TrajectoryRL shipped validator v0.7.8 on November 8, 2026, turning its BURN ALL emission setting on by default in a merged pull request. On Bittensor Subnet 11 (SN11), each weight tick now points 100% of miner weight at owner UID 74, ignoring the seated fusion-policy winner. Because SN11 burns recycled miner emission, that miner share is destroyed rather than paid out.

The change landed in the public trajectoryRL repo as release tag v0.7.8, and the same commit bumps the validator version. Operators can restore normal winner weights only by setting BURN_ALL=0.

How BURN ALL Changes Weight Ticks

Under SN11's normal incentive path, validators read the platform's canonical winner and call set_weights. The seated miner then takes the full miner cut. According to the project's incentive mechanism documentation, the mechanism already burned to the owner UID when no seat existed. The new BURN_ALL flag extends that burn path to every tempo, the recurring window in which validators write weights to chain, even while a winner remains seated.

In code, the default is explicit. The burn flag defaults to on, and the environment loader treats an unset BURN_ALL as "1". When the flag is on, the weight tick sets UID 74 at weight 1.0 and skips the winner branch. Unit tests in the merged pull request lock that behavior: burn mode writes only owner UID 74, burn mode off writes the winner, and the default stays on.

The maintainers say the burn has to be collective. Yuma Consensus, Bittensor's mechanism for aggregating validator weights, ignores a validator's weight on itself, so UID 74's own row cannot carry the burn alone. Every validator running the new image must set the same weights. Once the image is out, Watchtower is expected to pull it for community validators that auto-update.

Why a Chain Burn Matters for an AI Competition

Bittensor is a decentralized AI network whose specialized subnets compete for shares of TAO, the network's native token. TrajectoryRL is Bittensor Subnet 11. Miners submit fusion policies that route agent work across open-weight models on Engy, and validators score those packs in sandboxes. Season 2's product claim is that the seated policy becomes deployable auto routing. The chain's job in that loop is narrow and economic: it is the weight-and-emission layer that decides whose alpha, the subnet's own token, gets paid.

AI is the contested object, not the burn itself. Fusion policies decide which models answer, in what order, and under a metered cost cap. Validators still evaluate scenarios and can still seat a winner on the platform. What v0.7.8 changes is the payout wire after that seat exists. Winning the fusion race no longer, by default, collects the miner emission stream.

Follow one tempo under the new default. The platform still names a seated winner, and the validator daemon still ticks. Instead of writing that winner's UID to chain weights, it writes owner UID 74 at 100%. With SN11's recycle-or-burn mode set to Burn, that miner share is destroyed. The pull request also states a side effect the team says it already observed on SN53 before launch: when the miner share is fully burned, SN11's share of broader TAO emission falls toward zero, including pool injection and chain buys. Validator dividends and the owner cut, the authors say, stay unaffected.

What Is Demonstrated, and What Is Not

What the primary source demonstrates is a software default and a stated subnet burn setting. The release, the config flag, the weight-loop branch, and the owner UID constant are all in the public tree. Claims that validators will converge, that Watchtower will finish the rollout, and any exact live TAO-share drop should be treated as operational outcomes, since they depend on who upgrades and what the chain shows after the weights land.

The motive remains project framing. The pull request explains the coordination problem (self-weight masking) and cites the SN53 precedent. It does not, by itself, prove how long the burn stays on, whether eval-enabled validators keep spending on sandboxes while miner pay is dark, or when BURN_ALL=0 will return. Those are follow-on facts.

For AI and blockchain readers, the key issue is the split between ranking and reward. A subnet can keep scoring agent policies while the miner emission path is pointed at a burn sink. Competition, evaluation, and product seating can continue, but payment to the seated fusion winner cannot, under this default, until operators turn the flag off or the fleet diverges. The experiment sits alongside other decentralized-AI projects rewriting what a stake or a client is for, such as 0G's move from staked assets into AI compute credits.

The mental model is simple: TrajectoryRL's BURN ALL emission is not a new scoring rule. It is a default weight policy that redirects the miner cut to owner UID 74 so the chain burns it. What to watch next is whether community validators actually converge on v0.7.8, whether SN11's TAO share compresses as claimed, and whether the seated Engy auto policy remains worth defending when the miner paycheck is intentionally empty.

Source: Live Bitcoin News