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 A past verified announcement can now be outdated.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.
Minos miners reported that BAM files served after the weekend update did not match the declared SHA256 hashes, preventing participation in rounds. The mismatch remained an active operational incident at cutoff.
Ditto released v0.133.0 with per-node memory-screening controls, staged overflow capacity and OAuth support for model-context-protocol integrations.
An on-chain wallet coldkey changed from 5DXiV…XUWg to 5Ft26…h9f4 at block 8,965,844. Public evidence does not establish the wallet role or motive, so this is reported only as an executed key rotation.
At the production snapshot, 51 of 128 subnet markets were higher over 24 hours, 64 were lower and 13 were nearly flat. Subnet 70 led at about 44.4%, followed by Capcomp at 36.4% and RedTeam at 30.9%; MicroTensor fell about 16.9%. SubnetRadar separately counted 40 subnets at or above 90% miner burn.
Teutonic identified teutonic-II-110B-A7B-test1 as its new top model. Public discussion included miner-reported earnings, but those dollar figures were not independently verified and are not treated as a confirmed metric here.
OpenRoboto initiated a burn-rate reduction toward zero and activated a new candidate submission. The candidate had not yet been confirmed as champion at the reporting cutoff.
Verathos released audit hardening, wallet-account and API-key changes, then pinned a portable instruction baseline for processors affected by an AVX2 serving failure. The compatibility rollout was scheduled for the following week.
Beam reported automated transfers back online after its infrastructure move and said one terabyte had been processed during the transition. This is the verified recovery from the prior day’s outage report.
Minos traced miner 51’s scoring failure to GVCF output containing excessive NON_REF records. The team planned to disable GVCF and BP_RESOLUTION flags and change emit_ref_confidence settings in an upcoming release.
Claims increased burn_half_life to 4,050 blocks, set a three-hour query interval for six-hour weight updates, and clarified that rankings average the last three rounds. It also identified a dashboard status-label bug.
Sundae Bar changed similarity scoring to penalize concurrent near-duplicate submissions and removed flagged entries and accounts. The team said QA was not yet complete when the fix was released.
At the catch-up publication snapshot, 51 of 128 subnet markets were higher over 24 hours, 65 were lower and 12 were nearly flat. Capcomp led at about 46.9%, followed by RedTeam at 32.4% and Cortex at 29.4%; ChronoLLM fell about 11.8%.
An OpenRouter API incident temporarily stopped validator evaluations. ORO said service recovered and failed evaluations did not charge miners or change their scores.
A 401 failure affecting UID 234 was traced to signed requests missing a receiver field. The miner changed verification handling, while separate proposals to restore uniqueness filters and widen the seed range remained under debate.
OpenRoboto activated an xArm6 evaluation track capped at 30 models, with a 1.5 TAO submission fee and 20% of subnet emissions assigned to its prize pool.
Albedo added restricted-bucket uploads backed by on-chain commitments so models can remain private unless they win. Public submissions remain available, and validator and frontend fixes shipped alongside the feature.
Beam moved Beamcore services toward a multi-zone Azure setup. The public brief recorded more than seven hours of service disruption without a restoration estimate at that point.
Claims went live with 256 UIDs and said it planned to prune to 64 after one to two weeks of weight stabilization. The team prioritized newer miners while emissions awaited manual activation.
Ditto prepared Screener v11 with stronger top-five re-evaluation and moved policy updates to a daily review cycle. The team said confirmed cheating would remove emission eligibility.
The SBC21 challenge opened through August 30. The team disqualified mass submissions and began tightening similarity checks and validator-variance handling after accounts tried to exploit simultaneous submissions.
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.