Bittensor Toolkit →Find the right tool for your question.

BITTENSOR · THE NEWS ARCHIVE

Find the story.

Search 86 dated editions without scrolling through them one by one.

Reset filters

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 A past verified announcement can now be outdated.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.

2,335 matching stories

2026-10-26 edition →SN67 · Harnyxverified3 / 5 stars

SN67 Harnyx adds native Cloudflare Workers AI access

Harnyx says miners can use Cloudflare Workers AI through llm_chat, with Clef and Clairic through decision_query for decision and extraction work. Merged code documents paired account/token credentials and Cloudflare provider billing. Published rates apply, and the team says promotional credits do not reduce the task budget. This broadens provider choices, without guaranteeing task quality or lower costs.

A documented provider integration changes task execution and spending accounting for miners.

2026-10-26 edition →SN69 · Heralddeveloping3 / 5 stars

SN69 Herald announces a coverage campaign for SN51 lium.io

Herald says a new Lium campaign brief is open and invites journalists and PR professionals to register as miners and inspect its brief board. The public board was reachable, but its live registry did not finish loading in text inspection. We report the announcement without inventing a budget, eligibility tiers, completed placements or resulting customer acquisition.

A named campaign is actionable for media workers; delivery and payment outcomes are not inferred.

2026-10-26 edition →SN71 · Leadpoetdeveloping3 / 5 stars

SN71 Leadpoet prepares a reasoning-model whitelist

The current README has an Upcoming model whitelist section and provider-specific settings for returning reasoning text. The October 26 team-attributed digest asks operators to switch to approved models that provide full reasoning before enforcement. Follow the current documented list and endpoint settings, rather than assuming every model is eligible. The live enforcement start and the separate reward-rule dispute covered yesterday were not independently resolved.

A model-eligibility requirement helps miners prepare, with documentation distinct from live enforcement.

2026-10-26 edition →SN72 · StreetVision by NATIXverified3 / 5 stars

SN72 StreetVision raises its owner allocation to about 33%

A successful finalized transaction at block 9,366,833 on October 26 at 09:06:24 UTC changes the alpha allocation from 18% owner / 41% validators / 41% miners to approximately 33% / 30% / 37%. Exact five-slot vectors are retained in the evidence; burn and recycle slots remain zero. The owner bucket grows while both reward buckets shrink. Root receives its live baseline first, capped by the validator allocation; alpha validators share the remainder. Individual returns depend on stake, performance and take, and historical yields can lag.

A chain-confirmed allocation changes reward buckets without promising an individual yield.

2026-10-26 edition →SN75 · Hippiusverified4 / 5 stars

SN75 Hippius launches an S3 CDN with per-zone spend caps

Hippius announced its CDN live on October 26. Its product page describes serving a Hippius S3 bucket, or a folder within it, through nearby cache nodes, with HTTPS on a default or custom hostname, cache rules and purge controls. A monthly spending cap pauses a zone at its limit. Any-web-server origins are still labelled coming soon. This expands storage into website and asset delivery; regional performance and billed workloads were not independently tested.

A customer-facing storage-delivery product adds a concrete capability with documented limits.

2026-10-26 edition →SN78 · Umideveloping3 / 5 stars

SN78 Umi completes C6 execution; rewards still await settlement

The October 26 team-attributed digest reports evaluation execution complete for 103 models, producing 1,236 results and 618 comparisons. Certification and reward activation still await a settlement-bookkeeping fix described as tested and being prepared for deployment. C7 remains in intake without a confirmed evaluation start. This advances yesterday’s partial evaluation status, without establishing that C6 reward weights are live. The team says C5 endpoints can be taken offline while preserving hotkeys and records.

A substantive follow-up changes evaluation status and operator action while retaining the payout gate.

2026-10-26 edition →SN92 · REAXdeveloping3 / 5 stars

SN92 REAX schedules higher GPU speed caps for October 27

The published 1.2.0rc11.post7 changelog schedules validator activation for October 27 at 16:00 UTC. It raises the s1-pro speed cap from 320 to 640 per second and s1-fast from 1,280 to 2,560, leaving s1-lite and s1-cpu unchanged. It also expands capacity probes and records Tuesday/Friday 14:30 UTC engine crown slots. The daily team-attributed brief separately describes upcoming quantization eligibility. Published rules and scheduled activation do not prove tomorrow’s consensus change has already executed.

A dated release changes capacity and evaluation mechanics; activation stays explicitly planned.

2026-10-26 edition →SN98 · NeverPlayAloneverified3 / 5 stars

NeverPlayAlone changes SN98 reward allocation to 33% owner, 7% validators and 60% miners

A successful finalized transaction at block 9,371,009 on October 26 at 23:01:36 UTC changes the five-slot allocation from 0% burn / 0% recycle / 18% owner / 41% validators / 41% miners to approximately 0% / 0% / 33% / 7% / 60%. This increases both the owner and miner shares while reducing the validator bucket. Root receives its live baseline first, capped by that validator allocation; alpha validators share any remainder. These are allocation weights, not guaranteed individual returns, and historical yield estimates can lag the change. Exact vectors, successful execution and historical registration identity are retained in the evidence.

A finalized reward-allocation change affects subnet participants without establishing a new product or promising individual yield.

2026-10-26 edition →SN118 · Dittodeveloping3 / 5 stars

SN118 Ditto reverses overbroad integrity rejections and fixes reviews

The October 26 team-attributed digest says a budget-reservation bug could end reviews while budget remained, and reports a deployed fix. It also describes an overbroad integrity rule penalizing agents that lost earlier evidence during fallback; related rejections were overturned with scores preserved. A new competition remains in shadow mode without emissions. Affected submissions and the live fix were not independently audited or reproduced in this review.

Review and integrity corrections affect eligibility and trust, beyond an ordinary prize announcement.

2026-10-26 edition →SN120 · Affinedeveloping3 / 5 stars

SN120 Affine miners report stalled epochs; resolution remains unconfirmed

The October 26 digest records multiple miner reports of no new evaluation epoch over roughly a day. It also attributes a request for UID-ban logic to a confirmed team account after high scores reportedly received no emissions. A member’s optimizer-reset explanation is speculation, not an established cause. No independently verified fix or recovery time was found. This is an evaluation-incident lead, not a claim that the entire subnet or video product is offline.

Potential evaluation and payment disruption warrants coverage despite the unresolved explanation.

2026-10-25 edition →SN1 · Apexdeveloping2 / 5 stars

SN1 Apex submission timeouts are reported by a miner

A community member’s October 25 digest report says Competition 10 round 93 submissions time out while read endpoints remain available, raising a question about unused paid fees. The digest contains no team response. We could not independently reproduce the authenticated submission path, so this remains an unresolved incident lead rather than a confirmed network-wide outage or promised refund.

A specific submission failure can affect participation and paid access; its reach and resolution remain unconfirmed.

2026-10-25 edition →SN8 · Vantaverified3 / 5 stars

SN8 Vanta: 343.333333333 alpha burn verified on chain

A successful finalized AlphaBurned event on October 25 removed 343.333333333 alpha on SN8 Vanta. The exact amount, historical subnet identity and execution are preserved in the evidence. This is a transaction burn from its recorded wallet, separate from miner-emission withholding, fees and recycling; it does not establish that the current owner wallet was the actor.

An executed native-token burn is independently verifiable; the actor and economic meaning are not inferred beyond the evidence.

2026-10-25 edition →SN14 · Cacheonverified3 / 5 stars

SN14 Cacheon: 1,448.4478 alpha burn verified on chain

A successful finalized AlphaBurned event on October 25 removed 1,448.4478 alpha on SN14 Cacheon. The exact amount, historical subnet identity and execution are preserved in the evidence. This is a transaction burn from its recorded wallet, separate from miner-emission withholding, fees and recycling; it does not establish that the current owner wallet was the actor.

An executed native-token burn is independently verifiable; the actor and economic meaning are not inferred beyond the evidence.

2026-10-25 edition →SN15 · OROdeveloping3 / 5 stars

SN15 ORO reports evaluation-key abuse; affected race is rescored

The October 25 Discord digest attributes an audit result to the team: 51 agent versions across 32 hotkeys were removed after evaluation-key spending limits were increased beyond the allowed budget, and Race 2 was rescored. We have not independently audited those accounts or payouts. Separately, public code now records upstream inference costs for bring-your-own-key calls and corrects their classification. Fair comparison depends on measuring the real compute budget; those code changes alone do not prove the reported bans.

A consequential evaluation-integrity dispute affects miner eligibility and confidence in comparisons; enforcement counts remain attributed.

2026-10-25 edition →SN25 · URverified3 / 5 stars

SN25 UR moves its settlement-v2 target to epoch 4; activation remains pending

The project’s October 25 timeline says production configuration and maintenance are complete, but validator history recovery and finalized pool registration are still outstanding. Epoch 3 was missed. The revised target is epoch 4 at block 9,398,428, estimated around October 30 at 18:25 UTC, conditional on registration before that boundary. First claims are estimated no earlier than November 8 after root commitment, challenge expiry and finalization. These are the project’s block-based estimates, not confirmed activation or a promised payment date.

The revised primary timeline changes what providers can expect from settlement; deployment of services is distinct from contract activation.

2026-10-25 edition →SN34 · BitMinddeveloping2 / 5 stars

SN34 BitMind reports submission reopening after database work

Following the evaluation problems covered yesterday, the October 25 team-attributed digest says the database migration finished and submissions reopened at 18:00 UTC. It also describes Yugos as one track balancing output quality, cost and latency. This is an attributed recovery report, not independent proof that every earlier failed evaluation or payment was repaired.

A concrete recovery update is useful to affected miners; universal recovery is not established.

2026-10-25 edition →SN46 · Instantdeveloping3 / 5 stars

SN46 Instant reports higher emission withholding ahead of engine work

The October 25 team-attributed digest says the subnet moved to a 50% emission burn setting ahead of planned upgrades, with a new engine and RTX Pro 6000 mining next. It gives a target of returning to zero withholding in the coming weeks, rather than a fixed activation date. This is emission handling, not a qualifying transaction burn in our Burns archive. At the recorded 14:31 UTC alpha/TAO snapshot, the token was down about 15.65% over 24 hours; the price move does not prove a cause or an owner sale.

The reported operating change matters to miners; plans, market context and explicit transaction burns are kept distinct.

2026-10-25 edition →SN51 · lium.ioverified3 / 5 stars

SN51 lium.io enables GPU profiling for kernel developers

In an October 24 post found during the overlap review, the team says NVIDIA ncu profiling is enabled on its rental nodes, linking a profiling-enabled GPU listing for developers entering the Open Kernel Challenge. Another post shows a claimed 12-second setup of an eight-H100 rental. Profiling access is a concrete developer feature; the setup time is a provider demonstration, not a benchmark we reproduced. Separate referral payouts are conditional offers, not guaranteed GPU income.

Profiling helps developers inspect real GPU workloads; rental speed and referral examples retain their stated limits.

2026-10-25 edition →SN53 · engyverified3 / 5 stars

SN53 engy: 4,633.974128881 alpha burn verified on chain

A successful finalized transaction on October 25 burned 4,633.974128881 alpha on SN53 engy. Its matched buyback marker records 146.69 TAO spent to acquire the alpha; that TAO was not itself burned. The burn is counted once, not once per event marker. This is all-wallet subnet activity, not proof that the current owner coldkey performed it or that the funding came from business revenue.

An executed native-token burn is independently verifiable; the actor and economic meaning are not inferred beyond the evidence.

2026-10-25 edition →SN61 · RedTeamdeveloping2 / 5 stars

SN61 RedTeam announces a replacement miner dashboard

The October 25 team-attributed digest points miners to theredteam.io and says the older dashboard will be retired soon. No retirement deadline was supplied. The site responds, but its application could not be fully inspected in the text fetch; we do not claim to have verified every screen or a measured speed improvement.

The destination change is useful to miners without turning a dashboard refresh into a major product breakthrough.

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.