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 A past verified announcement can now be outdated.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.
Compelle refreshes debate topics every six hours; API and sealed strategies are planned
Compelle says its new topic system is live, drawing from prediction markets, current discussions and longer-running debates every six hours. It targets sealed miner strategies for October and an MCP/API integration by year-end, with private beta testing already underway. Lower-cost models and deeper reasoning settings remain under evaluation.
A live content-system change and clearly dated product plans make this a useful roadmap update.
SN84 owner coldkey swap is scheduled, not completed
SubnetRadar reports a scheduled coldkey swap for the account owning SN84, from 5G1hDN…tbbd to 0xf072…6cf1. Its chain-event page records block 9,030,810 at September 9, 15:25 UTC and labels the status Scheduled; its X alert describes a mandatory delay of roughly five days. This is a custody/control change to watch. It does not establish a completed subnet sale, buyer identity or purchase price.
A pending owner-key change is relevant to holders, but must be distinguished from a completed transfer or sale.
NeverPlayAlone retries failed artifact uploads instead of aborting the round
A new NeverPlayAlone change retries transient artifact uploads and skips individual entries that still fail, allowing the rest of a round to proceed. The repository confirms the handling change; no measured network-wide availability improvement was published with it.
A practical reliability fix for competitors, with limited evidence of broader impact.
Cortex adds research submission guidance and fixes sealed burn-weight submission
Cortex published a tbench research guide and fixed validators skipping a correctly sealed UID0 burn outcome. Unsealed results remain ineligible for submission. Its September 10 website table says Proof and Bounty submissions are accepted and judged, but other text on the same page still describes partial judging, disconnected Bounty scoring and automatic payouts not being wired. The table also says weights are not yet sealed. With these conflicting status descriptions, a complete paid-marketplace launch is not established.
The miner guide and weight-submission fix are real changes, while primary status disclosures prevent overstating a launch.
Beam marks six months and shifts its focus toward adoption
Beam’s six-month update summarizes its transport network, public transfer verification, large cross-cloud transfers and developer tools. The team says its next phase is increasing use in applications, enterprise workflows and AI infrastructure. It does not announce a new enterprise contract or establish that future petabyte/exabyte targets have been reached.
A current project milestone and strategic update is useful context, with a lower rating because much of it recaps existing work.
Claims recovers failed adjudication batches without discarding successful cases
Claims’ validator change splits malformed adjudication batches into smaller groups. An irrecoverable single case is excluded with a recovery explanation, while successfully adjudicated cases can continue. This reduces the impact of a model-output failure without accepting an incomplete result as valid.
A focused reliability fix improves the processing pipeline but has narrower impact than a product launch.
CookingTAO targets beta v2 within a week
CookingTAO’s September 9 announcement invites developers to submit miner code through its Mining Management Portal and targets beta v2 within a week. The team says subnet parameters were modified for its incentive mechanism, with more details to follow. The invitation and timing are confirmed as a plan; this announcement does not establish that normal miner payouts or the full platform are already live.
A near-term beta and miner recruitment are worth tracking, without turning a target into a completed launch.
TAO becomes a Bankr launch pair — new token markets can launch directly from X
ForeverMoney announced at 11:25 AM ET that TAO is live as a Bankr launch pair for Base and Robinhood Chain. Bankr’s own documentation explicitly lists TAO as a supported quote token on Base and describes launching through X. New projects can create trading pools where their token trades against TAO, giving TAO another use as liquidity for new markets. This also gives builders and AI agents a simpler route to launch around TAO. If those projects attract users, they could bring more TAO-denominated liquidity, trading activity and new communities into the ecosystem. The Robinhood Chain availability is attributed to ForeverMoney’s announcement. Added after today’s full research cutoff.
A live launch-pair integration expands TAO’s utility and distribution through Bankr and X. Its ecosystem-wide reach earns five stars; future liquidity and adoption depend on actual use.
Emission follow-up: some allocations return, while others fall to zero
At 11:07 AM ET, UMI’s observed subnet allocation is 0.107% and DSperse’s is approximately 0.00002%, versus zero in yesterday’s saved snapshot. Leadpoet, Hotfloat and Feval now show zero after previously positive readings. These are allocation observations, which alone do not prove an administrative switch. Epago is different: the successful disable transaction is documented in its five-star story above. Pool-side TAO emissions, participant alpha emission and validator burn settings are separate mechanisms; one number does not show whether every miner is getting paid.
The on/off story is still changing. Separating observed allocations from an actual administrative transaction makes the update usable and accurate.
Crucible Wallet reaches iOS in the US
Crucible Labs announced US iOS availability on September 9, adding to its existing Android and Chrome distribution. The team lists Unified Swap, portfolio tracking and Ledger support. This announcement was posted shortly before yesterday’s cutoff and is included here as a late-found update with its actual date retained.
A new mobile distribution channel makes Bittensor tools available to more users and warrants a clearly labeled catch-up.
StonkFun expands stock-paired launches and reports September 9 revenue
StonkFun says new coin-launch pairs include tokenized RDDT, BABA, COST, DELL, UPS, RBLX, SHOP and others, with GRND added in a September 10 post. Sunrise separately confirms 20 US stocks issued by Backpack Securities went live on Solana. StonkFun’s September 9 figures report $1,355,566 in platform revenue, $814,028 in buybacks and 4.55 million tokens burned. These are platform-wide reported figures, not TAO-only activity. The new story is the expanded asset range; TAO pairing was already covered in earlier editions.
The user-followed launchpad has a material product expansion and fresh business figures, with its connection to TAO kept proportionate.
Tensia’s bots now send subnet and chart links to Taoswap
Tensia says all subnet and chart links in its bots now open Taoswap. Its announcement points readers to subnet comparison, portfolio metrics and validator baskets showing allocation weights. This is a routing and data-access integration, not a new token listing or trading recommendation.
A shipped integration helps users inspect the network and validator allocations.
dFusion opens Persona conversations to their owners
dFusion says users can now talk with the Persona built from their profile, answers, corrected examples and selected public posts. It describes answers that show their supporting material and acknowledge missing evidence. Opening these Personas to everyone else is the next step, so the announcement is not being described as a fully public launch.
A concrete feature moves from profile-building to interaction, with the present access limit retained.
Watch the RL Bryer replay: 2009 Bitcoin OG’s next big bet
Yesterday’s conversation with RL Bryer is now available as a public replay, running 1:02:17. The video library has been updated from Upcoming to Replay. The latest members briefing remains the September 8 news and TAO price-action video at 18:35; no newer upload or scheduled stream appeared in the checked channel samples.
The viewing information is refreshed to match the actual broadcast and current member upload.
ORO joins Y Combinator’s Fall 2026 batch — SN15 takes the spotlight
ORO has been accepted into Y Combinator’s Fall 2026 batch. This is confirmed on YC’s own company directory, as well as ORO’s September 8 announcement and the OpenTensor Foundation’s follow-up. The team is building open models and daily evaluations for AI shopping agents. YC lists Christopher Golda as its primary partner. SN15 was up 17.34% against TAO over 24 hours at 11:13 AM ET. The accelerator acceptance is confirmed; no separate funding amount was disclosed in the announcement.
A major outside accelerator has accepted a live Bittensor project, with confirmation from both sides and a strong market response. This deserves the lead position.
BitMind removes its 80% burn; SN34 emissions are positive again
BitMind’s September 9 v5.0.3 change sets the baseline validator burn to zero, restoring the full published 40% image / 40% video / 4% audio / 16% generator allocation. This replaces yesterday’s 80% baseline burn. Separately, the 11:13 AM ET network snapshot shows SN34 receiving 1.208% of subnet emissions, compared with zero yesterday. The code and subnet-level restoration are confirmed; this snapshot does not establish that every validator has upgraded or every individual miner has received a payout.
Both the incentive code and observed network allocation changed materially. This is a direct follow-up to yesterday’s burn story, with the new settings clearly separated from individual payouts.
Leadpoet adds LinkedIn and email integrations, new customers and an OpenTensor engagement
Leadpoet says LinkedIn and email integrations are built into its platform and will soon let users engage leads directly. Its September 8 update also reports new customers across logistics, proptech, AI and textiles, plus onboarding OpenTensor to help drive Exploit conference attendance. An improved incentive mechanism and stronger sales agent are targeted for this week. Customer and integration claims are attributed to Leadpoet; the post does not announce that the new mechanism is already live. SN71 was up 7.03% against TAO at 11:13 AM ET.
Specific product integrations and reported customer acquisition are commercially meaningful. The launch target remains a plan and is not confused with an already shipped release.
Pareton reports 252.4 tokens per second on H200, beating its stock-B200 comparison
Pareton published a Qwen3.8-27B-FP8 comparison: its engine on an H200 produced 252.4 tokens per second versus 106.2 for stock vLLM on a B200, using the same prompt, output length and batch size of one. The team attributes the gain to multi-token prediction and links both runs. This is a specific workload result reported by Pareton, not a claim that H200s outperform B200s generally. The linked benchmark page could not be independently retrieved during this scan.
A substantial, reproducible-looking inference result on cheaper hardware is material product evidence. Hardware, model, workload and the independent-verification limit are explicit.
Pareton gets a new top-conviction hotkey in the subnet-king ranking
SubnetRadar’s September 9 record identifies 5HK5tp…EpeN as Pareton’s new highest-conviction hotkey, displacing 5GKbzK…4LWm at block 9,029,027. The reported conviction is 15,068.585. This confirms a ranking change in the provider’s chain-event feed; it does not by itself establish a completed subnet ownership transfer. SubnetRadar describes additional age and eligible-alpha threshold conditions for a takeover.
A specific governance-ranking change deserves coverage, with the distinction between leading the ranking and actually taking ownership kept explicit.
Vidaio publishes a full redesign centered on independently checkable scoring
Vidaio announced a rebuilt subnet architecture and a new whitepaper on September 8. Its stated design lets others recompute scores from public evidence, with results anchored on-chain, reducing reliance on a private scoring process. The published whitepaper is available in the project’s miner repository. This confirms the redesign announcement and its documentation; the announcement alone does not demonstrate that all parts of the new mechanism are active in production.
A redesign of the core scoring and verification process changes what miners produce and how holders can assess the subnet. It is more consequential than routine maintenance.
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.
