## (daily refresh, 1)

- aider: rotation re-verification, every cited reference live, stalled status unchanged (no commits since 2026-05-22, v0.86.0 still the newest tag).
- auto-dev: rotation re-verification, JetBrains counters refreshed (Xiuper 671, Sketch 4,548) and VS Code installs about 5,543, push and release record unchanged.
- avante-nvim: tagged train reached v0.4.8 (October 28, lockfile and CI-secret release-engineering bodies), v0.4.2 still the newest feature release, push date refreshed.
- crush: rotation re-verification, commits refreshed to 4,375, v0.98.2 still the newest release.
- deepseek-harness: tagged train reached dsh-v0.2.2-alpha.2 (October 28) opening the 0.2.2 line after dsh-v0.2.1-rc.4 fixed ACP stdout diagnostics; prerelease-only.
- warp-agent-cli: rotation re-verification, every cited reference live, pricing page unchanged against the 2026-10-20 reading.
- ante: recorded v0.3.0 (October 28), the 0.3 credential reference-language break.
- atomic-agent: recorded v0.6.13 with desktop-v0.0.7 (October 28), the agent-runs-setup-steps release.
- goose: recorded v1.57.0 (October 28), allowlist hardening and OpenRouter variant-suffix resolution.
- juggler: recorded v0.7.9 (October 28), auto-continue on reset-time usage limits, Pinboard file links, Compact/Hand off, /init AGENTS.md drafting.
- om-o: recorded v5.1.51 (October 28) with 5.1.49/5.1.50 unpublished inside it, the task-host liveness and DAG-recovery gate rework.
- oh-my-pi: npm train reached 18.13.3 (October 28), push date refreshed.
- qwen-code: recorded v0.26.3 stable (October 28) with sdk-typescript-v0.1.25, push date refreshed.
- autohand-code-cli: recorded first stable v0.9.9 and npm 0.9.9 (October 27); corrected the chart markup (light source and fallback dropped theme=dark).
- nanocoder: npm downloads refreshed to 8,343 monthly (window September 29 to October 28), push date refreshed.
- zerostack: crates.io counters refreshed to 3,755 total (726 recent), push date refreshed.
- gemini-cli: commits refreshed to 6,553.
- harnesses: push-date refreshes recorded for exo, fx, gptme, jcode, kilo-code, mimo-code, omnigent, and openhands (all 2026-10-29).
- claude-mem: recorded v15.5.0 (My Work and My Focus component refresh, opt-in OpenAI-compatible Observer batching, explicit summary-provider selection, redaction broadened to bare token/secret/api_key keys, Authorization bearer tokens, MYSQL_PWD/dbPassword pairs, and Slack xapp-/xoxe- tokens); npm re-verified in sync at 15.5.0.
- okf-agent-memory: recorded v0.7.2 (October 28): pull-path security hardening from a responsible disclosure (remote-derived paths rejected for traversal, absolute, hidden-dotfile, and symlink escapes across hub pull/sync, pull, restore, and vendor remove; staged atomic registry installs; symlinked-CWD false positive fixed); push date refreshed.
- deja: the tracked post-v0.23.1 train shipped as v0.23.2 (recall twice as fast and the index a third smaller on a 19k-session store, receipts provenance labeling, recalled rules demoted to untrusted reference data, subagent recall injection stopped after a zero-use measurement, Qwen Code External Context MCP profile, Cursor marketplace plugin, DEJA_RECALL_GLOBAL).
- sno-station: corrected out of rotation: the cited v1.6.0 release no longer exists upstream as a tag or release; v1.7.1 (October 29, failed saves survive with retry caps) recorded as the newest published release and the version-history deletion added as a caution.
- cognee: rotation re-verification, every cited URL re-fetched live, pricing unchanged (Free $0, Standard $1.00/1M plus $5/workspace, Enterprise BYOC); the README's new BEAM self-run benchmark table (0.79 at 100K, 0.67 exploratory at 10M) and Postgres single-instance demo path recorded, integrations now include OpenClaw and Hermes Agent.
- longmemory: rotation re-verification, every cited URL re-fetched live; contributors 51 by API with the README table restored; the README's documented surface grew (multilingual recall, answer_from_evidence, governed assets, new integrations) while npm stays frozen at 1.3.3; PyPI longmemory-sdk 404 again on 2026-10-29.
- memvid: rotation re-verification, every cited URL re-fetched live, no facts changed (no push since 2026-07-14, both registries frozen at 2.0.160, the archived dispute page intact).
- plur: rotation re-verification, every cited URL re-fetched live; spec v2.1 temporal-validity windows and episode records, the separate answer-accuracy figure (60.5 percent with the reranker stack vs 52.0 full-context and 5.5 none), provenance.origin actor stamps, and size-tiered indexing recorded; npm in sync at 0.22.1.
- memmachine: rotation re-verification, every cited URL re-fetched live, no facts changed (PyPI still 0.4.0, thin-discussion record unchanged).
- file-based-agent-memory: See also trimmed to five items (OpenCode link dropped) for the section's range.
- memory-feature-matrix: PLUR's contradiction and audit cells moved to the spec v2.1 temporal-validity windows and the provenance.origin actor stamps.
- memory: matrix matched to membership (34 columns, case-insensitive sort, all rows at header width), category index list current, chart markup clean (dark source only carries theme=dark), no membership changes.
- agentsmesh: repository removed from GitHub by 2026-10-29 (404, no redirect); status moved to repository-removed with agentsmesh.ai, docs, and Docker Hub still online; out-of-rotation re-verification.
- orchestration: matrix AgentsMesh status and license cells moved to repository-removed, choosing entry rewritten as wind-down risk.
- autogen: rotation re-verified, all sources live; a 2026-10-28 docs commit added the archive caution to the documentation landing page.
- ax: rotation re-verified, all sources live; push refreshed to 2026-10-29, launch thread steady at 666 points.
- claude-squad: rotation re-verified, all sources live; Homebrew installs refreshed to 5,142 over 365 days as of 2026-10-29.
- conductor: rotation re-verified, all sources live; recorded 0.92.0 (October 23, iOS remote control of the Mac's local workspaces).
- cmux: rotation re-verified, all sources live; pricing page re-checked unchanged against the recorded tiers; v0.65.2.2 (2026-10-27) and daily pushes confirmed.
- open-swarm: rotation re-verified, all sources live; pivot state unchanged (proprietary signed-builds shell), no factual drift.
- yao: rotation re-verified, all sources live; no factual drift, v1.0.0 (2026-10-13) and the 2026-10-19 push unchanged.
- atlas: rotation re-verified, all sources live; no factual drift.
- crystal: rotation re-verified, all sources live; no factual drift in the death record.
- gastown: dependency-upkeep record extended through 2026-10-28, still no release since June.
- agent-orchestrator: release train moved to v0.14.7 (2026-10-28), push 2026-10-29.
- agent-swarm: release train moved to v2.7.3 (2026-10-29).
- crewplane: release train moved to v0.4.0 (2026-10-29, workflow-execution safety fix).
- jean: release train moved to v1.0.13 (2026-10-28, mid-turn steering and auto-continuation after usage-limit resets).
- mastra: npm line moved to @mastra/core 1.80.0 (2026-10-28, per-resource user-scoped OAuth connections in @mastra/connect).
- oh-my-codex: latest release moved to v0.21.39 (2026-10-28, GPT-6.1 Terra catalog recognition and state-root publication fix).
- omnara: omnarad line moved to v0.2.14 (2026-10-29, canonical daemon binaries).
- openharness: desktop line moved to v1.3.36_desktop (2026-10-29), push refreshed.
- orca: latest desktop release moved to v1.4.244 (2026-10-29, four interface languages, faster workspace switching).
- ruflo: release line extended with v3.63.8 (2026-10-28, standalone @claude-flow/memory 3.1.0, pattern store and search on one backend).
- sim: release line extended with v0.9.35 and v0.9.36 (2026-10-28, library guides), push 2026-10-29.
- ag2: push refreshed to 2026-10-28.
- agent-manager: push refreshed to 2026-10-28.
- orchestration: agent-teams, agno, apowerb, archon, autogpt, bernstein, buzz, camel, cli-agent-orchestrator, crewai, dify, gascity, graphcode, happy-coder, herdr, langgraph, lobehub, looptroop, multica, orkas, parallel-code, praisonai, raven, scion, squad, superset, t3code, tuios, worktrunk, yylo: push or commits-through evidence refreshed to fresh as-of dates in the volatile pass.
- paseo: commit count refreshed to more than 6,100 on main as of 2026-10-29.
- vibe-kanban: post-shutdown community commit count refreshed to fourteen through 2026-10-28.
- chutes: monthly plans withdrawn from sale (PAYG only, legacy plans run out their billing period per the FAQ, no free tier stated) and the private GPU rate rose to $2.25/hour with a $6.75 deploy fee; two price history rows appended
- chutes: rotation re-verification, every cited URL re-fetched live, Kimi-K3 PAYG holding at $1.85/$9.50 with cached-input columns now on the table
- opencode-go: Go Teams joined at $20 per member per month and the catalog churned from 33 to 30 entries (Kimi K3.1 Pro and Gemini 4 Argon in, Space Bunny Free and Cat Jam Free out, Step 5 Preview paid); price history row appended
- opencode-go: rotation re-verification, docs and product-page model counts matched at 30 for the first time since September
- opencode-zen: gateway renamed OpenCode Console on the docs, the credit-purchase processing fee rose to 4.6% + $0.31, and the free roster fell from fourteen rows to ten; price history row appended
- opencode-zen: rotation re-verification; the GLM-5.3-Flash price-gap caution re-based onto the OpenRouter catalog (3.3x input, 6.7x output) with MiniMax M2.7 converged to identical rates
- supergrok: the pricing FAQ moved API usage outside the shared weekly pool (billed separately at standard rates), against Warp's still pool-inclusive doc of 2026-10-20, and the seller rebranded to SpaceXSI; price history row appended
- supergrok: rotation re-verification, ladder and API prices unchanged
- kimi-code: OpenRouter kimi-k3 listing's twentieth recorded move, $0.192/$15.00 per million (cache reads $0.182), input floor to Relace at $0.1824/$9.34, roster steady at 30 endpoints; kimi-k2.7-code and kimi-k3.1-pro held
- openrouter: fee state re-verified unchanged (5.5% Standard, 8% Business, BYOK $25,000 then 5%)
- experiential: pricing page rebuilt again, Free lost its 500-credit grant for 1M-tokens/day core metering, paid plans gained 9B/36B/54B daily ceilings, CClef replaced Jev in the zero-fee core set, a per-plan discount table returned, and the one-key count moved to 943 at unchanged plan prices; price history row appended
- experiential: PyPI advanced to 0.7.223 at 226 releases, repo pushed 2026-10-29
- litellm: v1.106.6 stable 2026-10-28 with v1.105.7, PyPI latest 1.106.6, and v1.108.0-rc.1 published 2026-10-29 on the never-tagged v1.108 line
- ollama: v0.40.13 (2026-10-28) extended LFM3 including lfm3-audio and lfm3-omni to Linux and Windows
- magnitude: @magnitudedev/cli 0.2.14 added Fastest and Most-intelligent catalog sorting
- bifrost: pushed 2026-10-29, releases and description unchanged
- llm-gateway: pushed 2026-10-29, v1.23.0 still latest
- localai: pushed 2026-10-29, v4.12.1 still latest
- model-access-feature-matrix: 26 cells moved across the chutes, experiential, opencode-go, opencode-zen, and supergrok columns with their notes
- model-access: rotation slice (chutes, opencode-go, opencode-zen, supergrok, the four oldest last-verified dates) fully re-verified; index one-liners moved for the four members; no membership change
- cross-category: none - no entrant decisions, membership, status, or price changes this run; scan carry-overs stay out of the day log per AGENTS.md

## (daily refresh, 2)

Recovery: the run died after the first commit wave with four worker reports uncollected and the people-and-publications wave written without a report; the on-disk state was verified 2026-10-30 and these lines come from the three surviving reports plus diffs (people-and-publications). The assistant-runtimes and automated-research scopes' uncommitted state was absorbed into the 2026-10-30 refresh commit, which also carries the LangBot and InternAgent notes those waves created.
- nanobot: recorded v0.4.0 (2026-10-28, GitHub and PyPI the same day: durable SQLite conversation history and browser transcripts, opt-in Preview update flow, workbench upgrades, Mermaid viewing), ending the seven-week release hold, Alpha classification confirmed current on PyPI.
- openworker: rotation re-verification, every cited URL re-fetched live (product site, OpenShell and approval-guidance docs, repo and releases APIs), v0.3.1 still latest, Windows still coming soon.
- open-webui: rotation re-verification, every cited URL re-fetched live (license FAQ, license-change thread, Algolia story search, repo and releases APIs), v0.12.4 still latest, license record unchanged.
- assistant-runtimes: volatile refreshes to fresh as-of dates across astrbot (1,083), eigent (258), hermes (51,677), octop (799), openclaw (7,798), openwork (331), qwenpaw (1,125), zeroclaw (1,106); the 2026-10-30 pass has since re-fetched every figure.
- assistant-runtimes-feature-matrix: status cells moved (Nanobot v0.4.0, AstrBot 1,083, Octop 799, OpenClaw 7.8k) and prose counts matched, stars row left at its 2026-10-14 counts.
- assistant-runtimes: q#13 openhuman re-check, stays held on the absorbed-family shelf.
- internagent: admitted as automated-research's 34th member; Shanghai AI Laboratory's twelve-task science framework with the MLEvolve subsystem folded in, its "#1 MLE-Bench" claim verified against the official table (fourth overall, 61.33, first with source available)
- internagent: matrix column added between Harmonic Aristotle and Kapso, index entry added, both in the same run
- openai-for-science: recorded the October 26 openai/math additions (721 manuscripts in 373 families, formalized wave-equation local smoothing, claimed ETH refutation out for independent examination)
- mole: recorded the unreleased verifier rework (contradictions judged from the quote with same-quantity-first ordering, competing estimates count as contradictions)
- kapso: recorded the October 28 external-contributor fixes on main (lane-env validation, knowledge-search description truncation, fallback removal)
- aris: push moved to October 29, documentation-only (search authentication and DeepXiv env loading)
- local-deep-research: October 29 batch recorded (retryable library failures, encrypted backup status, chat endpoint gating), open issues 1,267
- openresearch: post-v0.2.22 commits recorded (keyboard navigation, mobile dashboard, harness-setup streamlining), canary prerelease channel live
- aiscientist, alphaproof, arbor, evomap-autoresearch, openai-for-science: rotation re-verification complete, every cited URL re-fetched, no dead sources (two bot-walled news pages documented in place)
- automated-research: Dolphin, DiscoveryOS, and Virtual-Scientists stay closed on the dormant-below-500-stars rule with zero independent coverage
- automated-research: SGI-Bench routed to evaluation-review on the benchmark family
- model-selection-for-coding-tasks: kimi-k3's OpenRouter readings moved (aggregator listing $0.15/$20.00 to $0.192/$15.00, cache reads $0.14 to $0.182, both floors onto one Relace kimi-k3-20260715 endpoint at $0.182/$9.34, roster steady at 30).
- model-selection-for-coding-tasks: kimi-k2.7-code's OpenRouter floor held at $0.56/$3.10 with cache reads at $0.15.
- model-selection-for-coding-tasks: Anthropic, Google, Z.ai, DeepSeek, and Alibaba pricing pages re-fetched with every cell unchanged; as-of re-anchored to 2026-10-29.
- model-provider-feature-matrix: Moonshot flagship cell synced to the guide's new K3 readings (floor $0.182/$9.34 on Relace, listing $0.192/$15.00); every other vendor cell unchanged.
- section-root: nine matrix one-liners in agents/_index.md re-synced to their matrices (dates on code-review, control-planes, evaluation-review, memory, model-access, orchestration, sandboxing, spec-driven; counts and newest claims on evaluation-review, memory, orchestration, task-management).
- ai-release-tracker: rotation re-verification; archive trail resumed past the October 7 snapshot (Mistral Large 4 model page captured 2026-10-15/16, assets through 10-17), flat files still 2026-08-26, live flat files still 429, BI citation and single-story footprint unchanged
- whatmodel-sh: rotation re-verification; coverage to 888 text models from 22,353 results across 575 benchmarks as of 2026-10-29, 2026-10-19 terms disclose paid sponsor placements, and the Show HN's repository link resolves to a public MIT repository (43 stars, pushed 10-29), correcting the no-repository finding
- bestvaluemodel: volatile re-verification after the 10-29 push; 10-28 re-base corrected to 354 removals plus 73 re-scores, same-day follow-up added Gemini 4 Carbon at 59.4 (the page's new top score) and doubled Argon's prices
- vectara-hallucination-leaderboard: volatile re-verification; nothing moved since the 2026-10-25 look (README stamp still October 22, leader unchanged, pushed_at still 2026-10-23)
- trackers-and-leaderboards: matrix whatmodel.sh column refreshed for the rotation findings; membership unchanged
- trackers-and-leaderboards: no-dedicated-source check re-ran 2026-10-29, still none credible; standing decision holds
- people-and-publications (reconstructed from diffs): Pricing and Price history sections added the note format mandates to addy-osmani, boris-cherny, dex-horthy, geoffrey-huntley, thorsten-ball, dax-raad, harper-reed, jesse-vincent, mario-zechner, matt-pocock, paul-gauthier, peter-steinberger, shreya-shankar, and kent-beck (prices unchanged or newly baselined), the q#78 fill-in-rotation plan continued
- people-and-publications (reconstructed from diffs): Status refreshes to 2026-10-29 as-of dates across the category, including addy-osmani's co-authored Anthropic code-review post (2026-10-15), ghuntley's October 2-21 essay burst (fourteen essays, led by "patch the planet"), dex-horthy's aired-episode recount (#78 aired 10-20, #79 on 10-27, #80 slated 11-03), kent-beck's renaming announcement (2026-10-28), and cadence-cell updates in the feature matrix (peter-steinberger, owain-lewis, ray-amjad, matt-pocock, harper-reed, simon-willison, the-pragmatic-engineer, others)