
## (daily refresh, 1)

- assistant-runtimes: matrix re-verification moved status cells (NanoClaw 1.1k open issues, Nanobot v0.4.2, Octop 641, AstrBot 1,298 open issues and PRs, ZeroClaw 1.4k), stars row untouched.
- nanobot: recorded v0.4.2 (2026-11-21 on GitHub and PyPI the same day), the release retiring the standalone OpenAI-compatible API server with no drop-in replacement, Alpha classification current at nineteen releases.
- openwork: pricing churned a twenty-fourth time (add-ons returned, grandfather clause left), three main prices unchanged.
- anything-llm: rotation re-verification, v1.20.0 (2026-11-17) still latest, Cloud tiers (Basic $50, Pro $99, Enterprise custom) unchanged.
- astrbot: rotation re-verification, v4.30.0-beta.1 (2026-11-18) still newest, zero-English-footprint record intact.
- assistant-runtimes: openhuman reads 42,801 stars on 2026-11-22 against OpenClaw's 395,010, q13 trigger unfired.
- gpt-researcher: rotation re-verification, status unchanged (v3.8.0 latest, PyPI 0.17.0), every cited URL re-fetched live, the November 12 pushed_at movement traced to a non-default-branch push
- futurehouse-robin: rotation re-verification, status unchanged (last push still 2026-04-21), every cited URL re-fetched live, both cited HN threads re-confirmed
- karpathy-autoresearch: rotation re-verification, status unchanged (dormant since 2026-03-26, no LICENSE file, README MIT declaration intact), every cited URL re-fetched live
- mole: rotation re-verification, status unchanged (last push still 2026-11-08, v0.2.1 still the latest release), every cited URL re-fetched live
- nanoresearch: rotation re-verification, status unchanged (last push still 2026-10-09, the single assets release still the whole train), every cited URL re-fetched live
- local-deep-research: v2.0.2 recorded (November 21): compact research-notices strip, Subscribe on every completed research, pre-upgrade backup view, session and credential hardening; November 21-22 post-release batch recorded; open issues 1,637
- openresearch: v0.2.29 recorded (November 21, 59-PR train shipping the November 18-20 wave plus its own batch); November 21-22 post-release commits (per-project settings, crash-safe chat-to-project moves, telemetry off for alphaXiv sign-ins); forks 595, open issues 54
- aide-ml: the November 21 README cleanup removing the Weco product layer recorded out of rotation (landed hours after that day's rotation fetch), the layer table now two rows, the routing claims in the caution and bottom line rewritten, push date moved
- automated-research: matrix surface cells moved to LDR v2.0.2 and OpenResearch v0.2.29, membership unchanged
- automated-research: q26 evidence refreshed, the WecoAI list woke (first content since 07-30: PRs #19/#20/#21 merged 11-21 bringing the AutoSR, AgonAlpha, SciPi vot, STEER, and eigh kernel entries plus Thoth in Implementations & Forks), yibie ticker-only 11-22, handsome-rich ticker, webfuse-com stale 09-21, AI4Scientist quiet since 11-16
- code-review: ellipsis - rotation re-verification: every cited URL re-fetched live, all metered prices unchanged (10% platform fee, $0.142/vCPU-hour, $0.024/GB-hour, $5k/$10k/$15k support packages), the $100 starting credit back on the homepage (New customers get free $100 credit, no credit card required) after being absent since 2026-11-08 while the pricing page still omits it, the hero reframed to Model Agnostic Managed Agent Platform with 11.6k-agents and 67k-repos counters, and the CLI and SDK repos pushed November 21.
- code-review: git-lrc - volatile re-verification: pushed_at moved to November 22 and main's latest commit moved from October 23 to November 20 (a release-readiness commit, local security scanners required to pass before a release is cut), v0.6.17 still the latest release at 55 releases.
- code-review: greptile - rotation re-verification: every cited URL re-fetched live, all prices unchanged (Starter free, Pro $1 per credit, Base 1, Plus 3, Apex 10, the 500-credit startup program), the 22,000-plus-teams claim and Enterprise forge row standing, YC 35 team and Active, both HN threads unchanged (253 and 17 points); no factual drift.
- code-review: kodus - volatile re-verification: the push and main's latest commit both moved to November 21 (a Kody Rules fix reading the PR description past its first 1,000 characters), no release after 2.2.17.
- code-review: open-code-review - volatile re-verification: the push and main's latest commit both moved to November 22 (a workspace named-pipe fix), v1.14.0 still the latest release at 152 releases.
- code-review: openqodex - volatile re-verification: pushed_at moved to November 22 with main's latest commit November 21 (the README redesign merge), openqodex@0.22.0 still latest, npm reads 6,874 over the trailing month to November 22 with npm's backfill lifting the prior window's first-read 6,465 to 7,322.
- code-review: roborev - volatile re-verification: the push and main's latest commit both moved to November 22 (a dependency bump), v0.75.0 still the latest release.
- code-review: code-review-feature-matrix - corrected the OpenCodeReview maturity cell (still read 151 releases and v1.13.13 after the 2026-11-21 bullet recorded 152 with v1.14.0, a missed write caught by the cell-content alignment check) and refreshed the OpenQodex npm figure to 6,874 trailing-month downloads.
- code-review: star-history chart markup audited across the ten repo-subject notes, only the dark source carries theme=dark everywhere, no reintroductions to remove.
- code-review: the kodustech/awesome-ai-code-review staleness re-check and the dedicated-list search both re-confirmed the q21 evidence (see Digest lines); no entrant candidates surfaced to this category.
- code-review: coderabbit, cubic, gito, graphite-diamond, multi, openreview, qodo, sourcery - not flagged by a signal and outside this run's rotation; no re-fetch (last verified 2026-10-24 through 2026-11-21 per the work list).
- entroly: release 1.0.91 (2026-11-21, PyPI and npm synchronized across 108 releases) shipping the recorded unreleased batches plus the corpus binding, audit-record sealing, and small-sample trial statistics; main's unreleased 2026-11-21 to 11-22 dedup and cache-cost round recorded.
- headroom: release v0.45.0 (2026-11-22, PyPI at 0.45.0 across 208 releases) shipping the recorded unreleased batches (security sweep through the 11-19/11-20 round) plus the opt-in tray icon, the extension-replaceable shaper steering level (default L3), and the repeatable OpenCode --extra-model flag; main's unreleased 2026-11-22 CI-security round recorded.
- graphify: releases v0.9.114 and v0.9.115 (2026-11-20 and 11-21) recorded, the extract/build/update performance release with extractor fixes and then the GHSA-g6c9-j4vm-mwq4 forced-SMB-authentication fix on Windows; PyPI tracks 0.9.115.
- repowise: main's unreleased 2026-11-21 to 11-22 batch recorded (code-origin ranking below shipped code, repo="." reading, scoped get_answer containment, path-query alternatives, keyless setup and Ollama docs, hook commit-burst coalescing); PyPI unchanged at 0.61.0.
- qmd: main's post-release 2026-11-19 to 11-20 fix recorded (pooled embedding input kept in a single evaluation batch); v2.9.0 remains the latest release.
- code-review-graph: full rotation re-verification, every cited URL re-fetched intact, v2.4.0 still latest, no release since 2026-11-11 and main newest-committed 2026-11-12; no claim changed.
- augment-code: full rotation re-verification, every cited URL re-fetched intact, plan structure and no-free-tier unchanged, the October 8 acquisition announcement still the newest post, intent-hq pushed 2026-11-22; no claim changed.
- jevgrep: full rotation re-verification, every cited URL re-fetched intact, v0.8.5 still latest and no push since 2026-11-12; trailing-month npm downloads refreshed to 14,409.
- graft: 2026-11-22 pushed signal resolved as non-default-branch activity with main newest-committed 2026-11-19, note untouched.
- sol-pi: 2026-11-22 pushed signal resolved as side-branch activity with main newest-committed 2026-11-20 (the recorded v2 announcement), note untouched.
- context-engines: star-history picture blocks audited with only the dark source carrying theme=dark, no reintroductions; matrix columns and the category index match the 21-member membership; internal links and front matter verified.
- agent-barn: counters refreshed (21 open issues and pull requests), no release past v0.23.0, the 2026-11-21 push a side branch.
- agent-control: release train moved to v8.17.0 (2026-11-20, evaluator runtime-URL revert) with the npm package moving 3.5.0 to 3.6.0 on the ts-sdk-v3.6.0 tag.
- agentfield: candidate line holding at v0.1.147-rc.1 with PyPI verified in step at 0.1.146, counters refreshed (39 open issues, default branch on a docs commit 2026-11-21).
- arifos: self-measurement stamp moved to 2026-11-22 and regressed to 11 of 13 floors measured pass, F3 and F6 failing, vault999 stale again, deployed build's drift aligned with the repo tree ahead (deploy pending), against the unchanged 2026-10-11 audit; PyPI trailing at 1!2026.11.18.
- cc-safety-net: npm holding at 2.8.1, counters refreshed (5 open issues, default branch on 2026-11-20 with the 2026-11-22 push a side branch).
- control-planes: jamf-ai-governance full rotation re-verification, no drift, Business Wire failing at the connection level with its recorded 403 state carrying.
- control-planes: mission-control full rotation re-verification, no drift, counters unchanged.
- control-planes: okto-pulse full rotation re-verification, PyPI still 0.4.2 with the v0.4.3 tag pending five days, product page still v0.4.2 with 17 gates and 280+ tools, oktolabs.ai 403 to plain curl with the alternate fetcher serving both product pages in full.
- control-planes: settlebridge full rotation re-verification, pricing and the unresolved license metadata unchanged, HN still zero threads, 325 commits on main with the Cyber Lab verifier agents hidden behind a flag.
- hol-guard: counters refreshed (77 open issues, pushed 2026-11-22), train holding v3.60.0, coverage-contract page still Cloudflare-blocked at 44 days past expiry.
- ibm-contextforge: unreleased default-branch work moved to a security-hardening commit (2026-11-20), counters refreshed (988 open issues and pull requests).
- microsoft-agent-governance-toolkit: default-branch push corrected to 2026-11-18 (a rustls bump) with the 2026-11-20 and 11-21 repo-level pushes side branches, counters refreshed (175 open issues).
- openappa: release train moved through v0.45.0 and v0.46.0 (2026-11-20, the latter breaking with a RemedyRefusal ApprovalPending variant) to v0.46.1 (2026-11-21), the v0.47.0 tag landed with its release page pending.
- paperclip: backlog regrew to 6,996 with the push at 2026-11-22, release pause at day five of v2026.1117.0.
- speakeasy-gram: release train moved to server@2.39.0 (access requests stored and decided, Slack DM approval cards, temporary grants) and server@2.40.0 with dashboard@0.154.0 and the first tunnel@0.3.0 (2026-11-22), counters refreshed (287 open issues).
- tracekit: counters refreshed (107 open issues, default branch pushed 2026-11-22 on an unreleased reviewer fix persisting answered approval ids), PyPI holding 1.7.2.
- vestige: release train moved to v6.2.3 (2026-11-20, published without release notes), default branch on a README restructure, counters at zero open issues and pull requests.

## (daily refresh, 2)

- tau2-bench: corrected the star-history chart markup (light source and img fallback carried theme=dark again).
- harbor: corrected the star-history chart markup (light source and img fallback carried theme=dark again).
- crit: refreshed the activity line (default branch moved 2026-11-21 with an eslint dev bump, PR 1216); v0.24.0 still newest.
- hunk: refreshed the activity line (default branch moved 2026-11-21 with the copy-selected-hunk command, unchanged-lines rendering, a pager fix, and four extension-directory listings); v0.25.0 still newest.
- langfuse: refreshed the activity line (default branch moved 2026-11-21 and 11-22 with the billable-units ledger and org endpoint and continued-topic naming); v4.83.0 still newest.
- phoenix: recorded the arize-phoenix 22.0.1 release (2026-11-21, cancelled loader batches no longer freeze the server).
- plannotator: refreshed the activity line (default branch moved 2026-11-22 with inbox linked-thread search, reply origin, and math-label rendering fixes; 0.29.2 version-bumped, release unpublished).
- harnesstax: no change kept; the worker's proposed Changes repair was dropped at review, the 2026-11-20 bullet already recording the 2026-11-19 republication.
- deepeval: rotation re-verification, every cited URL re-fetched live; activity refreshed to 2026-11-22; Python 4.3.4 still newest on PyPI; trailing-month downloads refreshed to about 3.3 million.
- kiln: rotation re-verification, every cited URL re-fetched live; default branch quiet since 2026-11-18; v1.2.1 still newest on GitHub and PyPI.
- kitaru: rotation re-verification, every cited URL re-fetched live; default branch at its 2026-11-19 feedback-command merge; trailing-month downloads refreshed to roughly 21,000; skills-repo push date moved to 2026-11-19.
- workshop: rotation re-verification, every cited URL re-fetched live; everything unchanged (v0.1.24 newest, repository pushed 2026-11-05).
- evaluation-review-feature-matrix: Phoenix maturity cell refreshed to v22.0.1 (2026-11-21).
- executions: aeon - rotation re-verified with every cited URL re-fetched; scheduler tick corrected to the shipped 30-minute off-peak default (scheduler.yml contradicts the docs' own five-minute file-table line), skill catalog 86 to 93, Discord and Slack inbound recorded beside Telegram, counters refreshed (310 forks, 1,578 commits, pushed 11-21), proof-of-work page unchanged
- executions: rivet-agents - rotation re-verified with every cited URL re-fetched; v0.5.7 stable recorded (11-20), pricing page unchanged since the 11-18 rebuild, footprint re-confirmed, counters refreshed (143 forks, 102 open, pushed 11-20 UTC)
- executions: airut - 11-22 push was a side branch (default-branch head still the 11-19 dependency bump), open issues 1 to 2, v0.28.0 still the newest release
- executions: centaur - a third calendar cut v2026.11.20 (11-20) also failed to hold the top, centaur-0.1.224 (11-21) on top above 0.1.223, counters refreshed (297 forks, 121 open)
- executions: claude-code-action - release train moved to v1.0.286 (11-20), counters refreshed (2,289 forks, 972 open), Latest badge still on the v1.0 announcement release
- executions: claude-code-hooks - re-verified out of rotation after v2.1.330 (11-20), the 33-event surface and both version gates hold on the live docs, open issues roughly 16.0k
- executions: github-agentic-workflows - counters refreshed (639 forks, roughly 20,500 commits), train unchanged (v1.3.1 prerelease top, v1.2.2 newest stable), no advisory since 11-11
- executions: n8n - counters refreshed (61,614 forks), stable train unchanged (2.48.7 newest stable, 2.49.3 prerelease on the train)
- executions: switch - counters refreshed (368 forks, open issues 718 to 569), stable release line unchanged
- executions: trigger-dev - v4.8.8 (11-20) recorded as current, counters refreshed (1,537 forks)
- executions: matrix Aeon tick cells corrected to the 30-minute default, Render added to the Rivet Agents execution cell, no membership change
- ante: recorded the v0.3.8 release (published November 22, notes dated November 21), untitled-session naming, OpenAI Responses prompt caching with reasoning handed back, Gemini tool-call ids, the top_p/top_k/MODEL_TOP_P removal, restore points across provider switches, offline-model refresh dropping sm_86, and cache-write token pricing.
- atomic-agent: refreshed the push date to 2026-11-22.
- avante-nvim: refreshed the default-branch push date to 2026-11-21.
- auto-dev: rotation re-verification (Xiuper 717, Sketch 4,661 downloads, VS Code unchanged) and repaired the Changes ordering that had 11-17 above 11-13.
- codex: recorded the v0.173.0 alpha train (four alphas, November 20-21) behind the unchanged rust-v0.171.0 stable.
- copilot-cli: recorded the prerelease train reaching 1.0.115-6 (published November 21) behind the unchanged 1.0.114 stable.
- deepseek-harness: recorded dsh-v0.2.3-alpha.1 (published November 22) opening the 0.2.3 line: Git plugin sources, the Timeline plugin, the PR review page with GitLab merge, the session status area, remote Web sign-in, and the connector plugins.
- exo: rotation re-verification, all references live, still no tagged releases, default-branch commit still 2026-11-18, FrontierHarness figures unchanged.
- fx: refreshed pushes through 2026-11-22, v0.0.17 still the newest release.
- gptme: rotation re-verification, push date 2026-11-22, PyPI last-month 21,498 (down from 23,018), contributors TimeToBuildBob 2,969 against the author's 1,977, order unchanged.
- jcode: recorded the v1.0.3 release (published November 22, no v1.0.2 on the feed): Antigravity Google login with Gemini 4 (Neon) default and pooled quota bars, scripted API-key provider login, and Jarvis live-caption voice.
- junie: recorded the installer train continuing through 4193.1 (published November 21), again without release notes attached.
- kimi-code: repaired the Changes ordering that had an 11-19 bullet after the 11-20 one.
- mimo-code: refreshed the push date to 2026-11-21.
- mini-swe-agent: rotation re-verification, the default-branch quiet-since-2026-09-03 claim re-grounded on the commits API, v2.4.6 still newest, PyPI refreshed to 14,060,333.
- nanocoder: refreshed the push date to 2026-11-22, v1.32.0 still the newest release.
- oh-my-pi: recorded npm reaching 19.1.0 (published November 21): standby tools, provider hooks, opt-in prompt perturbation, the Codex underscore-header fix, and LiteLLM key rotation; push date 2026-11-22.
- om-o: recorded the v5.1.81 release (published November 21): frontend-skill routing with the visual-qa finish gate, memory-sync trust-pin fixes, and the omo sync status readout; push date 2026-11-22.
- omnigent: refreshed the default-branch push date to 2026-11-22, v0.23.0 still the newest release.
- openinterpreter: rotation re-verification, rust-v0.0.61 still the newest release, pushes through 2026-11-22.
- zcode: rotation re-verification, v3.18.2 still the newest desktop release, GLM Coding Plan prices re-read unchanged, disclosure threads unchanged.
- harnesses: matrix verified against its 45 columns and the touched columns' cells against their notes; category index list verified complete and sorted, with the DeepSeek Harness and Mini-SWE-agent one-liners refreshed.
- anthropic-structured-outputs: rotation re-verification, all five cited URLs re-fetched and live, the supported-model list still nineteen with fable-5-2 and mythos-5-2 the newest additions, the platform split, token table, and schema subset unchanged
- anyjev: out-of-rotation board re-fetch at the drift signal, the Decision Index's 2026-11-21 regeneration (346 configurations) crowd-shifted the family's unchanged scores again (lm6q thirty-sixth, Tacitus forty-eighth, L0 sixty-fifth)
- baml: volatile re-verification, a fresh 0.26.1 nightly line opened at the 20261120.a build (published 2026-11-21), the stable train holding at 0.26.0 and the wrapper track at 0.4.0; forks 524, npm trailing month 1,681,907
- clm: out-of-rotation board re-fetch at the drift signal, the unchanged 6.49 moved from rank 319 to 338 of 346
- cua-s1: volatile re-verification, the cua-s1 component untouched since its two 2026-11-20 dependency-security commits, the host pushed to 2026-11-22 with the 11-21 to 11-22 cua-spaces, cua-spacesd, cua-driver-rs, cua-sdk, and bench releases
- decider: out-of-rotation board re-fetch at the drift signal, the family's unchanged scores crowd-shifted again (31B fifty-first, pplx v1.2 thirtieth, the v1.3 rows thirty-fourth and 120th, 2B 280th of 346)
- decision-index: out-of-rotation board re-fetch at the drift signal, the 2026-11-21 regeneration (346 configurations, generated 14:35 UTC) kept the top two and every named score while a new kudhab took third at 73.04 (REAX 1.3 fourth), Logits Sigma-1 Duo, KnowLine Gen2, and the Strands Decider v2 31B filled the top ten, the vendor's rows crowd-shifted to seventy-ninth and eighty-ninth, Valen's 27B to fifty-fifth, Kev 27B to eighty-seventh
- h2o-lightning-4b: out-of-rotation board re-fetch at the drift signal, the three sibling rows' unchanged scores crowd-shifted (31B v1.3.0 nineteenth, Bolt 74B ninety-eighth, E4B 161st of 346), the 4B still absent
- hybrid-execution: the matrix's Decision Index and Strands cells and the category index's board one-liners re-based to the 2026-11-21 regeneration, JevBench cells held at 261 ranked systems on the re-fetch, and pre-existing Changes ordering defects repaired in jev, kev, and nimble
- jeff: out-of-rotation board re-fetch at the drift signal, the unattributed jeff-uncapped row's unchanged 6.21 crowd-shifted to rank 339 of 346
- jev: out-of-rotation board re-fetch at the drift signal, the Decision Index's 2026-11-21 regeneration crowd-shifted the vendor's rows to seventy-ninth (Jev 1.14, 60.24 unchanged) and eighty-ninth (1.13, 59.22 unchanged)
- kev: volatile re-verification and board re-fetch, the 2026-11-20 kev-deploy/kev-finetune KEV_REF pin (#292) landed, pushed date moved to 2026-11-21, and the 27B's unchanged 59.56 crowd-shifted from seventy-eighth to eighty-seventh of 346
- laya: volatile re-verification, 0.5.2 released 2026-11-21T19:24 UTC (laya.Escalator and laya.vllm_teacher, a vLLM-served teacher for what the Learner is unsure of, #1518), PyPI at 0.5.2
- nimble: out-of-rotation board re-fetch at the drift signal, the family's unchanged scores crowd-shifted again (v4.1 Max fifteenth to eighteenth, v3 132nd to 145th, v2 178th to 195th of 346)
- ollaya: volatile re-verification, one 2026-11-21 commit adding an OLLAYA_READ_ONLY mode refusing model-store changes over HTTP with 403 (#178), no release past v0.18.0
- openai-structured-outputs: rotation re-verification, all five cited URLs re-fetched and live, the guide still leading with gpt-6-astra, the gpt-6-luna pricing example unchanged at $0.10/$0.50 standard with Batch at half price
- outlines: rotation re-verification, all seven cited URLs re-fetched and live, the repository unchanged (pushed 2026-09-21), the backends feature doc still naming outlines-core, llguidance, and xgrammar, and the vLLM docs unchanged
- strands-decider-2b: volatile re-verification and board re-fetch, v0.4.0 released 2026-11-20 (PyPI 0.4.0 the same day, a breaking Python 3.11 requirement) pushed 2026-11-21, and the Decision Index's 2026-11-21 regeneration dropped the v1 2B row (20.72, rank 271) and entered five Strands Decider v2 rows (the 31B seventh at 72.12, above the vendor's Jev 1.14), the attribution resting on engine naming and the org's first v2 checkpoint (E2B, created 2026-11-17) with the v2 weights mostly unpublished and no v2 results section
- tool-eval-bench: volatile re-verification, two 2026-11-21 commits (a Flare (Hyperflare) backend joining the recognized set #348 and a web model-picker fix #347), no release past v3.1.0
- valen: rotation re-verification, every cited URL re-fetched and live, the repository unchanged (pushed 2026-11-15, v0.4.2 current), and the board's 2026-11-21 regeneration crowd-shifted the five unchanged-score 1.5 rows to fifty-fifth through 264th of 346 with the 27B still above Jev 1.14
- von: out-of-rotation board re-fetch at the drift signal, the unchanged 11.65 moved to rank 327 of 346
- winnow: out-of-rotation board re-fetch at the drift signal, the unchanged 46.57 and 32.72 crowd-shifted to 169th and 239th of 346
- memory: deja - recorded v0.25.0: the Ante, Doubao Desktop, and TeleAgent readers bring the wired count from 49 to 52, a composite GitHub Action installs deja, cold recall dropped to 658 ms and synced manifests to 2.3 MB, hooks gained edit radar and subagent briefs, and the Raycast extension reached the store; the count moved through the note and the matrix cells, and the two 2026-11-20 lags (latest-release line, push date) were repaired.
- memory: okf-agent-memory - recorded v0.9.0: full OKF v0.3 conformance with breaking trust-tier semantics and file-relative resolution, programmatic write protection, time-range change queries, the stale_after gap closed, and optional git-freshness signals; the version moved through the note, matrix cells, and index one-liner, with the upstream spec and README badge still reading v0.2.
- memory: sno-station - recorded v2.3.0: memory scoped by repository unless said to apply everywhere, unscopeable statements staying in their project, and faster cross-project correction propagation.
- memory: cognee - repricing recorded: Standard $0 per month with 100M included tokens then $0.10 per 1M processed (from $1.00), Premium introduced at $99 per month plus usage, Enterprise unchanged as BYOC; the Price history gained the row and the matrix pricing cell and choosing bullet moved.
- memory: zep - rotation re-verification: pricing ladder unchanged on the fetched text, Agent Skills recorded free on every plan until January 4, 2027.
- memory: belief-context-graph - rotation re-verification: no facts changed.
- memory: emulo - rotation re-verification: no facts changed.
- memory: engrim - rotation re-verification: no facts changed.
- memory: claude-mem - recorded v16.0.2 and v16.0.3: the auto-update cache sweep no longer deletes the plugin version under a live session and the OMP hook now talks to the server runtime; npm in sync at 16.0.3.
- memory: gbrain - recorded the train through the v0.60.753.0 and v0.60.759.0 stable cuts to v0.60.761.0: chat-day pages as .gbrain-owned files, remembered facts in CLI search, the parked-image scan fix, the two-minute Git-history bound, and skills surfacing model refusals.
- memory: operator-memory - recorded the November 21 coordinated batch: pi@v1.3.0 (Project Init streamlined, standalone index command retired), opencode-v2@v2.4.0, helper@v1.13.0, codex adapter at 0.5.0.
- memory: memory - volatile pass over the other ten flagged members (basic-memory, funes, hindsight, memmachine, memos, memu, openviking, plur, screenpipe, supermemory) found routine fixes only, and Basic Memory's and Sno Station's pricing pages re-verified unchanged; no note changes.

## (daily refresh, 3)

- bifrost: pushed 2026-11-22 on a Bedrock inline-document-sources fix and a Chat Completions tool-result image fix, no release past helm-chart-v2.1.49.
- litellm: v1.112.0-dev.2 took the train head and v1.108.15 patched the v1.108 line on 2026-11-21 (keyless JSON-configured deployments now 401 without an api_key, ResetData provider added, Groq and Meta transcription $0-spend fixed); PyPI holds 1.110.2.
- llm-gateway: default branch added the Lyria 3.5 music model, Qwen3.8 2.4T A95B on Fireworks, a T3 Code guide, and a content-filter 4xx fix, no release past v1.26.0, no plan-price or fee change.
- localai: default branch added an Ollama search add-on, an Ollama importer fix, a cluster and access administration UI redesign, and engine bumps, v5.0.1 still latest.
- ollama: re-checked with nothing past v0.40.17 or the 2026-11-21 push, no change.
- experiential: re-checked with nothing past v0.7.261 or the 2026-11-21 push, no change.
- llamafile: rotation re-verification with every cited URL live, nothing moved (0.10.6 still latest, default branch at the 2026-11-05 decision-models commit).
- vercel-ai-gateway: rotation re-verification recorded two pricing-page surcharges new to the record (Regional Hosting beta at 5% of inference cost, BYOK-without-fallback at $0.20 per 1,000 requests) plus provider-subscription API credits riding BYOK, price-history row appended, all recorded figures held.
- chatgpt-plans: rotation re-verification recorded GPT-6.2 Astra and GPT-6.2 Sol joining the rate table, message ranges, and feature matrix with USD billing figures and Daybreak Blue/Red and GPT-Rosalind-Research rate rows, the GPT-5.5 retirement notice leaving the page, and the consumer pricing page answering 200 a third time with figures still client-side.
- opencode-go: rotation re-verification recorded catalog churn (Qwen4 Max, MiMo-V2.6-Pro, LongCat 2.5 Free, Hy4 flash Free in, Crouton Alpha Free and the MiMo-V2.5 rows out), limit moves (Gemini 4 Argon $30, M3.1 Flash $20, Qwen3.8 Flash $30, MiMo-V2.6-Flash $60, Muse Spark rows $60), the GLM-5.3 promotional $500 Go limit through November 22, and Go Teams's two-member minimum, plan prices unchanged.
- model-access: matrix ChatGPT plans models cell extended for the GPT-6.2 pair, every other volatile cell re-checked unchanged against the notes' 2026-11-22 fetches.
- model-access: star-history chart markup verified clean across all nine repo-subject notes (only the dark source carries theme=dark); links, front matter, tables, and style checks clean across all 26 notes.
- archon: README rebrand recorded out of rotation on the people-and-publications handoff, factory-builder framing replacing the retired first-open-source-harness-builder line with the 363-of-524 claim kept.
- agent-swarm: v2.14.3 (2026-11-22) recorded, npm matching.
- oh-my-codex: v0.22.12 through v0.22.15 (2026-11-21/22) recorded.
- openharness: v0.6.80_cli (2026-11-22, dial eval) recorded, desktop line unmoved.
- openrig: v0.7.21 (2026-11-21, team-fixes patch) recorded.
- orca: v1.4.269 (2026-11-21, bots on eight more agents) recorded.
- pandacode: v0.3.136 (2026-11-22, ZCode 3.18.2 restored) recorded.
- paseo: v0.17.1 stable (2026-11-21, worktree workspaces plus Antigravity approvals) recorded.
- atlas: alpha-0.5.1 (2026-11-21, chores-only) recorded.
- agentbox: npm 0.33.1 and tray 0.4.4 (2026-11-21) recorded, default branch reactivated by the 0.33.1 merge.
- agentlas-os: desktop 2.0 push recorded (Agentlas Desktop 2.0.11, Android app, affiliate banner caution), engine line unmoved at os-v1.2.99.
- worktrunk: v0.83.0 (2026-11-21, PyPI wheels and config schemas) recorded with rotation re-verification.
- flowise: rotation re-verification, npm 28,814 in the month ending 2026-11-21.
- crystal: rotation re-verification, no drift.
- agent-manager: rotation re-verification, v0.49.0 stands.
- evoagentx: rotation re-verification, no drift, front-matter updated-date lag corrected.
- gastown: rotation re-verification, death record stands.
- yao: rotation re-verification, v1.0.0 stands.
- agtx: rotation re-verification, Trendshift 403 wall stands.
- agentgrid: rotation re-verification, v2.9.64 stands.
- agentsmesh: rotation re-verification, removal record stands.
- crewplane: rotation re-verification, v0.4.3 stands.
- gwq: rotation re-verification, no drift.
- hcom: rotation re-verification, v0.7.31 stands.
- orchestration-feature-matrix: 10 status cells moved to newest releases and trains, Worktrunk's cell carrying v0.83.0.
- addy-osmani: rotation re-verification, every cited URL re-fetched and live, no claim moved
- cole-medin: rotation re-verification, channel surface renders under browser UA (about 233,000 subscribers, 343 videos as of 2026-11-22), Archon README rebranded to "factory builder" YAML framing and dropped the October "harness builder" billing, 363-of-524 claim kept
- dax-raad: rotation re-verification, every cited URL re-fetched and live, followers about 5.8k as of 2026-11-22
- dex-horthy: rotation re-verification, AI That Works #80 aired 2026-11-10 (moved off the 11-03 slot), aired count 80, #79 lists as SlopCodeBench: The New Metrics, pricing unchanged
- geoffrey-huntley: rotation re-verification, the October pause ended with "the answer to ai is engineering" (2026-11-11), added to Status and References
- people-and-publications: matrix cadence cells moved for geoffrey-huntley (one essay 2026-11-11) and cole-medin (343 videos as of 2026-11-22); no membership change, columns stay at thirty-four
- protocols: ag-ui - train moved to release/2026-11-21: @ag-ui/langgraph 0.0.47 breaking (durability agent option, forwardedProps allowlist, thread-history-constrained continues), Python adapters moved (ag_ui_adk 0.14.0 closing the GHSA-jhvp-j8pf-5vh5 endpoint auth bypass, ag-ui-langgraph 0.0.48, ag-ui-openai-agents 0.2.0), first-party TS packages stay 1.0.5, downloads refreshed to about 34.0M for the month ending 2026-11-22.
- protocols: acp - trains moved 2026-11-21 afternoon: stable train to v1.21.2 with schema-v1.34.1 (IndexMap re-export and patch fixes), v2 beta train to schema-v2.0.0-beta.3, SDKs tracking (Rust 5.5.1, TypeScript 2.9.1), agents list re-counted at 51.
- protocols: llms-txt - rotation re-verification clean with no dead sources, Anthropic's served file now 1,014 lines, Status and cautions gained the first-party skeptic record (Backpack for Laravel 2026-10-20, its HN thread at 31 points and 21 comments 2026-11-20).
- protocols: pic-standard - rotation re-verification clean, spec train stands at v0.9.1, pypistats trailing month re-read 492 as of 2026-11-22 (rolling window), forks and open issues refreshed (47, 19), default branch quiet since the 2026-11-12 MCP-guard fix.
- protocols: protocols - matrix AG-UI prose figure to 34.0M (month ending 2026-11-22) and the PIC PyPI cell to 492 (as of 2026-11-22); A2A, A2UI, MCP, Poppy, and WebMCP pushed dates resolved to side-branch pushes with default branches quiet, UHP's v0.41.24 train already recorded; no membership change.
- retrieval: code-graph-rag PyPI train jumped to 0.7.79 (2026-11-21, 43 releases, the 0.7 line's catch-up publish) and pushed 2026-11-22 recorded
- retrieval: docling pushed 2026-11-22 with 1,071 open issues, trailing-month downloads re-read at 4,323,066, PyPI 2.151.1 still latest
- retrieval: knowhere releases v1.2.65 and v1.2.66 (2026-11-21 and 2026-11-22; publish-after-index-ready with refunds, then Word tracked-changes acceptance, LaTeX equations, embedded-object rendering) recorded
- retrieval: lightrag pushed 2026-11-22 with 245 open issues and PyPI 1.5.8rc2 (2026-11-21) recorded, 1.5.7 still latest stable
- retrieval: open-codebase-index pushed 2026-11-22 (PR #503), trailing-month installs re-read at 5,489 roughly flat, npm 0.40.1 still latest
- retrieval: ragflow pushed 2026-11-22 with 1,314 open issues, v1.0.0 still latest and Docker pulls unchanged at 4.1M
- retrieval: semantic-code-search rotation re-verification corrected the VS Code description to the external-ingest wording (page dated 2026-11-18) and rewrote the two Strengths bullets still describing the dropped encrypted-Cursor design
- retrieval: tree-sitter-chunking rotation re-verification found all claims unchanged
- retrieval: matrix maintenance row refreshed to 2026-11-22, no membership change

## (daily refresh, 4)

- sandboxing: landrun - recorded v0.2.0 (2026-11-21, breaking UDP restrictions on ABI v10 kernels, port-range flags, port-validation fix for v0.1.17's silent wrapping, attested static binaries), note, matrix cells, and index one-liner updated in the same run
- sandboxing: fence - volatile: 2026-11-17 push still branch activity, quiet-on-main watch holds
- sandboxing: brood-box - volatile: 2026-11-21 push branch activity, main's newest commit stays 2026-10-21
- sandboxing: microsandbox - volatile: 0.8 train advanced on main (cloud restart route), v0.7.16 still newest
- sandboxing: nemoclaw - volatile: fix train through 2026-11-21 evening (Ollama proxy health), pushed 2026-11-22, one-prerelease story holds
- sandboxing: nvx - volatile: prerelease train advanced to 2026-11-22
- sandboxing: opensandbox - volatile: post-rc.2 main activity is repo tooling, release-1.3.0-rc.2 still newest
- sandboxing: sandstorm - volatile: fix train through 2026-11-20 (detached archives, MCP progress markers, console refreshes)
- sandboxing: clawk - rotation re-verified: quiet window 101 days, stalled reading holds
- sandboxing: container-use - rotation re-verified: docs and 18-client coverage unchanged, 15 months without a release
- sandboxing: greywall - rotation re-verified: quiet window 101 days, forks to 40
- sandboxing: aigate - rotation re-verified: resumed-train reading holds
- sandboxing: mapped source arjan/awesome-agent-sandboxes unchanged since its last check (newest commit 2026-11-17), no new entrants
- sandboxing: star-history markup clean across all 25 notes, only the dark source carries theme=dark
- session-analytics: agent-analytics - rotation re-verification: the MIT LICENSE file landed on the default branch 2026-11-20 (PR 6), closing the missing-license caution carried since creation, and the npm CLI moved 0.5.42 to 0.5.43.
- session-analytics: codexmonitor - rotation re-verification: dormancy unchanged (last commit 2026-03-26, releases still ending v0.7.67), forks 421, open issues 94.
- session-analytics: clawtrace - rotation re-verification: all cited URLs, repository state, npm plugin (0.1.23), and credit rates unchanged.
- session-analytics: ctx - recorded v2.5.0 (2026-11-21): the ctx sessions listing command, ATIF v1.8 trajectory import and export, Antigravity App 2.x and T3 Code ACP session roots, and memory-bounded indexing with security dependency upgrades.
- session-analytics: memex - recorded v0.34.1 (2026-11-21, released without notes): resume from deleted workspaces and TUI filters fixed.
- session-analytics: token-monitor - recorded v0.95.1 (2026-11-21): settings data hints, Devin quota emails, Windows regional fonts, and Qoder and Z.ai Team quotas kept separate across synced devices; named rows re-verified 55 plus the aggregate.
- session-analytics: claude-code-agent-monitor - recorded v2.5.0 (2026-11-21), shipping the pending session-folder launcher and Windows test-gate fix.
- session-analytics: novafabric - default branch woke 2026-11-20 with a SOC 2 and ISO 42001 compliance wave (unreleased) and the tracker moved to 78 open issues and 17 open pull requests, mostly a planned Console 2.0 roadmap.
- session-analytics: matrix Tokscale Kind cell moved to 64 named stores (missed by the 2026-11-15 count pass) and the Agent Analytics Open source cell moved to checked after the 11-20 LICENSE.
- skills: agent-native - volatile refresh: the 11-21 pack push (#128) a turn-into-app content pass across its skills/ and plugins/ copies, catalog 26 held, npm 0.4.10 held with pinned-window downloads 7,560 (2026-11-16 to 11-22), framework pushes through 11-22 monorepo app development, HN zero-hit search re-run
- skills: gepa - volatile refresh: main unchanged since 2026-11-09 with the 2026-11-21 pushed_at on side-branch activity (commits API), open issues 166 to 177, PyPI 0.1.4 held
- skills: skillopt - volatile refresh: the 11-17 and 11-21 batches moved main to 429 commits past v0.2.0, open issues and PRs 93 to 80, v0.2.0 still the newest release
- skills: skillspector - volatile refresh: main's 11-19 batch grew past the recorded parser-boundary PRs (pyproject.toml dependency-table scanning for Poetry/PDM/Hatch/uv, tool-poisoning ATLAS tagging, input-origin retention, AST line-break indexing, dependency bumps), pushed_at 11-20 on side branches, open issues 163 to 156, v2.13.6 held
- skills: evoskill - rotation re-verification: every cited URL live, all claims held a fifth time (v2.0.0 still newest, main still 2026-07-06, v2 branch head still 2026-10-20, HN still zero-hit)
- skills: headcount - rotation re-verification: death state re-confirmed a third time (repo, account, raw README, org chart all 404), Zentor review and Enterprise DNA writeup live, HN search unrelated-only
- skills: skills-feature-matrix - Agent-Native no-coverage date re-run to 2026-11-22, no cells moved
- fluent: rotation re-verification clean, every cited URL re-fetched, no material change
- fusion: rotation re-verification recorded the README's Kimi K3 and custom-provider additions, experimental runtime-plugin marking, and shipping-weekly badge, the stable channel's lag at eleven weeks, and downloads at 1,865 for the month ending 2026-11-21
- machinist: volatile re-verification recorded the four-harness roster, headless runs, the default-on auto-merge with opt-out, Linear and run-API intake, and machinist.sh catching up, at 394 commits past v0.4.0 with v0.4.0 still the newest release
- open-inspect: volatile re-verification clean (pushed 2026-11-22, no release train, zero coverage holding); chart dark source fixed
- open-swe: recorded Robin Desktop v0.3.20 (2026-11-21) as the newest stable desktop tag
- ouroboros: recorded the v0.65.0-v0.65.2 run (2026-11-21 to 2026-11-22, GitHub and PyPI, the browser play-and-fix line)
- zeroshot: recorded the train to v17.1.0 (2026-11-22, npm in sync): v17.0.0's peer-compatibility break, canonical-identifier advance, and saved profile imports, v17.1.0's hosted checkpoint opt-out and Steps view
- loki-mode: recorded the train to v11.176.0 (2026-11-21, npm in sync): the switch-off-until-reviewed label rules, the SYSEDIT preservation, and the already-done veto; Docker pulls refreshed to 339,759
- loopforge: recorded the 2026-11-21 CodeBuddy optionalization (all pickers accept other CLIs' models, the platform runs without CodeBuddy) and the bundled-CLI correction (CodeBuddy, Codex, DeepSeek Harness)
- software-factory: index Ouroboros one-liner corrected to 15 runtimes
- software-factory: matrix moved three Machinist cells and one LoopForge cell on the volatile re-verifications
- ai-dlc: volatile refresh (pushed 2026-11-22, 356 open issues, nightly-preview fixes on the default branch, v2.11.4 still the stable release).
- bmad-method: volatile refresh (111 open issues and PRs, side-branch pushes through 2026-11-22, main at 2026-11-12) and the README's mid-November 2026 v7 ETA lapsed without a v7 release.
- gsd: volatile refresh (pushed 2026-11-22, v1.22.0 still newest, npm 93,554 successor and 99,585 legacy last month).
- openspec: volatile refresh (main still at the v2.1.1 release commit with side-branch pushes through 2026-11-22, npm 5,606,970 last month).
- spec-kitty: volatile refresh (pushed 2026-11-22, v4.0.0rc8 still the newest prerelease, PyPI 2,540 last month).
- cc-sdd: rotation re-verification of every cited source (all alive); recorded the Windsurf / Cascade integration deprecated for the Devin adapter, now scoped to Devin Local / CLI with Devin Cloud outside it, npm 75,678 last month.
- intent-router: rotation re-verification of every cited source (all alive, the delivery report intact, footprint still empty); recorded the count climbing back to 108 stars and 21 forks on 2026-11-22, a three-week pace unlike the one-day refills.
- mcp-probe-kit: rotation re-verification of every cited source (all grounded, the bytezonex 403 wall re-confirmed, footprint still zero); npm moved to 987 last month.
- spec-driven-development: link, front matter, chart-markup, and matrix parity checks clean across all 18 notes; no membership change, matrix and index page current.

## (daily refresh, 5)

- bb: volatile re-verification on the pushed signal, release train unchanged at desktop-v0.50.1 (November 18) with post-release feature commits landing November 21-22 (per-session agent-writable folders granting each thread only its own storage, Android text-selection fix, split-pane focus).
- cate: volatile re-verification on the released signal, recorded the stable graduation v2.1.0 and v2.1.1 (both November 20): shared workspace runtimes surviving restarts, encrypted device pairing over Cate Connect, SSH and WSL remote machines, an iPhone companion, terminal-based agents including Pi, and focus-free browser automation.
- cc-haha: volatile re-verification on the pushed signal, release train unchanged at v0.7.9 (November 20) with feature commits landing November 21-22 (local Antigravity ACP sessions, Grok Official context-window alignment with per-model overrides).
- codenomad: volatile re-verification on the pushed signal, stable v1.1.0 unchanged with the v1.1.0-dev daily train building through November 22.
- continue: volatile re-verification on the pushed signal, metadata-only, pushed_at 2026-11-20 against a default branch quiet since 2026-07-21.
- delta: rotation re-verification, recorded Remote control that actually works (November 17, Personal Runners on your own machines beside the Zed-hosted cloud, `delta runner serve` for servers and VMs); pricing re-verified unchanged.
- nimbalyst: volatile re-verification on the pushed signal, recorded v0.84.3 (November 21, GPT-6.2 Sol as the Codex default, Electron 45, macOS 13 Ventura minimum), and repaired the Status chain that had dropped the v0.84.2 record its 2026-11-19 bullet logged.
- openchamber: volatile re-verification on the released signal, recorded v2.8.0 (November 22, Leads, OpenCode Server shared-service mode, Show notes, Run in terminal, per-project secrets, Android share-into, security tightening) as the new latest release.
- vibeyard: volatile re-verification on the released signal, recorded v0.3.11 (November 21, current-project session search scope, python3 hooks and status-line fixes, Claude Code hook session state with plan usage limits) as the new latest release, the train's first written feature notes since the pause; cautions and bottom line adjusted.
- whiteboard: volatile re-verification on the pushed signal, no desktop release after v0.5.2 with diffr releasing 0.1.24 (November 22), open-work queue at 11 issues and 29 pull requests; the new repository tagline keeps the README's agent-drawing pitch, no repositioning.
- windsurf: rotation re-verification, all cited URLs re-fetched live, and the devin.ai/pricing reading updated: headline tier prices unchanged (Free $0, Pro $20, Max $200, Team $80) but the Teams plan's $40/month per-full-dev-seat add-on no longer stated, the plan now reading $80/month with up to 100 users; Price history row appended.
- void: rotation re-verification, all cited URLs re-fetched live, archived flag, release freeze, product page, and both threads standing unchanged.
- zed: volatile re-verification on the pushed and released signals, weekly train added v1.29.3 (November 20), prices re-verified unchanged; added the MCP support fact (Tools and Prompts per the MCP docs) the feature matrix's MCP cell carried without a note trace.
- surface-feature-matrix: moved the Cate Mobile surface cell (iPhone companion, stable 2.1.0 of November 20) and extended the Delta Cloud agents cell (personal runners, November 17); cloud-agents reading extended.
- surfaces: no credible surfaces-dedicated source has appeared; the GitHub search re-run returned the known low-star IDE and client lists, none a category list.
- surfaces: OpenBitFun's Cowork-style general-workspace rejection holds on the fourteenth q#14 re-read (README re-fetched 2026-11-22, same four harness modes, phone remote control, Mini App marketplace, 4,104 stars, pushed 2026-11-21).
- task-management: agentrq - recorded v0.17.0 and v0.17.1 (both 2026-11-21: per-workspace agent concurrency removed so an agent runs one task at a time, a server cap on the agents one account runs on a machine, failed native tool calls keeping their error, a contributor licence agreement, and team-shared Pano docs readable by agents by id) and refreshed counts (2 open issues and pull requests, pushed 2026-11-22); v0.17.1 now the newest release
- task-management: backlog-md - recorded the four releases of 2026-11-21 through v1.57.2 (task filenames following title edits, decisions as first-class CLI records, the read-only static browser export, the folder-separator ID guard, and the rename-race ID locking) and refreshed counts (107 open issues and pull requests, 160,990 npm over the trailing month to 2026-11-22); v1.57.2 now the newest release
- task-management: beads - refreshed counts (1,951 forks, 1,788 open issues, pushed 2026-11-22, 50,983 npm over the trailing month to 2026-11-22); v1.3.4 still the newest stable release and rc.4 still the newest candidate
- task-management: its-a-plan - refreshed counts (332 forks, 88 open issues and pull requests, pushed 2026-11-22, 2,147 runner npm over the trailing month to 2026-11-22); v1.12.0 still the newest release
- task-management: kata - refreshed counts (135 open issues and pull requests, pushed 2026-11-22); v0.23.0 still the newest release
- task-management: ordewell - recorded v0.14.5 (2026-11-21: the Pi runner shipped over Pi's RPC mode, its writable mode running Pi's own tools without asking even under Guarded, plus briefs-first messages, model-call counting, known-gap remainders, earlier long-task flags, largest-slices-first scheduling, and the stalled-stream resend) and refreshed counts (23 open issues and pull requests, pushed 2026-11-22, about 9,200 npm); v0.14.5 now the newest release
- task-management: pad - refreshed counts (17 open issues and pull requests, pushed 2026-11-22); v0.21.1 still the newest release
- task-management: plandb - rotation re-verification with every cited URL re-fetched live: the stall holds (v0.2.1 still the newest tag, no commit since 2026-09-24, six open issues), the HN no-story claim re-confirmed to 2026-11-22, the stall duration refreshed to over six months, and the org's site restructured to front CodeAF, its coding agent product, beside the AgentField platform (agentfield pushed 2026-11-21, SWE-AF 2026-11-12)
- task-management: task-master - rotation re-verification with every cited URL re-fetched live: all claims confirmed, the repo still quiet since April 2026 with 0.43.1 still the newest release, the Commons Clause license and the product page's "Open source. MIT licensed. fork it." claims unchanged, Hamster's tiers unchanged, and npm read 62,912 downloads over the trailing month to 2026-11-22
- task-management: ai-dev-tasks - rotation re-verification with every cited URL re-fetched live: still dormant with the last push at 2025-11-05 and counts refreshed (1,698 forks)
- task-management: task-management-feature-matrix - refreshed the release and activity cells for the seven members with pushes or releases since 2026-11-21, and the Ordewell integration cell gained the Cursor (ACP) and Pi (RPC) runners
- task-management: the dedicated-list check found no credible task-management-specific list for the eighth consecutive run, with no incidental candidate surfacing in scope
- task-management: q24 evidence refreshed with no independent adoption or third-party coverage for TomzxCode/agentic; the rejection stands
- trackers-and-leaderboards: benchmark-radar - volatile re-verification: the 11-21/22 pushes are bot daily snapshots and the daily institution refresh only, badge to 54,628 records over 117 days as of 2026-11-22, v0.15.0 still newest.
- trackers-and-leaderboards: bestvaluemodel - volatile re-verification: the 11-22 push is a routine AA data refresh with no changelog entry after the 11-21 Argon doubling, cadence and no-license state unchanged.
- trackers-and-leaderboards: epoch-ai - rotation re-verification: team page lists 64 people (from 62) and moved to epoch.ai/about/team (301), data stamps through Nov. 22, GitHub to 41 repos (newest push Nov. 19), newest publications the Nov. 18 agentic-inference insight (4.6x in a year vs 2.6x compute), the Nov. 17 Anthropic-compute report, the Nov. 13 Fable-5.2/Astra report, and the Nov. 9 incident report on AI agents attempting exploits during a third-party alignment evaluation.
- trackers-and-leaderboards: lmarena - rotation re-verification: recorded the Interaction-as-a-Judge (11-13) and Agent Arena Index rebuild (11-18, topline index plus agent-mode Alignment Index on a thousandth-of-a-standard-deviation scale, rankings unchanged, Tool Hallucination signal retired) posts, and the methodology fleet's move from the lmarena org (now 0 public repos) to Arena-OSS (created 2026-08-20, 13 repos), arena-rank still unpushed on main since 2026-08-04.
- trackers-and-leaderboards: the per-refresh dedicated-list check re-ran 2026-11-22 with no change (git-git-hurrah 6 stars, metaory 2 stars, SAILResearch 383 stars as the pending proposal, no new credible list in search).
- trackers-and-leaderboards: matrix refreshed from the epoch-ai rotation (run-by cell to 64 listed as of 2026-11-22); membership unchanged at thirteen, columns and index list verified sorted and matched.
- section-root: model-selection-for-coding-tasks: the ChatGPT plans' rate table, Plus message ranges, and Work/Codex feature matrix now carry GPT-6.2 Astra and GPT-6.2 Sol (Astra 6-50, Sol 17-180 local messages per five hours on Plus, equal ranges on Standard Business; 200/10/1,000 and 50/2.5/250 credits per 1M at Standard speed; USD usage-based $8/$0.40/$40 and $2/$0.10/$10 per 1M; Enterprise/Edu off by default), recorded in the subscription fence paragraph and replacing its stale 10-100 Sol-messages figure, the Codex pricing page cited
- section-root: model-provider-feature-matrix: OpenAI subscription cell gained the ChatGPT plans' 6.2-pair rate-table and message-range facts (Astra 6-50, Sol 17-180 local messages per five hours on Plus, as of 2026-11-22), the Codex pricing page cited
- section-root: trackers-and-leaderboards: the Benchmark Radar coverage cell re-dated to 54,628 records over 117 days as of 2026-11-22 and the matrix's 11-22 bullet extended with the Benchmark Radar and bestvaluemodel volatile outcomes
- cross-category: 2026-11-22 stars scan empty past checkpoint, the newest star is the checkpoint star itself (abhinavxd/libredesk, 2026-11-21T04:48:22Z), page one reached the checkpoint on its first entry, checkpoint held
- cross-category: six broad sources carry over, AMO's one content addition (2026-11-22 #134) is the executions member switch's second listing, hesreallyhim twelve ticker commits only, andyrewlee, ai-for-developers, e2b-dev, slavakurilyak zero commits since the 11-21 scans
- cross-category: SeeleAI/Thoth (WecoAI/awesome-autoresearch 11-21 wave, Implementations and Forks) rejected on the thin-record tail, 51 stars, 4 forks, default branch dormant since 2026-06-15, zero independent coverage, orchestration kind flag, reopen on traction or independent coverage

## (daily refresh, 6)

- resolution: slopcodebench: admitted as the evaluation-review category's iterative-extension benchmark (SprocketLab/slop-code-bench, 324 stars as of 2026-11-22, MIT, pushed 2026-11-19, arXiv 2603.24755); note, matrix (twenty-six columns), and category index extended.
- resolution: codeaf: rejected below the coverage band (Agent-Field/CodeAF, 820 stars as of 2026-11-22, Apache-2.0, pushed 2026-11-22, v0.8.3 on 2026-11-19): footprint is directories, vendor compare pages, and aggregator news with zero HN stories and comments, and the "#1 on DeepSWE" headline is the maker's own same-model run with no reproduction in the repository.
- resolution: h5i: rejected on the subject mismatch (h5i-dev/h5i, 933 stars, Apache-2.0, pushed 2026-11-20): the kyrolabs listing's multi-agent worktree-race description matches nothing in the linked repository today, which is a Lean 4-verifiable web framework plus a web-security/red-team CLI; the described tool resolves to no repository.
- resolution: tale: rejected on the thin-record tail (tale-project/tale, 47 stars, 10 forks, MIT, pushed 2026-11-21, zero HN coverage for tale.dev/tale-project); the listed repository path does resolve, correcting the dispatch premise.
