Bittensor Intern Proposes “Fractal Emissions” to Give Subnets More Runway

A community member, Bittensor Intern, has proposed a new approach to Bittensor emissions that would allow subnet owners to choose how quickly they consume their available rewards.

The idea, called “fractal emissions,” is based on a simple premise: not every subnet needs the same amount of funding at the same stage. A compute-heavy production subnet may need maximum emissions to scale, while an early-stage research subnet may benefit more from receiving fewer rewards over a longer period.

Under the proposal, subnet owners could voluntarily accept 100%, 50%, 25%, 12.5% or lower of their available emissions. Choosing a lower rate would reduce TAO and alpha issuance, slowing dilution and supply pressure while extending the period over which the subnet can use its subsidy.

The proposal also suggests adjusting deregistration scoring based on the selected rate. For example, a subnet accepting 25% of its available emissions would receive a 4x adjustment to its moving alpha price for deregistration purposes. This is intended to give lower-emission subnets more time to demonstrate progress.

The broader argument is that Bittensor should not force every subnet to consume emissions at the same speed. As TAO’s value grows, a fixed emission allocation can become increasingly large relative to what a subnet actually needs, potentially creating unnecessary sell pressure and dilution. Giving subnet owners control over their emission rate could make emissions more responsive to actual demand.

Bittensor Intern also argues that fractal emissions could work alongside more subnet slots and a softer emission-gate curve, allowing more experimental projects to compete while keeping the overall cost of supporting them under control.

The proposal is still an early framework, with details such as the exact deregistration adjustment, cooldowns, and interaction with the emission gate left for further modelling and community discussion.

The key idea: instead of asking every subnet to consume its allocation at the same speed, let them choose their funding pace, and let the market determine whether that choice was justified.

Babelbit as a case study

The proposal uses Babelbit (SN59) as a recent example of why the “fractal emissions” could work. Babelbit was deregistered roughly one week after announcing its Dubbing API, the first product built on its Language Transformation infrastructure, following a major multilingual-model breakthrough just days earlier.

Under the proposed model, Babelbit could have selected a lower emission rate during its earlier research phase. At 50% emissions, for example, it would have received fewer rewards and issued less alpha, but its deregistration score would have received a 2x adjustment. Depending on the surrounding rankings, that could have given the team significantly more time to turn its technical progress into adoption and market support.

The point is not that Babelbit would necessarily have survived under fractal emissions. Rather, its experience illustrates the trade-off the proposal is trying to introduce: a subnet could voluntarily give up rewards today in exchange for more time to prove its thesis.

Read the full proposal and join the discussion →