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 A past verified announcement can now be outdated.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.
Bitcast’s current dashboard reports $422,300 revenue and $261,800 buybacks
Bitcast’s September 8 dashboard displays approximately $422,300 in cumulative marketing revenue and $261,800 in SN93 buybacks, a 62% recycling ratio. Its revenue page supplies transaction links and shows its latest listed receipts dated September 4. These are project-published cumulative figures, not new September 8 revenue, an independent audit or a promise that future revenue will be bought back.
The primary dashboard lets readers check the revenue discussion against disclosed totals and transaction dates.
Albedo announces a scoring pipeline based on multiple leading-model trajectories
A September 7 Albedo-linked announcement describes using multiple state-of-the-art trajectories to determine the progress expected from competing 35B models. Today’s channel brief also discusses the scoring change. The detailed activation rules and complete evaluation specification were not independently retrieved, so the implementation status remains developing.
A changed evaluation target matters to the coding-agent competition, but the short announcement does not supply enough detail to state the full rules.
TAOstatus publishes v3 code; reported SN104 launch still needs reconciliation
TAOstatus merged v3 and published documentation for pooling miner-supplied LLM API capacity. Today’s team-channel brief reports a live SN104 validator and real traffic. However, the public README still specifies testnet 501, and the product documentation says recurring health checks are the current source of usage data, with other consumers planned. The software is available; the mainnet and external-workload claims remain developing until those sources are reconciled.
The launch report is material, but the primary documentation contains concrete limits that should not be erased by an aggregator’s headline.
TensorUSD adds daily burn growth when the winning evaluation gets stale
TensorUSD merged a rule that increases the burn share by ten percentage points for each UTC calendar day since the winning evaluation date, capped at 100%. The remaining share goes to the selected miner; if that miner is no longer registered or no reward remains, the code burns the full allocation. This is a validator reward rule, distinct from TUSDT borrowing terms or advertised yield.
Stale evaluations now affect the distribution of rewards directly, creating a reason to keep results current.
SOMA adjusts how coding quality and token savings combine in its score
SOMA merged a smoother quality penalty for its software-engineering tasks. Below half of baseline task success, only the quality penalty contributes; from 50% to 80%, the score blends toward token efficiency, and at 80% or above the token score applies fully. This changes how compressed context is judged when it saves tokens but reduces successful outcomes.
A scoring change at the boundary between cost savings and useful work matters to both miners and product quality.
Coldkey swap moves control from 5F6aR…iZ4D to 5HTeZ…cc8a
A finalized-chain coldkey swap executed at block 9,015,081 on September 7 at 10:52 AM ET, moving from 5F6aR…iZ4D to 5HTeZ…cc8a. The public event does not identify a subnet-specific role or explain the reason for the swap, so this report does not infer identity, motive, a security incident, ownership transfer or a sale.
Coldkey changes can alter operational control and deserve a permanent audit trail even when the affected role and reason are not public.
Subtensor release 455 merges tighter proxy permissions and a crowdloan fix
The core repository merged release 455 and a fix that prevents NonTransfer and NonFungible proxies from using EVM, Contracts or Crowdloan calls to move value indirectly. The fix also removes a temporary block on crowdloan contributions in the code. Mainnet runtime activation was not independently confirmed, and the change itself does not establish that an exploit occurred.
Proxy permission boundaries affect network security and delegated account control. The merged code is important even before mainnet activation is independently established.
CONST follow-up: Cats backs the TAO-reward community as the token retraces
Cats publicly described building CONST into a longer-term community project and said he exchanged wrapped TAO received from holding it for native TAO. The token’s market page also links to @Cats_CR. At 10:42 AM ET, DEX Screener showed the CONST/TAO pool at about $620,000 market cap, $671,000 daily volume and a 54.09% rolling 24-hour decline in USD. This updates yesterday’s rally coverage: it remains a Solana community token, and no Jacob Steeves ownership or official Bittensor affiliation is established.
The TAO meme follow-up now includes direct community leadership statements and the sharp reversal, rather than carrying forward yesterday’s price narrative.
Sunrise launches inside MoonPay’s PayBox
Sunrise and MoonPay announced a live PayBox integration for accessing supported Solana token markets from Claude, ChatGPT or Grok. This expands the distribution of the asset gateway behind TAO’s recent Solana rollout. MoonPay says the feature is available only in eligible jurisdictions outside the United States. The announcement concerns the gateway; no TAO-specific transaction was executed or independently tested in this scan.
Confirmation from both providers establishes a new distribution channel for Sunrise assets, with eligibility limits that materially affect access.
StonkFun reports $1.20 million daily revenue and adds tokenized Nike pairs
StonkFun’s September 7 report lists $1,198,987 in platform revenue, $719,516 in buybacks and 5.67 million STONK burned. Separately, Stonk and Sunrise announced NKE-paired coin launches on September 8. These are platform-wide figures and a new tokenized-stock quote asset. TAO pairs were already announced on September 6; today’s NKE expansion is not another TAO-pair launch.
The fresh revenue report and expanded pair options update the launchpad story while keeping platform economics separate from TAO-specific demand.
Exploit Summit previews Yanez’s next product layer and adds ChronoLLM as a sponsor
Exploit Summit says Jose Caldera will present Yanez’s next product layer and a permissionless model for the agent economy at the Montreal event on September 28–29. The organizer also announced ChronoLLM, SN38, as a sponsor. These are conference and product-preview announcements; they do not establish that Yanez’s next layer has launched.
The organizer supplies concrete upcoming product and ecosystem milestones without requiring speculation about what will be unveiled.
Neuralteq adds a stake-weighted validator-support index
Neuralteq’s updated subnet dashboard adds a relative support index based on the top 20 root validators, weighted by their stake, with the leading subnet normalized to 100. A separate allocation factor measures each subnet’s share of their combined root stake. The layout also exposes risk and immunity information more clearly.
A new public comparison tool is useful to readers following validator allocation and subnet support.
BitMind releases v5: continuous King of the Hill rewards replace round-based escrow
BitMind published v5.0.0 on September 6, resolving yesterday’s pending-release status. Each discriminator lane splits rewards 85/10/5 between its current king and two predecessors; the overall allocation remains 40% image, 40% video, 4% audio and 16% generators. A challenger must improve the SN34 score by 0.01 on the same benchmark. Discriminator payouts activate only after GAS explicitly enables them and the start time passes; installing v5 alone is not activation. Generator rewards continue during warmup.
A published major release changes how detector miners win, retain and receive rewards. Its explicit activation conditions are essential to understanding the rollout.
OpenRoboto allocates 85% of track emissions to two physical-robot competitions
OpenRoboto’s September 7 documentation states that the active track allocation is now 15% simulation, 42.5% real-robot π0.5 and 42.5% real-robot LingBot-VLA 2.0. Each physical track has its own seasonal pool, with 95% of that pool for the champion and 5% for qualified entries under the existing vesting terms. Previously accrued rewards are not redistributed. A shared workstation interface is also published; the documentation itself is not notice that a new submission window is open.
Allocating the large majority of emissions to physical robotics is a major economic and product development, affecting participants across all three tracks.
Chutes appoints Christian Hyland as CEO
Chutes announced Christian Hyland as its new CEO on September 7. The company says he has spent several months helping reshape its approach to scaling and operations, and cites his background at Coinbase, Toku and First Wave AI. The appointment is a leadership change at one of Bittensor’s major inference providers; the announcement did not disclose a financing round or new customer contract.
A new CEO at a major inference subnet is a significant strategic development and deserves prominent placement.
Beam launches production P2P tunnels and previews a Zapier integration
Beam announced that peer-to-peer tunnels are live in production, connecting private infrastructure, cloud services, AI workloads and remote participants for real-time data streaming. Its new September 7 follow-up previews a two-way Zapier integration: Beam transfer, Room, Tunnel and stream events would trigger application workflows, while application events could start Beam operations. Tunnels are announced live; Zapier support is described as coming, with no release date or separate Zapier confirmation located.
A live connectivity feature plus a planned route into widely used business applications substantially expands the product’s potential utility.
Pareton reports 3.5× faster Qwen inference after its first week of competition
Pareton reports a 3.5× end-to-end improvement over stock vLLM for Qwen3.8-27B-FP8 on one H200, measured over SWE-agent traces. Its figures include median request latency falling from 727 ms to 190 ms and throughput rising from 58.7 to 221.8 tokens per second per request. The winning patch combines the model’s MTP head with fused kernels. The project reports 25 rounds, 74 submissions and 31 miners. These are project benchmark results, not a universal speed guarantee.
Specific performance gains, workload conditions and participation counts provide substantial evidence of the competition’s output.
Lium reports 14,499 tokens per second on a single GPU
Lium’s September 6 demonstration claims Qwen3.6 35B ran at 14,499 tokens per second on one rented GPU at 85% lower cost than OpenRouter. The public announcement attracted substantial attention, but the retrieved post does not specify enough batching, hardware and pricing conditions to reproduce the comparison. Treat the number as Lium’s demonstration claim, not a per-user serving speed or a general cost benchmark.
A widely circulated inference demonstration is material compute news; its unusually high throughput needs the missing workload context.
Blockmachine adds Mantle RPC with request-based pricing
Blockmachine announced dedicated Mantle RPC support on September 7. Its product article describes a free development tier, Pro plans starting at $25 per month and request-based billing, with account-generated endpoints for existing EVM tools. This extends the SN19 service to Mantle developers. The provider’s claimed savings of up to 60% were not independently benchmarked, and offering chain support is not evidence of a commercial partnership with Mantle.
A concrete new supported chain and published product access create a new customer-facing use for the subnet.
SayGM approaches 100 billion processed tokens
SayGM reported 96.48 billion recorded network tokens processed on September 7, including more than 30 billion output tokens. The total remains below the 100 billion milestone. It is a project-reported usage measure, not a count of customers, revenue or independently audited paid requests.
A concrete usage milestone is significant evidence of the scale being reported by a live inference service.
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.
