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 A past verified announcement can now be outdated.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.
Cathedral announced a live compute service priced per verified CPU run, with free tests and an hourly option for persistent workers. The team also said it is moving under a new owner and ultimately into its own slot; live chain metadata still labeled SN39 deprecated at publication, so the slot transition remains developing.
A traffic surge exhausted GPU memory across GLM-5.2 capacity and produced 404 failures in multiple datacenters. The team said it added rate limits and safeguards; planned Kimi K3 vision support had not been independently verified as live by the cutoff.
A concentrated public-code review produced fixes that bind signed attestation bodies to their views, separate registration deposits from slashing reserves, repair live-surface registration and deposit handling, and add oracle guardrails and stake-layer wiring. These are merged code changes; production activation was not independently confirmed.
Round-eight submissions closed and evaluation began, with one active agent per hotkey enforced for new entries while existing multi-track agents were grandfathered. Phylax also published all 80 slot states and a local evaluation harness using a 47-artifact corpus.
The public BASE repository changed its configured emission share to 100% for PRISM and pinned the change in staging. This is strong code and staging evidence, but it does not by itself prove that live validator weights or network emissions had already changed.
The team said owner alpha may be sold through the liquidity pool after a public 24-hour notice because proposed over-the-counter transactions did not proceed. It supplied neither a fixed sale time nor an alpha amount, so this is a developing treasury-policy signal and does not qualify as a scheduled unlock.
SOMA began baseline and miner runs across its stage-three evaluation. The team hot-fixed the qualification list and said dashboard scores still included SC2 results that will not count in final scoring; several miners reported withdrawing because DeepSeek/OpenRouter evaluation cost was about $180 per hotkey.
Affine reset evaluation after removing models it considered exploit-based, added 30 GPU replicas and reported cutting evaluation time in half. Miners disputed the exclusions and the lack of advance notice, so the reset and capacity increase are verified through the official channel while the exploit characterization remains contested.
SubnetRadar now labels conviction mechanics live on spec 447: for an eligible subnet older than one year, one hotkey must strictly exceed 18% of eligible alpha for ownership to transfer. This is a material status change from yesterday, when only the main-branch merge was verified.
Fresh conviction events inside the window included an approximately 90,800-alpha lock on SN112 and a 12,200-alpha SN93 position switching to perpetual. Neither event is a scheduled unlock. A later 334-alpha Teutonic switch to decaying occurred after the 10:00 AM cutoff and was excluded from the reporting window.
SubnetRadar counted 44 of 128 reporting subnets at or above 90% miner burn, unchanged from yesterday. Chutes remained 57% of the way to its first 10.5 million-alpha halving threshold, with an estimated 427 days remaining. No verified registration, deregistration or completed halving surfaced inside the window.
At the 6:45 PM ET publication snapshot, 48 of 128 subnet markets were positive, 79 were negative and one was flat. Cathedral/SN39 gained 27.1% while conjectures/SN66 lost 15.0%. Because publication was delayed, these figures are labeled as a publication snapshot rather than the 10:00 AM cutoff.
The late screener snapshot placed Chutes at 11.84% of displayed emission, lium.io at 10.33%, Minos at 8.29%, Affine at 7.88% and Targon at 7.49%. SubnetRadar separately flagged Green Compute rising to 1.03%. These are point-in-time shares, not evidence of revenue or future performance.
NIOME moved beyond yesterday's scheduled-release notice: the validator produced valid scores and committed weights after query failures and operational instability. The team also added coldkey and IP filtering around the revised genomic-profile scoring system. The successful weight commit is verified in the public repository; broader stability remains developing.
Ridges merged an upload-credit path that lets an eligible uploader consume a one-time credit instead of burning alpha. The client stops before an upload if neither a usable credit nor a permitted burn path is available, and it can retry with the returned credit identifier. This changes submission mechanics; it does not guarantee that any miner receives a credit.
Harnyx changed the current data-v8 challenger rule to require a ten-percentage-point score improvement, or a qualifying cost or latency improvement without regression, while retaining historical thresholds for older versions. It also routed similarity checks through DeepSeek on OpenRouter. The merge makes champion replacement harder to trigger through small relative differences.
Verathos merged deterministic capacity-audit selection parity, including full active-hold timing and selection behavior. The change is designed to make validator decisions reproducible across the capacity-check path; no independent performance result was available by the cutoff.
After yesterday's burn rollout, Ditto merged a defended-score crown anchor and tightened the margin described in its official brief. The team also disabled a Python query that was saturating CPU and moved upload and LLM relay work to Go. A newest-agent tie-break remained under review, so the scoring result is still developing.
Poker44 released validator v0.2.10 with a migration that converts the exact legacy 30% burn default to zero while preserving custom allocations. It recomputes the durable target and submits the new weights when chain cadence and commit-reveal rules allow. This is a validator-allocation change, not an alpha-token unlock.
A merged Subtensor main-branch change replaces spec 446's aggregate 10% conviction gate with a rule requiring one hotkey's own conviction to strictly exceed 18% of eligible alpha. Exactly 18% does not pass, and unrelated locks can no longer tip a takeover. The code and documentation are strong release evidence, but the merge alone does not prove that Finney enacted spec 447 before cutoff.
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.