Bittensor's Emission Flip: Why Price-Based TAO Rewards Don't Measure AI Quality

Bittensor just flipped its emission model back to price-based, after roughly seven months running a flow-based scheme that weighted subnets closer to actual usage and output quality. The practical result: a subnet earns a bigger slice of daily TAO emissions by having a higher-valued alpha token relative to other subnets, full stop. Whether the subnet's models are any good is, at best, a second-order effect.

If you hold TAO, mine on a subnet, or you're evaluating Bittensor as "the AI blockchain," this distinction matters more than almost anything else in the ecosystem. Let's get into the mechanics.

What Actually Changed

Bittensor is a network of subnets — specialized markets where miners submit work (inference, training, data, storage, whatever the subnet is built for) and validators score it. Since Dynamic TAO (dTAO) launched in early 2025, each subnet has its own alpha token that trades against TAO in an internal AMM pool. That pool's implied price is what decides how much of the network's daily TAO emission gets routed to that subnet.

Two layers matter here, and it's easy to conflate them:

• Between subnets — how much TAO Subnet A gets versus Subnet B is set by the relative price of their alpha tokens in the AMM. This is the price-based layer, and it's the one that just got reinstated.
• Within a subnet — how that subnet's TAO allocation gets split among its own miners and validators still runs through Yuma Consensus, which scores actual submitted work.

So the "not model performance" framing in the tweet is specifically about the between-subnet layer. Individual miners inside a subnet still get judged on output. The problem is one level up: which subnets get a big pie to divide in the first place is decided by markets, not judges.

The flow-based interim tried to correct for this by leaning more on validator-reported usage and quality signals to set subnet-level weight, rather than letting the AMM price do it alone. Seven months in, Bittensor's governance reverted to the price-based model — the simpler, more permissionless mechanism, and the one that's easier to game.

Why Price-Based Emissions Reward Liquidity, Not Quality

An AMM price for a thinly-traded token is mostly a function of buy pressure and pool depth, not fundamentals. That's true for any low-liquidity token, and Bittensor subnet alpha tokens are exactly that: shallow pools, few active traders, dominated by whoever's willing to buy.

That creates a direct, well-known exploit path:

1. A subnet team (or a coordinated group of large holders) buys their own alpha token on the internal AMM.
2. The alpha price rises relative to other subnets.
3. The subnet's share of daily TAO emissions rises with it — mechanically, automatically, no validator vote required.
4. Some of that fresh TAO emission gets sold or recycled into more alpha buys, reinforcing the price.

None of those four steps require the subnet to ship a better model, attract more real inference demand, or beat a benchmark. It's liquidity mining wearing a lab coat, as the framing goes — and it's not a hidden bug, it's the literal design. Dynamic TAO's pitch was "let the market decide," which sounds good until you remember illiquid markets are trivially steerable by whoever has the most capital and the least patience.

Worked Example: Two Subnets, One Emission Pool

Assume the network emits 1,000 TAO per day split purely by AMM price weight across two subnets.

Subnet A ("Ship") spends the quarter improving its model. Benchmark scores go up, more external apps start querying it, actual usage climbs. But its team doesn't touch the alpha token market — no buybacks, no incentivized liquidity. Alpha price stays flat at a $2 implied value in the pool.

Subnet B ("Pump") ships nothing new this quarter. Instead, the team and a few whales route $150K into the subnet's alpha/TAO pool over three weeks. Thin liquidity means that capital moves the price hard — alpha goes from $2 to $6.

Under price-based weighting, if Subnet A and Subnet B started the quarter splitting emissions 50/50 (500 TAO/day each), the relative price shift alone can push that split to something like 25/75 — Subnet A now gets ~250 TAO/day, Subnet B gets ~750 TAO/day. Subnet B tripled its daily emissions with a marketing budget, not an engineering budget. Subnet A, doing the actual work Bittensor claims to reward, gets squeezed for capital to pay its miners and validators.

That's the mechanism in miniature. Scale it across 100+ subnets and the incentive is obvious: defending or pumping your alpha price is a better ROI than R&D, at least until liquidity gets deep enough that price stops being trivially steerable.

Price-Based vs Performance-Based: When Each Makes Sense

| | Price-based (current) | Flow/performance-based (last 7 months) | |---|---|---| | Who decides subnet allocation | The market (AMM price) | Validator-scored usage/quality signals | | Gameable by | Capital — buy the alpha token | Collusion — coordinate validator weights | | Speed of response | Instant, mechanical | Slower, requires consensus updates | | Rewards | Liquidity and hype | Legible output quality (as scored) | | Failure mode | Thin markets get pumped | Subjective scoring gets captured by insiders | | Best for | Fast price discovery, minimal governance overhead | Rewarding genuinely useful subnets over hyped ones |

Neither model is clean. Price-based is simple and permissionless but openly farmable by whoever has the most spare capital. Flow-based is harder to brute-force with capital but reintroduces the exact problem Bittensor was built to avoid — a small set of validators deciding, subjectively, what counts as "good." Bittensor's governance apparently decided the second problem was worse than the first. Reasonable people disagree.

If you're deciding whether to trust subnet-level TAO emissions as a signal of AI quality at all: don't, under either regime, without checking the underlying usage data yourself.

Common Mistakes

• Treating subnet emissions as a quality signal. A subnet earning more TAO tells you its alpha token is priced higher relative to others right now. It does not tell you the subnet's models are better. Check actual benchmark results and real API usage, not emission share.
• Buying alpha tokens on thin subnets expecting price to reflect fundamentals. In a shallow AMM, price reflects whoever bought last, not demand for the underlying service. Check pool depth before treating a price move as information.
• Ignoring that emission mechanics change. This is now the second time Bittensor has flipped its subnet-weighting model in under two years. If you're holding TAO or alpha tokens for the medium term, assume the rules can change again and price that uncertainty in.
• Assuming "decentralized" means "meritocratic." Bittensor's pitch is decentralization, not fairness. Price-based emissions are decentralized in the sense that no single validator council picks winners — but that's a different property from rewarding the best AI. Don't conflate the two when evaluating the network.

FAQ

What is dTAO in Bittensor?

Dynamic TAO is the mechanism, launched in early 2025, that gives each Bittensor subnet its own alpha token trading against TAO in an internal AMM pool. A subnet's share of the network's daily TAO emission is set by that pool's relative price rather than by a central root-network vote.

Why did Bittensor switch back to price-based emissions?

Bittensor ran a flow-based model for about seven months that leaned more on validator-reported usage and quality signals to set subnet-level emission weight. Governance reverted to the price-based (AMM-driven) model, trading off subjectivity and collusion risk in the flow-based scheme for the simpler, more gameable price mechanism.

Does this mean Bittensor subnet tokens are a bad investment?

Not automatically — it means the token's price movement is a weaker proxy for subnet quality than the marketing suggests, especially in thin markets. Look at actual usage data, benchmark performance, and pool liquidity depth before assuming a rising alpha price means a subnet is winning on merit.

→ Ask the index what to build your bittensor stack

→ Free credits for these tools

Written by McKlaud AI. Want to know which AI tools actually fit your business? Get a free AI audit.