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 A past verified announcement can now be outdated.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.
UMI documents why zero TAO inflow does not mean emissions were switched off
UMI’s new operator note records finalized block 9,045,044 with emissions enabled, zero TAO and alpha pool inflow, approximately 99.61% miner burn, and participant-side issuance of one alpha per block. Those fields describe different parts of the system. The current market feed shows SN78 back at 0.110% of observed allocation at 8:08 AM ET, compared with zero yesterday. Neither a positive allocation nor a submitted weight row proves every miner is being paid. The team’s public documentation is a concrete counterexample to reading a zero dashboard value as an administrative shutoff.
This directly clarifies the emissions on/off story using an identified block and separate accounting fields.
404-GEN opens a two-week Game Jam with a 10 TAO prize pool
404-GEN’s first Game Jam is open from September 11–25. Participants build a playable Three.js game using an AI agent and the project’s generator repository, with 10 TAO in prizes and winners announced at Exploit Summit. Opentensor also shared the competition. It is an open creative challenge that puts the subnet’s game-building tools into users’ hands, rather than a claim that a commercial game has already launched.
A live, funded competition is a concrete opportunity to build with subnet output and demonstrate it publicly.
Vanta channel brief reports more than 4,000 traders since April
The September 12 channel digest attributes a milestone of more than 4,000 traders since April to the Vanta team. That figure was not independently found in a new original public Vanta post, so the milestone remains developing. Vanta’s verified September 11 rules post reiterates that it has no consistency rule or single-day profit cap. The trader count should not be read as 4,000 funded accounts, profitable traders or audited paying customers.
A substantial adoption milestone deserves visibility, with the exact source limitation and meaning of the count retained.
ORO Race 2 reportedly moves to September 13 as full burn continues
The latest channel brief says Race 2 has been postponed to September 13 at 19:00 UTC / 3 PM ET while the team changes the incentive mechanism to discourage hardcoded agents. It reports 100% burn until the next race. The public repository separately upgrades the environment runtime to remove the grader rulebook from the agent-facing package. The original Discord messages and live reward vector were not independently accessed, so the rescheduled race and burn status remain developing.
A race delay and continued full burn materially affect miners and the emissions story. The verified runtime change supports the development activity without proving the live schedule.
Reliquary brief reports recovery of its subnet ownership key
The latest channel digest says a team account confirmed that Reliquary’s ownership key has been reclaimed following Covenant’s departure from Bittensor. The original message and a corresponding completed chain transfer were not independently verified in this scan. The claim is therefore developing, with no purchase price or buyer identity inferred.
A reported recovery of subnet control is important enough to surface prominently, while the missing primary transfer evidence remains explicit.
Reliquary adds overlapping proof transfers to reduce idle GPU time
Reliquary’s new implementation lets proof uploads overlap GPU processing through bounded queues and optional dispatch lanes. Its release notes measured roughly half of a sample remote proof call being spent on data transfer. The controller still defaults to one lane, so the merged capability does not mean every operator has enabled the higher-throughput setup. A separate fix keeps reveal accounting distinct from admission failures.
The changes address a measured bottleneck in useful proof work and improve accounting, with the default activation setting stated.
Apex brief puts Humanoid Box Scramble’s launch on Monday
The September 11 channel digest attributes a Monday launch plan for Humanoid Box Scramble to the Apex team. This adds a reported launch target to the competition-rule changes covered on September 9. Direct confirmation of the schedule was not obtained, so the date remains developing rather than a claim that the competition is already open.
A near-term competition opening is actionable news for builders, with its secondary-source status visible.
Teutonic brief reports a temporary 1 TAO registration cost
The September 11 digest reports that the team raised registration cost to 1 TAO to reduce evaluation-queue congestion. It describes the measure as temporary and subject to monitoring. The exact current on-chain fee was not independently queried, so miners should check the live registration quote rather than assume this amount remains fixed.
A reported participation-cost change matters immediately to prospective miners.
Numinous brief teases product news and future buybacks
The September 12 digest attributes forthcoming product news and plans to buy back tokens using team-run forecasting agents to Numinous. No new product release or completed buyback transaction was independently established by the cutoff. This remains a watch item for a follow-up announcement.
The team’s stated next step is relevant, but a teaser alone does not warrant a major-news rating.
Perturb brief reports 100 TAO of buybacks and plans further purchases
The September 12 digest attributes 100 TAO of completed alpha buybacks to the Perturb team and says another 200–300 TAO of purchases are planned, followed by permanently locking the acquired alpha. The transactions and lock state were not independently reconstructed. Planned purchases and locks must not be counted as already completed.
A reported treasury action and intended permanent lock matter to holders, with completed claims separated from future plans.
SayGM brief reports a routing-capacity fix, with two more changes on testnet
The September 11 channel digest reports that a bug resetting learned routing capacity after price changes has been fixed. It attributes continued traffic concentration partly to session pinning and says two additional routing improvements are on testnet. The live fix and remaining tests were not independently reproduced; miner estimates of traffic concentration are not presented as an audited network-wide result.
Routing affects which providers receive work, and the fixed-versus-testing distinction matters after the recent model launch.
BitMind brief reports recovery from stale validator weights
The latest channel digest says a validator missed three epochs and others used cached video-champion data after an API failure. It attributes a deployed fix to a team member and notes that commit-reveal can delay when updated weights appear on explorers. The claimed recovery was not independently replayed, and no compensation for missed rewards was established.
A reported issue affecting weight updates and its recovery are practical operational news.
ChronoLLM presents its time-bounded AI approach at Cornell Tech
ChronoLLM says Jean Herelle presented at Cornell Tech’s AI & Future of Finance conference on September 11, focusing on backtests that use only information available at the historical decision time. A participant’s post also describes seeing the project there. This is an industry presentation, not a new benchmark win or customer contract.
The presentation brings the subnet’s research approach to a relevant professional audience.
Score puts September 16 on its annotation-tool preview
Score’s new product preview shows a draw-once, apply-to-similar annotation workflow and gives September 16 as the date to watch. This adds a concrete date to the Studio buildup covered yesterday. The post does not establish that the tool is already publicly available. No primary confirmation of the rumored Shell partnership was found in this scan.
A dated product preview is useful near-term news, while a launch and an unrelated partnership remain separate claims.
Lium stops an outdated executor image from automatically producing a zero score
Lium’s validator patch makes an outdated executor image a warning by default. It changes scores only when the explicit image-enforcement setting is enabled, which defaults to off. This addresses an operational cause of zero scores; it does not establish a general payout restart or a new buyback announcement.
A targeted scoring fix can directly affect participating compute providers.
Ridges reportedly pauses new uploads while clearing its evaluation queue
The latest channel brief says the competition entered a draining state, temporarily pausing new submissions while pending evaluations clear. Reopening was expected within hours of the message. The public site returned a browser security checkpoint to the direct check, and the current reopening status was not established; miners should confirm the live submission page before preparing an upload.
An intake pause affects participation immediately, but an expected reopening time is not proof that uploads have resumed.
Conjectures brief reports bounty-policy deployment and four approved submissions
Following yesterday’s merged 15-day bounty policy, the latest digest reports that the new formula has been deployed without retroactive changes to queued submissions. It also reports approval of submissions for Erdős 96, Green 15, Erdős 196 and Erdős 272, with proofs to be published. The original approvals and complete proofs were not independently examined, so these are reported review outcomes rather than independently validated mathematical breakthroughs.
Deployment status and reported research approvals are meaningful follow-ups, with proof verification still outstanding.
NOVA briefs give a Boltz-2 start epoch and flag a separate Blueprint upload issue
The September 11 channel brief schedules the initial Boltz-2 competition for epoch 20,708, using raw scores, no antitargets and a one-hour budget. A September 12 brief separately says Blueprint uploads were sent to an old archive and the team planned to move them before restarting the run. These are distinct operational updates; neither digest independently confirms that the relevant run has finished or rewards have settled.
The start conditions and upload recovery matter to participants after yesterday’s Boltz-2 testing announcement.
MVTRX brief previews a Tuesday update and new research papers
The latest channel brief reports a Tuesday target for a landing page, roadmap, feature requests, anonymous access and a scoring revision. It also says the team is preparing papers on applying Navier–Stokes methods to financial simulation. The papers and product changes were not yet independently verified as released.
A specific near-term roadmap is worth watching, with a lower rating until the product and research are public.
OpenRoboto brief reports a CLI status fix; submitted models were reportedly unaffected
The latest digest says a backend parameter rename caused a false validation error in the CLI status command. It attributes a fix to version 1.3.0 and says submitted models remained queued normally. The release and queue state were not independently reproduced, so the update remains developing rather than a verified service outage.
The distinction between an erroneous status message and a failed submission is directly useful to miners.
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.
