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 A past verified announcement can now be outdated.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 adds a check for whether bootstrap weights actually take effect
UMI’s status now separates an authorized bootstrap service row from its observed economic effect. It remains ineffective if another active validator uses a different row or any frozen eligible miner receives zero incentive. The current README says SN78 is active on mainnet but translation weights are not active. That helps explain why a bootstrap status label must not be read as proof of normal translation rewards.
The new status distinction addresses a concrete source of confusion during subnet activation and emissions changes.
Reliquary reduces proof-processing round trips
Reliquary changed proof verification to send one batch for a 16-rollout training group instead of splitting it into four requests. Its release notes identify network round trips as a major part of the measured delay. The repo also adds recovery and timing improvements around training groups. This is a focused processing update after the already-covered V1 work, not a newly launched public consumer product.
Reducing an identified bottleneck matters to whether training windows fill and useful work reaches verification.
Vidaio and PIP bring Cabin to IBC in Amsterdam
Vidaio says it is at IBC with PIP CEO Mark Sheffield and his team, presenting Cabin’s AI-powered localization. The September 10 arrival update follows the previously announced September 11–14 showcase. This is the start of the public demonstration and industry meetings, not evidence of new contracts signed at the event.
A subnet-powered product reaches a major broadcast-industry audience through its existing studio relationship.
MicroTensor brief reports rewards returning; paid compute is still in testing
A September 11 channel digest reports emissions re-enabled for participants and 42 rewarded systems in round 1240, using p95 task latency in scoring. The same digest says the paid-compute layer remains in testing and its temporary reserve will be reduced when that layer goes live. The market feed shows a positive subnet allocation, but the claimed participant payouts were not independently reconstructed.
The distinction between a reward restart and a paying-workload launch is material for both operators and readers.
Nodexo reports 123 GPUs listed on its public board
Nodexo says 123 GPUs are listed by independent operators, with an estimated hardware value of about $1.3 million. Its update describes a timed compute challenge used to check that listed hardware matches the claimed GPU class, with validators checking the proof on ordinary CPUs. These figures describe available inventory and estimated equipment value. They do not establish rental revenue, utilization or that Nodexo owns the hardware.
A quantified supply milestone shows the scale of the marketplace while keeping listed capacity separate from actual customer spending.
Ditto details chat-history imports and portable memory for V2
Ditto’s new walkthrough describes importing exported chats from ChatGPT, Claude, Gemini and Perplexity, plus notes from Obsidian, Joplin, Google Keep and Markdown files. It also describes exporting a signed Passport ZIP or plain Memory JSON. The announcement still says the V2 beta will open soon, so this is a feature preview rather than a claim that everyone already has access. Separately, the subnet dashboard gained coding shadow-coverage controls; those do not establish that a new scored coding competition has launched.
Moving existing history into and out of an assistant is a useful product feature, with the announced access stage retained.
Affine adds a scored terminal-agent command format
Affine’s September 10 update advances its weight-version key to 14 and adds the Terminus JSON command-batch format as a scored action. The operator notes say the core scoring math, incumbent reign and minimum submission block remain unchanged. The update broadens the action formats a terminal agent can use without treating it as a new competition reset.
A new accepted action format changes what participants can submit and how tooling connects to the competition.
Sundaebar brief reports an eight-hour duplicate-submission cooldown
The latest channel digest reports an eight-hour cooldown for near-identical submissions and a rule requiring a lead greater than 0.01 points to change the leaderboard. It also lists September 12 as SBC27’s closing date. These changes respond to queue saturation and close-result disputes; the original team messages and live enforcement were not independently checked.
Submission and tie rules directly affect competitors, with secondary-source limitations retained.
TAO pairs go live on BaseStonk through ForeverMoney
BaseStonk and ForeverMoney both confirmed that TAO pairs are live on the Base launchpad. ForeverMoney says creators can launch tokens against TAO and earn creator fees in TAO. This adds another use and distribution channel for TAO beyond the Bankr integration covered yesterday. Whether it creates sustained buying pressure depends on actual launches, liquidity and trading activity; those totals were not disclosed in the announcement.
Confirmation from both sides establishes a new integration outside the subnet ecosystem and another concrete use for TAO.
Emissions update: ORO and UMI show zero; Witness gains a small allocation
At 10:07 AM ET, ORO’s observed allocation is zero, versus 4.465% in yesterday’s saved snapshot. UMI is also zero, versus 0.107%, while Witness now shows a small positive allocation of about 0.000047%. Hotfloat, Epago, Leadpoet, Feval and CookingTAO remain at zero in this comparison. These readings describe allocation, not the reason for it. Lamida’s September 11 post says only SN29 and SN36 were manually paused; that administrative-state claim was not independently verified today. A validator burning miner rewards, a zero subnet allocation and an administrative disable are different events.
The on/off story is still changing, and the new zero readings affect substantial subnets. Timestamped comparisons prevent a community claim from being mistaken for a confirmed switch.
StonkFun reports $30M in holder rewards and 14.4% of STONK burned
StonkFun says cumulative rewards distributed to ecosystem holders have exceeded $30 million. A separate update reports 14.4% of STONK’s supply burned, worth more than $40 million at the then-current token price. The latter is a valuation of burned tokens, not $40 million in cash spent on buybacks. Both figures are platform-wide and company-reported, rather than TAO-pair-only results. The launchpad’s TAO pairs remain part of the broader story already covered in prior editions.
Fresh business and supply figures matter for the user-followed launchpad, provided their scope and valuation basis are clear.
Latest Shizzy video: TAO Broke Down… What Happens Next?
The September 10 public episode is now the site’s default video: “TAO Broke Down… What Happens Next? | Bittensor Subnet Market,” running 19:48. The RL Bryer conversation remains available as a replay. The latest members upload found in the channel check is still the September 8 briefing, accessed through YouTube membership.
The video library now opens the latest verified public episode and keeps member access clearly labeled.
Lium takes the subnet market-cap lead: $964K billed, plus TAO borrowing on Mentat
Lium reports $964,000 billed to 1,187 renters in its September 9 monthly review, with 36% month-over-month growth, 1,908 new signups and about 80% user retention. It says 63% of rentals were initiated programmatically. Capacity now spans 68 datacenters in 21 countries. Today, Mentat added live TAO borrowing against Lium collateral; its application lists the SN51 market. At 11:07 AM ET, the market snapshot puts Lium ahead of Chutes by subnet market capitalization: approximately 422,226 TAO versus 405,641 TAO. Lium’s token gained 8.65% against TAO over 24 hours. Billing and retention are company-reported; market rank is a timestamped snapshot.
Substantial reported customer spending, a live lending integration and the market-cap lead combine into the day’s strongest commercial story. Price supports its prominence.
Lium patches rental isolation and request replay; SDK fixes runaway billing risk
Lium merged changes that isolate rental containers from neighboring containers and reject signed metrics/log requests outside a five-minute freshness window. These address documented weaknesses, with no confirmed exploitation established in this scan. Its SDK also schedules pod termination as soon as a rental is created, rather than waiting for readiness; previously, a pod that failed to start could remain billable without an end time. The code changes are public, but fleet-wide rollout was not independently checked.
The fixes address tenant separation, replay protection and a concrete billing failure. They merit their own operational item beneath the adoption lead.
Epago up over 1,100%, with TAO emissions switched off
SN36 is now Epago, an open deep-research project with published code and a model competition built around Tongyi-DeepResearch. Its token was up 1144.71% against TAO over 24 hours at 11:07 AM ET. The team reported a 1,050-TAO purchase and perpetual lock. But the subsequent transaction 9032187-0016, executed September 9 at 19:57:36 UTC, sets subnet 36’s emission-enabled flag to false through a sudo call; the explorer shows a successful execution. The latest snapshot still records zero subnet emissions. This is a confirmed administrative switch, not just a falling allocation. Epago’s public agent API is not open yet and its site labels the demo as scripted. We did not establish the decision-maker’s identity or reason for the shutoff.
A new subnet identity, extreme market move and confirmed emissions intervention deserve top billing together. Five stars measure news importance, not an endorsement.
Conjectures publishes a proof for the near-half-density case of Green’s Problem 51
Conjectures says its miners solved a case that had been open for 16 years: the near-half-density question in Green’s Problem 51. Its September 9 working paper, by Jensen Kohlmeyer and Liam Kruer, describes the theorem and provides Lean verification and reproduction details. The result concerns guaranteed subspaces in sums of dense sets over a binary vector space. It does not solve Green’s full extremal problem or the related question over the integers. We checked the published paper and its stated scope; we did not independently rerun the formal verifier.
A substantive research result with a public proof and verification record is a major output from a subnet. The precise case and remaining limits matter.
TensorUSD announces a long/short strategy pivot with ViewPallet; TUSDT stays
TensorUSD plans to turn SN113 into a competition for profitable and delta-neutral strategies across subnet-alpha markets, while retaining the TUSDT stablecoin and existing AI liquidator. Its September 10 announcement describes a collaboration with ViewPallet through which users could allocate TAO to miner strategies and copy positions. A performance fee on realized profits would fund protocol operations. This is an announced direction and proposed revenue mechanism: the post calls the derivatives system upcoming, and we did not find a matching new launch announcement from ViewPallet.
A change in the subnet’s core product and business model is material, even though the trading mechanism remains planned.
Bitcast adds TestDrive as its third creator application
TestDrive joined the Bitcast ecosystem on September 9, bringing indie hackers, app builders and solopreneurs into its creator network. Participants try products, publish their own assessments and earn through the attention that is verified. Bitcast explicitly identifies this as creator application number three, and TestDrive’s account shares the announcement. The integration was announced just after yesterday’s research cutoff; it is a new addition to this edition.
A live application expands the creators and customers using the network. Both project accounts link the integration.
Bitsec names Talisman as a new continuous-auditing client
Bitsec says Sentios audited Talisman’s updated wallet extension and wallet app, and welcomes Talisman as its newest continuous-auditing client. The announcement quotes Talisman’s extension v3.8.1 release, which adds transaction-risk analysis, clearer decoding and root-validator basket visibility. Talisman’s release confirms the product update; the audit engagement is reported by Bitsec and was not independently reconciled against a public audit report.
A named wallet customer is concrete adoption evidence for the security subnet. The commercial claim is attributed to its publisher.
Ditto confirms 1,342+ app integrations are live
Ditto’s September 9 post explicitly says its app integrations are live, including Google Workspace, Notion, Linear and HackMD. It supports multiple accounts and describes a shared memory layer that helps agents coordinate across tools. This supplies the first-party activation statement missing from yesterday’s scan. It confirms this product feature, without implying every element of Ditto V2 or the next mining benchmark has launched.
A large, usable integration expansion is a material product release, with an explicit first-party live statement.
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.
