BITTENSOR · THE NEWS ARCHIVE
Search 88 dated editions without scrolling through them one by one.
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 A past verified announcement can now be outdated.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.
SubnetRadar’s August 29 issue recorded 57% of subnet markets higher, with Feval up 23.1% and Compelle down 12.1%. It also counted 39 subnets burning at least 90% of miner emissions; those figures are a dated market snapshot, not a forecast.
Teutonic said 78% of tasks in a new evaluation database came from a new source and that its training scripts and results were publicly verifiable. Miners continued disputing whether the leading model had an unfair timing advantage; no misconduct finding was established.
Buyers encountered repeated HTTP 502 errors reaching models while worker approvals remained queued. The team continued work toward permissionless TEE onboarding, but whitelist approval was still required at cutoff.
A fail-fast initialization error in ADA-3 v3.0.1 caused valid runs to receive 0.0 after the first task, consuming miner slots and triggering emission cooldowns. The team had not yet published a final re-scoring policy for affected submissions.
OpenRoboto shipped CLI v0.1.0a3, migrated to protocol 0.9.0, fixed the LingBot adapter for evidence-based training and reordered payable-refusal handling before fee processing.
KubeTEE reported that a Kraken root validator claimed about 1,856 TAO and liquidated 5,587 alpha across four coldkeys within 24 seconds on August 27. The team linked the thin-liquidity move to root-claim mechanics; this is a documented liquidation, not an inferred owner sale.
Albedo open-sourced its models and reduced evaluation time about 20% to roughly 40 minutes. The team said duplicate model submissions were flooding the queue and continued investigating possible scoring manipulation.
Ditto raised its submission fee from 0.05 to 0.1 TAO to cover evaluation API costs, added miner-experience and marketing staff, and placed screener policy v10 into review ahead of its next release.
At the checked production snapshot, 42 of 128 active subnet markets were higher over 24 hours, 78 were lower and eight were effectively flat. Green Compute led at about 28.5%, followed by OpenRoboto at 19.0% and Finsight at 16.4%; rec4ll fell about 14.4%.
Numinous moved forecasting evaluation to rolling multi-horizon windows and enforced one agent version per hotkey. The scoring formula itself was unchanged, while the aggregation and anti-gaming policy changed.
Buyers reported DeepSeek v4 API timeouts while miners raised concerns about traffic concentrating on one node. Worker onboarding remained manually approved, and permissionless Qwen worker access had no announced delivery date.
Following a Monday infrastructure overhaul, the team reviewed on-chain submissions and possible checkpoint duplication while miners debated dataset expansion and burn pressure. No final fairness finding or new dataset deadline was announced.
Verathos confirmed miners serving HTTP endpoints were excluded from routing and scoring under mainnet HTTPS requirements. The team said miners could convert endpoints without resetting EMA scores while a broader fix was developed.
Albedo resumed pre-evaluation after storage exhaustion, pinned faster download providers and reran failed submissions. A model-reveal mechanism to reduce duplicate spam remained planned rather than deployed.
Claims resumed testnet work after miner eligibility fixes, published a run-results dashboard and set its mainnet target for about 6 PM London time on August 28. The launch itself had not occurred inside this report window.
Ditto deployed v12, cutting validator scoring cycles from about seven hours to three and durably storing benchmark model calls for audit. The team reported 366 submissions during the prior week and placed a coding competition on the v13 roadmap.
Swarm disabled its overloaded owner validator and cleared more than 600 erroneous zero scores. Other validators continued evaluating normally; the team attributed the incident to validator resource limits and began work on a longer-term safeguard.
HTX opened TAO-USDT spot trading at 13:00 UTC on August 27 after enabling deposits. This expands direct exchange access to TAO; it is not a subnet-token listing.
SubnetRadar recorded SN61 RedTeam up 18.5%, SN92 MicroTensor up 16.3% and SN70 up 14.1% over 24 hours. SN78 Vocence fell 19.3%, SN49 Nepher Robotics fell 9.7% and SN80 OpenRoboto fell 7.1%; SN51 lium.io led tracked chain buys at 311 TAO.
SubnetRadar recorded a single-wallet purchase and stake of 388 TAO into SN4 during the reporting window. The on-chain move does not establish the wallet owner's identity or motive.
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.