THE DAILY BRIEFING

Bittensor Daily — October 17, 2026

A new Subtensor change calculates the root network's share of block emissions live from current network state each block, replacing a fixed per-subnet value that can shift how emissions are effectively split among validators. Separately, SN25 (UR) launched mainnet, bringing more than 100,000 existing network providers onto Bittensor as miners, while SN2 (DSperse) underwent a leadership handover to a new development team.

01
Subtensor PR 3264 changes the root network's emission share to be calculated live each block from current network state, instead of using a per-subnet TAO weight value that was frozen when set.
This is a breaking emission-related change that can shift effective payouts for validators under the new live formula.
02
SN25 (UR) launched mainnet, transitioning more than 100,000 existing network providers from UR Network into Bittensor miners, with miner burn set to 0% and a new Network Capacity Reserve funded by excess emissions.
The launch brings an existing large user base onto Bittensor, with BringYour named as the first paying customer once revenue begins.
03
Inference Labs transferred leadership of SN2 (DSperse) to a new team, DaniLabs, whose lead previously worked on DSperse before joining OTF and has now returned to the subnet.
The handover marks a continuity transition for a subnet that has reported over 6 billion proofs of inference and peak throughput above 1,000 proofs per second.
04
On SN92 (REAX), a chain-commitment limit caused competition entries, including UID 111, to be deregistered during evaluation before the team introduced a keep-alive weight and retroactively restored affected entries.
The gap meant hotkeys submitting competition entries could lose their slot instead of being properly evaluated, prompting the team to apologize and fix it retroactively.
05
Community members raised concerns over recently introduced Null Consensus and Gamma token mechanism changes, citing worries about validator power concentration and potential subnet-owner abuse.
Some are requesting a public town hall with the responsible team to clarify the design, reflecting unresolved concerns about centralization.

SN2 24h +29.41%

DSperse

What happened: Inference Labs transferred leadership of SN2 (DSperse) to a new team, DaniLabs, which will take over subnet development with a focus on fast, affordable AI inference. The new lead previously worked on DSperse before joining OTF and has now returned to the subnet. A team member separately recapped SN2's history, citing over 6 billion proofs of inference and peak throughput of over 1,000 proofs per second in its zkML work.
Community: The leadership change prompted disputes in the subnet channel, with some community members raising concerns about a legal dispute and questioning the new lead's fitness to run the subnet, while others voiced support. One member suggested that a stake could settle doubts if the claims were false; a moderator said the broader team would no longer intervene in subnet-specific disputes.
Official: Inference Labs confirmed the handover on X, thanking the community and noting the new lead joined the team from OTF over a year ago and helped scale DSperse. The subnet channel also published a recap citing more than 6 billion cryptographic proofs completed and peak throughput above 1,000 proofs per second.

SN92 24h −3.26%

REAX

What happened: The team opened two SN92 (REAX) competition tracks, demand forecasting and risk assessment, at 12:00 UTC on October 30, while delaying the REAX Engine track to Monday for an additional improvement pass. A chain-commitment limit meant hotkeys submitting competition entries could not also serve inference and were being deregistered during evaluation; the team introduced a keep-alive weight for entries that pass copy checks and beat the entry floor to prevent this. Some older entries scored before the rule change, including UID 111, were still pruned and had to be retroactively covered. The team also fixed a validator speed-measurement issue so hotkeys are credited at real throughput, removing the payout advantage of splitting one machine across multiple hotkeys, and set a 0.5 TAO minimum registration fee alongside the existing increase and cap.
Community: Miners questioned why UID105's incentive dropped despite high performance and raised concerns about hotkey deregistration during competition evaluation and the fairness of multi-hotkey versus multi-machine setups. After UID 111 was deregistered due to the keep-alive gap, a community member flagged it along with UID26 and UID27 losing emission.
Official: The team acknowledged the keep-alive gap affecting UID 111 and apologized, fixing it retroactively since 19:21 UTC, and confirmed step one of the speed-measurement fix went live at the 21:15 UTC weight set, with multi-machine support for a single hotkey planned next.

In debate · Fairness of multiple hotkeys vs. multi-machine mining on SN92 · 29 participants

Some miners argue splitting GPUs across separate hotkeys is more profitable than consolidating them under one hotkey, distorting registration incentives.

“having 2 hotkeys with 1 gpu each is still more profitable than having 2 gpus with 1 hotkey”
The team maintains running several hotkeys per machine is allowed and not cheating, but should not pay more than a single hotkey once the speed measurement is fixed.

“it's within the rules, so it isn't cheating. But it shouldn't pay more than one hotkey”
Miners with multiple machines say a single hotkey does not receive enough traffic, leaving hardware idle until multi-machine support ships.

“even in one machine, one hotkey isn't enough as it's not getting enough traffic.”

SN25 24h +8.23%

UR

What happened: SN25 (UR) launched mainnet, transitioning more than 100,000 existing network providers from UR Network into Bittensor miners. The subnet's validator is active, setting weights and verifying traffic, and mining and validating are permissionless. The team set miner burn to 0% and launched a Network Capacity Reserve, with excess emissions funding future miner capacity. A mainnet upgrade to fix miner payouts is planned for the following week.
Community: Community members congratulated the team on the launch and asked detailed questions about how Alpha ties into network usage, including whether operators will need to continuously acquire Alpha and whether usage fees paid in fiat will be used to buy SN25. Others asked practical questions about how mining works and whether mining is open.
Official: The team said BringYour, which operates ur.io and the associated VPN, will be the first paying customer, and that the subnet is not yet generating revenue since mainnet just went live. It confirmed usage fees will be paid in SN25 Alpha, with fiat payments converted into SN25 purchases, and that staked Alpha may bring usage discounts.

SN44 24h −0.98%

Score

What happened: On SN44 (Score), the team removed several petrol-detection ground-truth blocks (9292500, 9281100, 9292200) after miners reported incorrect bounding box annotations. A miner alleged that a top performer on the Cricket element controlled roughly 30 hotkeys and continued operating new hotkeys after being blacklisted. The team paused Cricket challenges while adding new ones and clarified that the tie-break mechanism compares outputs across 20 challenges, 10 latest and 10 random, reproducible via hash. The PPE and Fallen Person Detection tracks went inactive and were later restored after repeated miner reports.
Community: Miners repeatedly flagged ground-truth errors in the petrol track and asked for specific blocks to be reviewed and removed. Several participants raised concerns about multi-hotkey control on Cricket and pushed back on the tie-break threshold, arguing it favors established miners, while others defended the existing rule as effective.
Official: The team confirmed removal of the flagged petrol blocks, explained the tie-break's random-but-reproducible design, paused Cricket pending new challenges, and announced the Fallen Person Detection element was restored and would be monitored.

In debate · SN44 tie-break threshold and fairness of the scoring system · 11 participants

A miner proposed lowering the tie-break score threshold from 0.05 to 0.04 to reduce unfair advantages for established models.

“How about lowering the score threshold from 0.05 to 0.04?”
The team maintained that the current threshold effectively protects miners' models from being copied and reflects genuine output improvements.

“Tiebreak effectively protects miners models from being copied, isn’t exploitable and effectively value actual OUTPUTS improvements while ignoring noises”
Another community member argued that overly easy recent challenges, not the threshold itself, were producing unfair tie-break outcomes in the beverage track.

“Recent challenges are very easy and it makes unfair tie-break in beverage track.”

SN107 24h −1.30%

Minos

What happened: Discussion in the SN107 (Minos) channel centered on questions about why certain accounts maintained top mining ranks over extended periods, with some participants noting that multiple UIDs appeared linked to the same group. The owner team addressed the concerns directly, stating it does not control top-ranked miners and pointing to hashing of inputs and truth files as a way for participants to verify submissions were not tampered with. The team also said it is prioritizing release of IM2 for fully agentic mining and plans presentations on the 22nd and 23rd.
Community: Several miners pressed the team for ways to independently verify that top miners' configurations matched what was committed on-chain, with one participant repeatedly questioning whether top ranks were effectively fixed. Other community members pushed back, saying they had looked into the matter themselves and did not find evidence supporting the concerns, while one noted that over 250 hotkeys were receiving emissions.
Official: The team said it is not controlling the top-ranked miners, described ongoing work to add resilience against participants using multiple coldkeys to occupy top slots, and announced it will share further details on X ahead of presentations on the 22nd and 23rd.

More subnet community coverage every day in the Telegram channel →

Null Consensus and Gamma token changes spark centralization concerns

Community members raised confusion and concern over recently introduced Null Consensus and Gamma token mechanism changes. Some asked for a public town hall with the team to clarify the design, citing worries about validator power concentration and potential subnet-owner abuse.

Positions

Some community members want a public town hall to address confusion and fears of validator centralization.

“there is a lot of confusion + worries about centralisation + single Validators getting too much power”
A community member questioned the purpose of Null Consensus, suggesting it could let subnet owners extract value more easily.

“it makes it easy for subnet owners to abuse and extract money no?”
Another community member argued Null Consensus changes little in subnets where one validator already dominates, since it only formalizes an existing reality.

“Null just makes the fact explicit.”

Official response: No official response addressing the Null Consensus or Gamma token design concerns has appeared in the messages reviewed; one community member suggested asking questions in the channel, and another directed subnet-specific or chain-wide weight-setting issues to the appropriate validator or network channels, without addressing the centralization concerns directly.
What to watch: Whether the team holds a public town hall or issues a direct explanation of Null Consensus and Gamma token mechanics in response to community requests.

@inference_labs

Inference Labs said SN2 (Inference Labs) has produced more than 6 billion cryptographic proofs and announced that a team member will take over the subnet under the handle @danilabsai.

View the original post on X ↗
@JesusMartinez

A community member noted that subnet owners can now opt in to raising the miner cap on their subnet from 256 to 2,500.

View the original post on X ↗
@urnetwork

UR said it is live on mainnet on SN25, transitioning its 100k+ network providers to Bittensor miners with a permissionless validator and miner setup.

View the original post on X ↗
@urnetwork

UR said it has set miner burn to 0% on SN25 and is launching a Network Capacity Reserve funded by excess emissions, held in a separate multisig.

View the original post on X ↗
@nepher_robotics

Nepher Robotics reported Humanoid Run Jump Phase 3 ended with a winning score of 0.8245, up from 0.6750 in Phase 1 and 0.7735 in Phase 2, with 0.0018 separating first and second place.

View the original post on X ↗
@TaoOutsider

A community member cited $583,268 in platform revenue over the last 30 days for Targon Compute, the business behind SN4 (Targon).

View the original post on X ↗
@EndureNet

Endure said its first week on mainnet covered seven rounds from Oct 4 to 10 with an average score of 95.5%, scoring all 420 values against actual outcomes.

View the original post on X ↗
@CryptoZPunisher

A community member said SN28 (SayGM) offset 12.9% of miner emissions in Cycle 1 and 20.3% in Cycle 2, totaling $147,460 in alpha burned, including $104,372 in the latest cycle.

View the original post on X ↗

◆2026-10-17 · End of this edition

#3264 Breaking change Affects emission

etao: pay root its live baseline share off the top every block (spec 481)

What: Subtensor now calculates the root network's share of emissions live, each block, from current network state, instead of using a per-subnet TAO weight value that was frozen at the time it was set.
Why: A frozen, point-in-time weight could drift out of step with actual network conditions over time, so the change ties root's emission share to live data instead.
Who it affects: Validators see root's emission cut computed and deducted from the top of each block's emissions under the new live formula rather than the old fixed value, which can shift effective payouts; ordinary TAO holders and miners are affected only indirectly through overall emission flow, with no action required on their part.

Recent upgrades

Oct 13 v478 · Subtensor v478 lowers EVM gas needs for unstaking and restores missing staking associations

24H Market

Gainers

Losers

Bittensor Daily — October 17, 2026

DSperse

REAX

UR

Score

Minos

Null Consensus and Gamma token changes spark centralization concerns

etao: pay root its live baseline share off the top every block (spec 481)

24H Market

↗Gainers

↘Losers

Bittensor Daily — October 17, 2026

Today's highlights

Subnet Community Pulse

DSperse

REAX

UR

Score

Minos

Ecosystem debate

Null Consensus and Gamma token changes spark centralization concerns

Voices on X

Protocol

etao: pay root its live baseline share off the top every block (spec 481)

Recent editions

Gainers

Losers