Dates and verification labels belong to the original edition. A past verified announcement can now be outdated. Historical project names are retained; an SN filter searches the slot across registrations. All keywords must match the story text or subnet label. Ratings describe editorial priority, not investment quality.
Refinery adds protections for interrupted rounds and unavailable GPU capacity
Refinery’s new source provides urgent-update handling that preserves round audit state and returns credits for abandoned rounds. It also distinguishes GPU-capacity failures from evaluations that actually began, so unstarted work can be carried forward. This is a concrete follow-up to yesterday’s SN125 rebrand and mechanism reveal. Source availability and community “live” posts still do not prove that the complete deployed reward loop is operating.
Early operating rules that protect submission credits matter to prospective miners, especially during a new subnet’s rollout.
Emissions return across Bittensor: 19 subnet slots move back above zero
A comparison of yesterday’s saved network snapshot with 11:13 AM ET today finds 19 subnet slots moving from zero to a positive emission share: SN8, SN9, SN16, SN29, SN34, SN35, SN40, SN43, SN47, SN48, SN58, SN63, SN83, SN84, SN90, SN92, SN103, SN110, SN121. These include Vanta, iota, BitMind, Quantum Compute, Enigma, Green Compute and Capcomp. qBitTensor Labs independently confirms that SN48 and SN63 were re-enabled this morning, while asking for more clarity on the decision process. This is a subnet-allocation comparison, not proof that every miner payout is active or that one policy change caused all 19 transitions.
The breadth of the observed emission return makes this an ecosystem-level lead, affecting many projects and holders at once.
Buttensor reports 1,817 TAO distributed, with 206,373 cumulative payouts
Buttensor’s September 8 update reports 1,817.02 TAO distributed as of 18:08 UTC, an increase of 455.26 TAO since its September 7 update. The 206,373 figure counts payouts, not unique holders. The project describes a 3% transfer tax used for proportional TAO distributions after costs. These are project-reported dashboard totals, not an independent reconciliation, and the tax rate is not a yield. The new distribution milestone gives the StonkFun/TAO story a measurable follow-up.
A dated distribution total provides more useful context than price speculation, with the accounting limitations stated.
StonkFun reports $21 million in ecosystem rewards and adds more pairing options
StonkFun says cumulative rewards across its entire ecosystem have exceeded $21 million. That is platform-wide, not TAO-only revenue or rewards. Its September 8 announcements also add JTO and tokenized Sphere stock as pairing options and say StonkFun is available on Fomo. TAO pairing was introduced earlier and is not being presented as a new September 9 launch.
The broader platform’s reported scale and distribution channels matter to the new TAO use case, but unrelated rewards are not counted as TAO demand.
CONST’s community promoter reiterates the Bittensor focus
Cats says CONST was launched to bring the Bittensor community together and promote TAO and its subnets. A separate September 9 post warns about its volatility. These are community-positioning statements, not a new product integration, exchange listing or announcement by Bittensor co-founder Const. No new verified listing or utility launch was located in this scan.
The user-requested meme beat remains covered, while a community statement receives proportionate weight.
Forever Money’s Hash Rate appearance discusses TAO bridges and future subnet-token access
Hash Rate episode 185 features Forever Money’s philism discussing TAO on Base and Robinhood Chain, Chainlink bridge security, meme integrations and future subnet-token access. The September 9 episode is new; the underlying bridge announcements were covered earlier. Future subnet-token plans discussed in the interview are not a confirmed live bridge or launch date.
A new primary interview adds roadmap context to an important interoperability story without recycling the original launch.
Today with Shizzy: RL Bryer at noon Eastern, plus a new members briefing
The RL Bryer livestream is scheduled for September 9 at 12 PM ET / 9 AM PT, titled “2009 Bitcoin OG’s Next Big Bet: TAO | Ride of a Lifetime with RL Bryer.” The latest members-only briefing is the September 8 news and TAO price-action video, running 18:35. The public library’s latest regular upload remains “TAO Breakout Is NOW!” The stream was still upcoming when checked.
A practical viewing schedule and access update for readers following the channel.
Minos emissions are back — SN107 rallies roughly 13% against TAO
Minos has resumed receiving subnet emissions. The 10:49 AM ET network snapshot shows a 1.18% emission share, compared with zero at yesterday’s cutoff, and a 12.97% rolling 24-hour rise in α/TAO. Tensia’s alert recorded the restoration on September 7. Separately, Tensia co-founder Ror says the fix was approved and the miner burn should come off September 9; that next step is a reported plan, distinct from the emission restoration already visible.
A material emission pause has ended at a major genomics subnet, with a double-digit market response. The observed recovery earns top billing; the planned miner-payout change remains separate.
SN125 becomes Refinery: a new competition to improve how open models train
The former 8Ball slot now identifies as Refinery, with a published website and source code for incentivized optimizer research. Miners submit training algorithms; validators compare them on a roughly 360-million-parameter model using one B200 and a maximum 20-hour run, then require a separate confirmation. The initial research phase is planned for at least six weeks. This is a substantial change of mission. The repository still warns that source availability is not proof of an active deployment, and the current network snapshot shows zero subnet emissions.
A new research mission, concrete evaluation design and public implementation make this one of the day’s largest subnet developments. Live deployment and paid emissions are not inferred from the rebrand.
BitMind changes its validator burn to 80%, reserving 20% for rewards
BitMind’s latest merged code replaces the temporary 100% burn with an 80% baseline burn and bumps the package to v5.0.2. After that baseline, the documented maximum shares are 8% image, 8% video, 0.8% audio and 3.2% generators; empty lanes can burn additional weight. This supersedes the earlier 100% figure circulating in today’s briefs. The network snapshot still shows zero SN34 subnet emissions, so this code change does not establish that TAO emissions or miner payouts have resumed.
The reward allocation changes materially during the v5 rollout. Separating validator burn settings from network emission activation prevents a misleading payout claim.
OpenRoboto expands physical evaluation to seven tasks as SN80 climbs
New documentation gives both physical-robot tracks the same seven-task suite, with ten trials per task and 70 trials per submitted model. Tasks range from placing objects in containers to two-object rearrangement and plug insertion; the old single-block demonstration is now historical context. Public training-data guides are also linked. The two tracks retain separate results and their published season rules. SN80 was up 17.63% against TAO over 24 hours at 10:49 AM ET, with a 2.24% network emission share.
A broader real-world evaluation suite gives the robotics competition more substance, while the market and emission readings show significant current attention.
Cacheon opens its GLM-5.3 kernel-optimization arena
Cacheon announced on September 7 that miners can now submit kernel optimizations for GLM-5.3 and earn protocol rewards. The matching arena work has been merged into its public repository. The project sees the new model as an opportunity to turn optimization into production inference demand; the announcement does not disclose customer revenue or a signed commercial agreement.
A new competition built around a major open model creates a concrete new target for miners and a potential route to product usage.
Hippius launches encrypted Drive plans from free 10GB to $7 for 2TB
Hippius says its new Drive plans are live: 10GB free, 500GB for $4 per month, 2TB for $7, 10TB for $22 and 25TB for $50. The project describes the storage as end-to-end encrypted. This is a customer-facing pricing and product update, giving users a clear way to compare the SN75 service with conventional cloud storage.
Published consumer plans make the storage product easier to buy and use. A live commercial offering matters more than a general roadmap tease.
MicroTensor’s MT Code 3G appears on Verathos testnet
MicroTensor announced that its archived mt-code-3g-r1238-5ED3LzEW system is being served by Verathos. We checked the receiving project’s Models page: selecting Testnet shows MT Code 3G, labeled SN92, with a 32k context and one miner at the time of inspection. The listing was absent from the mainnet table. This is a concrete example of a model produced by one subnet being distributed through another, at the testnet stage.
The announcement is supported by the receiving product’s visible listing. Cross-subnet model distribution is meaningful progress, with the testnet boundary explicit.
RedTeam’s five enterprise pilots and upcoming Immune System get a Foundation showcase
OpenTensor Foundation’s next Novelty Search will feature RedTeam’s bot-attack defenses, five new enterprise pilots and its upcoming Immune System. The broadcast is scheduled for Thursday, September 10 at 9 PM UTC / 5 PM ET on X and YouTube, hosted by Const. RedTeam also shared the event. Pilots are not the same as disclosed paying contracts, and the Immune System is still described as upcoming.
Specific enterprise pilot activity and a new defensive product deserve prominent coverage, supported by both the project and the Foundation’s event announcement.
Ditto brings memory and tools into an inference endpoint powered by Bittensor services
Ditto and founder Peyton Spencer demonstrated Ditto Inference, combining memory, observability, session backups, built-in tools and compression in the inference endpoint. Ditto names SayGM, engy and Chutes as its underlying inference providers. Its separate on-premise offering runs on customer-controlled hardware; the team says it is qualifying its first lead. The provider integrations are project-reported, and a lead is not a completed customer contract.
An application combining several subnet services offers a tangible route from network capacity to a product people use. The enterprise sales stage remains clear.
Desearch 2.0 proposal points toward a continuously refreshed web index
Today’s team-channel brief reports a proposal to move Desearch from live query routing toward preparing public web content in advance. The first phase would cover crawling, content preparation and embeddings, with GPU capacity required from miners and validators. The brief reports a September target, but hardware requirements and participation details are still pending. A direct published proposal or release was not located, so this remains a developing direction rather than a confirmed launch.
A proposed change to the subnet’s core search architecture could substantially alter its product and miner workload, warranting visibility before the implementation is confirmed.
Instant founder refocuses SN46 on one-model inference competitions
In a September 7 essay, Seby says SN46 is again the center of his work after expanding beyond the original RESI team’s capacity. Instant’s proposed starting point is accelerated inference with one model per mechanism, explicit economic requirements and performance targets that can incur slashing. The planned sequence is a focused mechanism and API, then a front end. He says additional models and mechanisms will depend on research and comparable performance evidence.
A founder’s explicit strategic reset and concrete incentive direction are material for the subnet. The essay sets a plan; it does not announce that the new API is already live.
Cortex opens Bot Desktop beta access for miners
Cortex announced beta access to Bot Desktop for miners on September 8 and published a demonstration. This is a new product-access announcement following yesterday’s unresolved readiness report. Separately, the CLI repository adds persistent goals and continuing runtime improvements. The beta announcement is confirmed; it does not establish that every planned production feature has shipped.
A visible beta invitation is a meaningful change from a planned release and belongs with today’s product launches.
Capcomp falls roughly 58% in 24 hours; no confirmed cause established
SN103’s α/TAO price was down 57.90% over 24 hours at 10:49 AM ET. The same snapshot showed zero subnet emissions and roughly 647 TAO on the pool’s TAO side. This is a major market move worth flagging. Community allegations in the channel were not corroborated by a primary announcement or transaction investigation, and the price decline alone does not establish wrongdoing.
An unusually large move and shallow remaining pool liquidity make this a material market update. The rating measures news importance, not an accusation or a trading recommendation.
Searches published headlines and summaries, not private bookmarks, unpublished leads or source-page contents. Duplicate headlines within an edition appear once; separate editions remain separate so follow-ups are preserved. Share the page URL to retain your filters.
