A line of code pushed to GitHub on Wednesday could reshape the economics of Bittensor, the decentralized AI network behind the TAO token. The proposal is called Root Reborn, authored by a developer using the pseudonym “unconst,” and it transforms validators into something closely resembling fund managers.
Why should you care about a proposal still running on a testnet? Because it touches the most sensitive lever in the entire system: who decides where the money goes. Bittensor distributes roughly 7,200 TAO per day, according to on-chain data from Glassnode, worth over $1.8 million at current prices. That “who” carries real weight.
How Bittensor Works Today, and What Root Reborn Breaks
Bittensor is built from dozens of subnets, each a marketplace for a different AI task, from language model training to image generation, each with its own “alpha” token. Staking TAO on the root layer earns a yield. The problem lies in how that yield gets paid: the network automatically sells the alpha tokens of each subnet and converts them into TAO. The result is constant selling pressure on the very assets the subnets are built on, persistently dragging their prices down.
Root Reborn flips the logic. Instead of selling everything, each validator selects a basket of subnets to support, much like a fund manager picking stocks. The yield that was previously liquidated gets reinvested into the chosen subnets, held as a compounding basket over time, and returned to the validator as stake. The staker still receives yield and can convert back to TAO whenever they choose.
What concretely changes:
- The constant selling pressure on subnet tokens turns into net buying, supporting their prices.
- Validators stop being passive yield pipes and become active curators.
- Subnets deemed valuable attract fresh capital; those judged weak are starved of it.
A New Idea on an Architecture Already Rebuilt Once
Functionally, this is not the first time Bittensor has rewritten its own rules. In February 2025, Dynamic TAO (dTAO) arrived, stripping a committee of validators of the power to allocate emissions and handing it to the market. Each subnet now has an alpha token and a liquidity pool, with price determining how much each subnet receives, as outlined in the official project documentation. Root Reborn builds on exactly this layer, adding human judgment on top of market automation.
When we first built the Opentensor Foundation to support Bittensor we were responsible for everything:
— const (@const_reborn) June 17, 2024
- The Python API.
- The Blockchain code.
- The Explorer/Scanner.
- The Validator nodes.
- The Subnet protocols.
- The CLI.
- Fundraising.
etc. etc.
Today, the majority of…
The broader context matters: decentralized AI has moved well beyond narrative. An independent empirical analysis published on arXiv (paper 2507.02951) measured concentration and flow dynamics across the network, and projects like the Templar subnet have trained models with tens of billions of parameters entirely through distributed computing. The convergence of AI and blockchain infrastructure is accelerating, and Bittensor sits at its center.
Who Actually Earns at Every Block?
To understand why the stakes are high, look at how TAO emissions are split at each block. Validators and AI miners each receive 41% of every emission, with subnet creators taking the remaining 18%, according to the empirical analysis in arXiv paper 2507.02951 and Bittensor’s own documentation. Root Reborn targets that 41% validator share, seeking to transform passive recipients into active allocators. That ambition explains why the proposal is generating debate long before it touches the mainnet.
How TAO Emissions Are Split at Each Block
Source: empirical analysis arXiv 2507.02951 and Bittensor documentation
- AI Miners: 41%
- Validators: 41%
- Subnet Creators: 18%
Caution is warranted. Root Reborn remains testnet code, not a live change, and an initial automated review has already flagged two serious concerns: an upgrade migration path that could stall under large data volumes, and a payment routing issue that risks penalizing stakers. The direction, though, points to something larger than any single protocol. The same race to build AI infrastructure is forcing every network to ask who allocates capital and by what logic. Pure market mechanism or human judgment: Bittensor is trying to hold both at once. If the model works, the validator-as-fund-manager template won’t stay confined to one network for long.
