) you’ll feel it in adoption before you see it in the price.

Share

Which of these moves your conviction most: the protocol’s response to the governance crisis, or the institutional positioning that happened before it? I’m genuinely curious where your head is at after this month.

And if I missed something important from April, what would you add?

Reply to this email and let me know. I read every single one.

Until next time.

Cheers,

Brian

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

Last month, I watched my son discover that passwords exist.

He was trying to unlock my phone to watch Bluey. When Face ID failed, he demanded the passcode. I showed him the pattern: six digits I’ve used for years.

“But what if someone sees?” he asked.

“They’d need to watch very carefully,” I said.

“But what if they have a really good brain?”

Somehow my five-year-old explained the quantum threat in a nutshell: what if someone builds a brain so powerful it can guess every password on earth?

Right now, your TAO wallet is protected by math that would take a regular computer longer than the age of the universe to crack. But quantum computers don’t play by those rules. They can try millions of combinations simultaneously.

And according to Google’s recent research, that “really good brain” might arrive sooner than anyone planned for.

When Someday Became 2029

Let me start with what I actually understand (and what you need to know without a PhD in cryptography):

Every major blockchain: Bitcoin, Ethereum, Solana, Bittensor, uses an encryption standard called ECDSA to secure wallets and sign transactions.

What This Actually Means: ECDSA is like a really complicated lock. When you send TAO, you’re proving you have the key without showing the key itself. The math is so hard that even the fastest supercomputers can’t guess your key before the sun burns out.

Unless you have a quantum computer.

A sufficiently powerful quantum computer running Shor’s algorithm could reverse that math. We’ve known this since the 1990s. The threat has always felt distant, theoretical, a problem for a future generation of developers to solve.

Then the timeline compressed.

Google recently published research showing roughly a 20x efficiency improvement in running Shor’s algorithm, shifting credible threat estimates from 2035–40 toward 2029–30.

Solana’s Anatoly Yakovenko called it 50/50 within five years.

Nic Carter flagged something more alarming still: the possibility of a nine-minute on-spend Bitcoin attack, where a quantum computer intercepts a live transaction while it’s waiting to confirm within the standard ten minutes.

This isn’t just a far-off problem anymore. It’s a ticking clock.

The whole network has to be post-quantum before that machine exists.

Bitcoin has no roadmap.

Bittensor has no roadmap I’m aware of.

Ethereum has a public migration roadmap targeting 2029.

So when I started researching this piece, my first question was simple: Should I be worried about my TAO?

What Quantum Actually Threatens in Bittensor

Here’s where I had to stop and really understand the mechanics.

The panic is missing the target. Quantum doesn’t threaten the blockchain itself, the chain of hashed blocks is safe.

What quantum does threaten is authenticity: your signature proving you authorized a transaction. The ledger stays intact; the trust in who controls what does not.

Think of it this way: Quantum computers can’t rewrite the blockchain’s history. But they could steal your keys and write new history pretending to be you. This is a general blockchain problem, but Bittensor’s agile subnet architecture gives it tools to experiment and adapt faster than monolithic chains.

Blockchains like Bitcoin are like oil tankers. Changing course takes years of planning, coordination, and risk.

Bittensor is more like a fleet of speedboats. Each subnet can test different approaches, fail fast, and share learnings across the network.

When the quantum threat becomes real, that flexibility matters.

This is what clicked for me: Bittensor doesn’t need to be first to solve quantum threats. It needs to be fast to implement when someone else does. And it has infrastructure that could help test those solutions.

Share

The Quantum Subnets: Building Both Sides of the Equation

While everyone else is either panicking about quantum threats or hyping quantum computing, Bittensor has subnets building both.

QBittensor Labs runs two related subnets that caught my attention:

SN63 (Enigma): The Research Lab

Think of this as the quantum computing gym where miners train to get better.

Validators generate difficult quantum problems: things like peaked circuits and hidden-stabilizer tasks. Miners compete to solve them using simulation software called Quantum Rings, which runs on regular hardware (no actual quantum computers required yet).

The best solutions get scored on correctness and efficiency. Winners earn emissions.

What This Actually Means: Miners are getting really good at quantum algorithms right now, using simulated quantum hardware. When real quantum computers become more accessible, these miners already know how to use them.

SN48 (Quantum Compute): The Production Marketplace

This is where the training becomes real work.

At OpenQuantum.com, anyone can submit quantum circuit jobs: computational problems that might benefit from quantum approaches. Miners execute these jobs on simulators or actual QPUs (Quantum Processing Units) if they have access to them.

Then validators issue “proof-of-quantumness” (yes this is a real phrase) challenges to verify the results actually came from quantum computation, not classical shortcuts.

Both subnets are at full capacity: 256 miners registered in each. They’ve generated over 300 million proofs and the difficulty curve keeps climbing

Why This Matters for Quantum Threats:

If you want to test whether a new post-quantum signature scheme actually resists quantum attacks, you need quantum computers to try attacking it.

Right now, only IBM, Google, and a handful of others have that hardware. It’s expensive and you’re basically asking permission from the same companies whose centralized models you’re trying to escape.

SN48 is building a permissionless alternative. Need to simulate quantum attacks on a proposed new signature scheme? Submit the job. Miners with QPU access compete to run it with results verified on-chain.

The Structural Hedge Nobody Saw Coming:

QBittensor Labs is playing both sides of the quantum revolution:

If quantum becomes a commercial compute primitive (drug discovery, financial modeling, optimization), SN63 and SN48 are early infrastructure for those markets.

If quantum arrives as a cryptographic threat faster than anyone’s ready, the QPU access becomes directly useful for testing post-quantum defenses.

And because it’s happening on decentralized markets, the progress is visible. You’re not waiting for a Google press release. You can watch circuit complexity scores, fidelity rates, and proof validation metrics update in real-time on Taostats.

The Part I’m Still Figuring Out:

I don’t have strong conviction on whether these subnets are good investments yet. They’re early-stage infrastructure plays in a technology that’s still maturing.

I’m watching for signs these subnets keep delivering. Are circuit complexities climbing? Is proof-of-quantumness validation holding up under scrutiny?

If yes, this could become the R&D lab for Bittensor’s own quantum-safe future. Testing post-quantum signatures requires simulating quantum attacks. That’s exactly what this infrastructure enables.

Your Quantum Readiness Framework

Here’s the framework I now use to decide if quantum panic is warranted.

The Three-Signal System:

Signal #1:

Does Bittensor publish a post-quantum migration roadmap? TAO runs on Substrate with sr25519 signatures. This is the same threat class as Bitcoin and Ethereum. Ethereum’s targeting 2029. Bittensor has nothing public yet.

Here’s the test I ran this week:

1. I checked the last 8 weeks of OpenDev call notes for any mention of “quantum” or “cryptography upgrades”

   • Result: Zero mentions

2. I searched the Bittensor, Church of RAO and TAO Community Discord for “quantum” going back 90 days

   • Result: Numerous mentions from community members asking. No official response.

3. I looked at Polkadot’s quantum discussions (since Bittensor runs on Substrate)

   • Result: Polkadot has an active RFC discussing post-quantum primitives.

Signal #2:

Are institutional players treating this as real? Watch for:

• Exchanges announcing post-quantum wallet support.

• Large validators discussing migration plans.

• Foundation developers mentioning it in OpenDev calls.

Signal #3:

Does the timeline keep compressing? If Google publishes another efficiency breakthrough, or if someone demonstrates Shor’s algorithm on a meaningfully large key, the clock speeds up.

What Changes: Move your TAO to the most actively developed wallet (likely the one that responds first to quantum concerns). Consider keeping a larger percentage in liquid form rather than locked in long-term staking.

The Bittensor Advantage (And Why I’m Not Worried Yet)

Here’s what I’ve come to understand about Bittensor’s position:

The Speedboat Principle: Bittensor can upgrade faster than Bitcoin or Ethereum. When post-quantum standards solidify (likely NIST’s recommendations), Bittensor’s substrate-based architecture means implementing new signature schemes is a runtime upgrade, not a multi-year fork negotiation (looking at you, Bitcoin).

The Hidden Defender: Targon (SN4) is Bittensor’s private compute subnet. It’s not a quantum security tool, but it does help protect data while it’s being used, which can reduce the risk of it being stolen and decrypted later.

The Reality Check: Bittensor will need to upgrade its cryptography. Every blockchain will. But panic helps no one. The timeline is 2029-2030 and Bittensor’s architecture is designed for exactly this kind of coordinated evolution.

The story is quieter than the noise suggests.

We have time. We have tools. We don’t have a comprehensive plan yet, but we’re watching the right signals.

Your Three-Step Action Plan

This Week: Check your subnet holdings for any mention of quantum preparedness. Just search their Discord or GitHub for “post-quantum” or “quantum resistant”. Note which ones are silent and which are discussing it.

This Month: Bookmark these two resources:

1. Polkadot’s RFC discussions on post-quantum primitives (since Bittensor inherits Substrate’s cryptography roadmap)

2. NIST’s Post-Quantum Cryptography standards page (to track when official recommendations are finalized)

This Year: Watch for Signal #1 from my framework above. If Bittensor doesn’t publish a post-quantum migration roadmap in 2026, while Ethereum has one for 2029, that’s worth asking about loudly.

The quantum threat is real and the timeline is compressing.

But we’re not helpless. We’re just early in a process that every blockchain is navigating together.

And unlike Bitcoin’s oil tanker, Bittensor’s speedboats can change direction fast when the course becomes clear.

Subscribe now

Until next time.

Cheers,

Brian

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.

On April 9th, Sam Dare published a letter accusing Bittensor of being a lie.

It was long, carefully written, and timed for maximum damage. It arrived alongside a purpose-built whistleblower website and a coordinated dump of $11 million in subnet alpha tokens, most of it sold the night before the letter went public and before his own community had any idea what was coming.

Let me walk through what happened, what I think it means, and why I’m not selling.

Reconstructing the Exit

For readers newer to this ecosystem, some context.

Covenant AI was not a fringe player. The team operated three subnets on Bittensor: Templar (SN3), Basilica (SN39), and Grail (SN81). Templar was the crown jewel: a decentralized pre-training environment that produced Covenant-72B, a 72-billion-parameter model trained across more than 70 independent contributors on commodity hardware and coordinated through the Bittensor incentive layer.

The achievement was real. NVIDIA CEO Jensen Huang praised it publicly. Chamath Palihapitiya mentioned it on the All-In podcast. TAO rallied 90% in the month that followed. It was a mainstream moment for Bittensor, the proof of concept the whole ecosystem had been working toward.

Then, on April 9th, Sam Dare, the founder of Covenant AI, posted his open letter announcing the withdrawal of his three subnets from the Bittensor ecosystem. TAO dropped 27% in two hours with nearly $900 million in market cap evaporated.

Dare’s letter laid out four specific allegations against Bittensor co-founder Jacob Steeves, known universally in the ecosystem as Const.

He alleges Const had:

1. Suspended emissions to Covenant’s subnets.

2. Removed the team’s moderation capabilities over their own community channels.

3. Unilaterally deprecated their subnet infrastructure.

4. Applied economic pressure through large, visible token sales timed to moments of conflict.

Explosive claims, but before we get to Const’s rebuttal, it’s worth understanding how Bittensor’s emission mechanics actually work. Because this is the technical detail that makes or breaks the most serious accusation.

Under dTAO, emissions are determined entirely by an on-chain algorithm. There is no admin key or pause function.

Each subnet’s share of daily TAO emissions is set by net TAO flows into its liquidity pool calculated automatically by the protocol, in the same way for every subnet without exception.

What Const can do, what any TAO holder can do, is buy or sell alpha tokens in those pools. When he sold his positions in Covenant’s three subnets, that reduced TAO in the pools and affected the flow-based emission calculation. The same way anyone’s sell would.

This is market mechanics at work. Emissions cannot be modified in any other way. Dare’s central accusation requires a technical capability that does not exist in the protocol.

Const addressed each claim directly. He stated:

1. He has no authority to pause emissions (as described above).

2. He did not remove Dare’s moderator role outright, but removed his ability to delete posts and later restored it. Community member Alex DRocks corroborated in real time, it was Dare who deprecated his own Discord channels, and the posts deleted were questions from community members asking why SN39 was replicating work another compute subnet had already done. Dare was deleting scrutiny.

3. He’s unsure what deprecating their subnet’s infrastructure actually means (yet to be confirmed).

4. His token sales (less than 1% of what he had invested in Covenant’s teams) were made because Covenant’s subnets were not running and were operating on near 100% burn code, affecting emissions the same way any buy or sell would.

The Rug Frame: Reading Sam Dare’s Exit Clearly

Now pay attention to the choreography.

On the same day as Dare’s open letter, a newly launched site called Tao Papers published internal documents and on-chain forensics from multiple whistleblowers.

The headline claim: of 41 network upgrades carried out between 2023 and 2026, 38 were proposed, signed, and deployed entirely from infrastructure controlled by Const, with the other two signers approving each upgrade within minutes and no public discussion on record.

Taken at face value, that’s worth examining properly.

But it is not spontaneous. You don’t build a forensic website overnight.

The fact that the open letter and the whistleblower site launched simultaneously tells you everything about the intent. This was a campaign planned and timed for maximum narrative impact on exit.

And the exit was the most malicious part.

Prior to the public announcement, Dare had already sold 37,000 TAO ($11m) worth of subnet alpha tokens across Templar, Grail, and Basilica.

The dump hit retail holders tied to Covenant’s projects before they had a single word of warning.

TAO’s sales volume had already reached its highest level since December 2024 in the 24 hours before the letter went public.

This was not a principled departure, but an exit strategy, wrapped in governance language, timed for maximum extraction.

The Tao Papers data may contain real information about governance concentration, but its framing as independent corroboration collapses the second you notice it arrived gift-wrapped, synchronised with the open letter, hours after the pre-announcement dump had already cleared.

Share

The Legitimate Critique Underneath the Drama

Let me be honest here, because this newsletter does not earn your trust by only telling you what you want to hear.

The Tao Papers governance data, whatever you think of Sam Dare’s motives, raises a legitimate question about Bittensor’s current governance structure.

38 of 41 network upgrades originating from a single founder’s infrastructure, with co-signers rubber-stamping within minutes, is not the picture of distributed governance.

The triumvirate structure, as currently operating, functions more as a legal architecture than a genuine check on centralized decision making.

Const has not explicitly addressed that structural critique. He responded to Dare’s specific allegations point by point. He did not address the broader claim that meaningful governance authority has not yet transferred.

I think that’s intentional, because his response is in line with a design pivot. His statement in the aftermath of the exit:

“This will prove to birth the first subnets on Bittensor that run headless and as true commodities.”

These are the words of a builder saying the criticism identified a real structural dependency, and the answer is to architect it away.

The reality is, decentralization is a gradual process for the good of the protocol.

Every major network retains concentrated authority early to move fast and prevent paralysis, then systematically transfers power as architecture matures. That transition creates friction. Big players in the ecosystem: subnet founders, major investors, validators, all have different timelines for when centralization should end.

When those views collide with financial stakes this large, exits like Dare’s become possible.

Const stepped down as CEO of the Opentensor Foundation two months ago, explicitly to push decentralization forward. This episode accelerates that trajectory.

Revealing a structural weakness is different from being terminally broken. Every serious protocol has gone through this and now we will see how the Bittensor ecosystem responds.

The Road to Full Decentralisation: What Changes Now

The specific structural reform announced in the aftermath of this crisis is worth understanding clearly, because it directly addresses the failure mode that Dare exploited.

Bittensor is introducing lock-based subnet ownership.

The mechanics: ownership of a subnet will be determined by a team’s long-term economic commitment, not just their technical presence.

Investors will have transparent, advance notice if a subnet owner unlocks their tokens, allowing the market to reprice before founders can use their communities as exit liquidity.

The system will also allow investors to shift their staked capital fluidly to alternative management teams.

If lock-based subnet ownership had existed on April 9th, every holder in Templar, Basilica, and Grail would have seen the unlock coming before the dump hit.

The market would have repriced on Dare’s decision to exit, not on his public announcement of it. The $900 million wipeout would have been a controlled, visible repricing rather than a midnight ambush.

Autonomous, headless subnets, running as true commodities without a single operator as a point of failure, are the longer-term architectural answer. The Covenant crisis exposed a network that is mid-decentralization, under pressure and now moving faster.

Stillcore Capital (Bittensor Fund) partner Mark Jeffrey put it plainly:

“Bittensor is a quite a lot more than Subnet 3, and TAO will carry on fine without it.”

He is right. And multiple big players have come out in support of Const in the aftermath of this incident. But I’d go further.

Bittensor is stronger for this than it was two weeks ago. The lock-based model and the headless subnet architecture were coming. This crisis brought them forward.

What Happens Now for Covenant and the Three Subnets

Covenant operated three subnets: Templar (SN3), Basilica (SN39), and Grail (SN81). Templar was the crown jewel — the pre-training environment that produced Covenant-72B. Basilica was focused on decentralised compute. Grail handled more complex task-driven AI workloads. All three now show as deprecated on Taostats, removed from active participation in the network.

The team will continue developing Covenant-72B off-network, with new project announcements promised but no ecosystem named.

They’ll be building without the TAO incentive layer, without the established subnet infrastructure and carrying the reputational weight of what much of the community now views as a coordinated exit dump.

New investors in whatever comes next will be asking the same question Bittensor holders are asking today: who’s really in control, and when are they planning to leave?

For Bittensor, the three subnets now need transition to new operators or a headless model, and the process is untested at this scale. If the transition is messy, it reinforces the single-point-of-failure narrative.

If it goes smoothly, it validates Const’s “true commodities” vision: that subnets should eventually run without any dependency on their founding team.

That transition is the first real test of everything Const said would come next. I’ll be watching it closely.

What the Price Action Was and What It Wasn’t

When a high-profile governance headline breaks alongside a prominent team exit and a visible token dump, the market reacts predictably.

Traders operating on shorter timeframes, whether that’s algorithms, momentum players, or simply anyone with a stop-loss or a risk tolerance that got breached, will sell into uncertainty.

Not because they’ve evaluated whether Sam Dare’s accusations are credible or Const’s rebuttals hold up. Because uncertainty itself is the signal.

Even longer-term investors who might otherwise hold through volatility often reduce exposure when a story that felt clean suddenly feels complicated. They’re not abandoning their thesis necessarily, they’re stepping back until the fog clears.

That is also rational behavior. Preservation comes before conviction when the facts are in dispute.

Your job, if you hold TAO with conviction, is not to front-run these reactions or predict when sentiment will stabilize.

It is to ask different questions entirely:

• Has the network stopped working?

• Has the underlying technology been invalidated?

• Has anything happened here that changes the fundamental thesis?

• Has the team with the long-term vision left, or the team that just cashed out?

Those are the questions that matter. The price tells you what the market did on April 9th. It does not tell you what TAO is worth in 2027 and beyond.

Ecosystems Have Always Had Bad Actors: The Price of Upside

Here is what I want non-crypto readers to understand. And what I want crypto natives to remember.

Every transformative decentralized network has had this moment. More than once.

Bitcoin didn’t emerge from consensus and goodwill. It emerged from years of the block size wars, a community riven between competing visions, ultimately splitting into Bitcoin and Bitcoin Cash, and then Bitcoin Cash splitting again into Bitcoin Cash and Bitcoin SV, driven by Craig Wright’s years of fraudulent Satoshi claims, litigation, and manipulation.

Those were genuine existential moments. Bitcoin absorbed all of it. The network is worth over a trillion dollars today.

Ethereum’s defining early crisis was the DAO hack in 2016. A vulnerability was exploited, $60 million drained, and the community faced an impossible choice: honor the immutability they had promised, or fork the chain to recover the funds. They forked. Half the community followed.

The other half stayed on the original chain and called it Ethereum Classic. People said Ethereum would never recover its credibility. ETH is now the settlement layer for trillions of dollars of financial activity.

The pattern is consistent. Early-stage decentralised ecosystems attract the most visionary builders and the most opportunistic actors simultaneously, because the rules are still being written.

Power is concentrated by necessity before it can be distributed by design. And some people see that window and take it.

This is the cost of being early to asymmetric upside.

If you wanted zero drama and full governance maturity, you would be buying index funds. You are here because the upside on the early innings of decentralized AI infrastructure is unlike anything in a passive portfolio.

That upside does not come without episodes like this one.

Reading the Aftermath

What This Builds

Lock-based ownership is a genuine structural improvement, not a PR response. The headless subnet architecture, if it delivers, creates a more resilient and more trustworthy system than what Covenant left behind. The ecosystem survived one of its largest subnet’s exit without collapse.

What to Monitor

Watch two things.

First: how cleanly SN3, SN39, and SN81 transition. Smooth operator handoffs would validate the true commodities thesis and demonstrate that Bittensor’s incentive layer is more durable than any single team.

Second: where Covenant deploys next. If they name a competing ecosystem, the protocol-switching cost argument becomes live and worth monitoring.

What Remains Unresolved

The triumvirate governance concentration is real and not yet fully resolved. Const’s point-by-point rebuttals were credible. The structural critique has not been fully answered. Further prominent departures, especially if they cite similar governance concerns, would compound the credibility damage faster than any structural reform could repair it.

What Dies Here

The narrative that any single subnet team is Bittensor. This moment definitively ends that story, which is uncomfortable in the short term and healthy in the long term. The ecosystem should never again be this exposed to a single founder’s exit.

The Morning After

There is a difference between having a genuine grievance and performing one for maximum effect.

Real concerns deserve real processes. You don’t burn the room down to prove the room had a fire hazard. And when someone times their concern to the moment of maximum personal extraction, you learn important things about them more so than their peers.

The Bittensor network is still running. The technology that produced Covenant-72B still exists, built on infrastructure that no departing team can take with them.

The builders who stayed chose to stay. And the structural reforms now in motion were always where this ecosystem was heading.

The price will do what prices do. I’m watching the transitions.

Subscribe now

Until next time.

Cheers,

Brian

Disclaimer: This is not financial advice. I am a writer documenting the Bittensor ecosystem. Always do your own research.