↓ Skip to main content
  1. Agents/
  2. Agents log/

Agents log 2026-11-21

Author
Tom Rochette
Table of Contents

(daily refresh, 1)
#

  • nanoclaw: recorded v2026.11.0 stable (2026-11-20) ending the 2026.10.0 hold, with refreshed adoption numbers
  • picoclaw: volatile refresh, default branch still quiet since 2026-08-19 with side-branch pushes through 2026-11-20, open count 31 to 26
  • astrbot: volatile refresh, open issues-and-PRs 1,267 to 1,285
  • octop: volatile refresh, open issues 631 to 640
  • anything-llm: volatile refresh, open issues 342 to 332, pushed 2026-11-21
  • open-webui: volatile refresh, open issues 263 to 285
  • openclaw: volatile refresh, open issues 8,560 to 8,719, matrix and prose counts matched
  • zeroclaw: rotation re-verification, every cited URL re-fetched live, v0.9.0 still latest, open count 1,340 to 1,350
  • openworker: rotation re-verification, every cited URL re-fetched live, v0.3.1 still latest desktop release, Harness SDK recorded as a second README and site product line
  • assistant-runtimes: matrix re-verification, five status cells moved (nanoclaw stable v2026.11.0 at 1.0k, astrbot 1,285, octop 640, openclaw 8.7k, picoclaw 26), matching prose counts, membership unchanged at sixteen, columns verified sorted
  • assistant-runtimes: openhuman reads 42,783 stars on 2026-11-21 (about +16 in one day, the slowing trend continuing) against OpenClaw’s 394,941, the q13 trigger stays unfired, openhuman held
  • kapso: recorded the v0.9.5 release (November 18, on PyPI, the knowledge-graph per-page rank prior and the wiki-sync state-save fix), which shipped 13:07 UTC after the 11-18 fetch had recorded v0.9.4 the same day.
  • kapso: corrected the Hacker News footprint, which had stayed at zero stories while the makers’ Show HN launch thread (November 12, 61 points, 19 comments as of 2026-11-21) sat open; self-submitted, recorded below the independent-coverage bar.
  • ai-researcher: rotation re-verification, all six references live, dormancy (last push 2025-10-16) and the missing license file confirmed.
  • aide-ml: rotation re-verification, all six references live, PyPI still 0.2.2, repository quiet since the October 16 dependency upgrade.
  • arbor: rotation re-verification, five of six references live, the VentureBeat coverage now a site-wide 429 wall (article, domain root, and browser UA).
  • deepanalyze: rotation re-verification, all references live, Hugging Face usage re-read flat (239 trailing-month downloads, 93 likes) at the fresh fetch date.
  • dexter: rotation re-verification, status unchanged (last push 2026-09-23, v1.1.0 still latest, license still undetected).
  • automated-research: matrix Kapso surface cell moved to v0.9.5; structural checks pass (35-cell rows, sorted header, one-distinctive-cell pass over the Surface row).
  • code-review: Added cubic, the Cohere-acquired whole-codebase reviewer, matrix and category index extended to fifteen members [glm-5.3-flash]
  • code-review: Reopened the 2026-08-30 weak-tail rejection of cubic on Cohere-acquisition traction and admitted it [glm-5.3-flash]
  • cubic: Recorded the Cohere acquisition (2026-11-13, terms undisclosed, secure-coding vertical), the March 2026 Martian F1 lead at 65.7, the ReviewBench pre-ran-June caution, and the zero-HN-footprint finding [glm-5.3-flash]
  • open-code-review: Recorded v1.14.0 (November 21: behavior-and-impact analysis pass, severity-ordered findings, SQL, Dhall, and Ruby rules, CUDA and HIP support), ledger at 152 releases [glm-5.3-flash]
  • gito: Rotation re-verification, fork count 42 to 46, release state unchanged (v4.5.0, pushed October 1) [glm-5.3-flash]
  • openreview: Rotation re-verification, dormancy re-confirmed (pushed March 6, zero releases, no license file), no factual drift [glm-5.3-flash]
  • code-review: Rejected mountainowl/bubo (8 stars) and better-isms/heygrc (0 to 4 stars) on the thin-record tail, below-band with zero coverage [glm-5.3-flash]
  • wrenai: full rotation re-verification, every cited URL intact, main’s 2026-11-19 docs-only commits and the 2026-11-20 push read non-material, volatile numbers refreshed (2,119 forks, 247 open issues), no release since wren-v0.18.0
  • sourcegraph-code-context: full rotation re-verification, recorded the 2026-11-19 credit-rate cut (Deep Search 70 to 50 credits, Code Finder 11 to 3, Smart Hovers free, service accounts onto human-user billing) with a Price history row, the $16K Enterprise floor unchanged, blog newest-post claim advanced to the November 20 ruby2shell piece
  • llmlingua: full rotation re-verification, package frozen at 0.2.2, default branch unchanged (pushed 2026-11-09), forks refreshed to 456
  • semble: full rotation re-verification, trailing-month PyPI figure refreshed to 113,859, forks to 295, no release since v0.7.3 and no push since 2026-11-08
  • repowise: recorded main’s unreleased 2026-11-21 batch, PyPI unchanged at 0.61.0, counts refreshed (1,048 forks, 486 open issues)
  • rtk: dev-1.2.2 candidate train advanced to rc.662 carrying the post-v1.2.1 telemetry-marker and Windows git-help fixes, stable remains v1.2.1, counts refreshed (5,610 and 1,689)
  • serena: recorded main’s unreleased 2026-11-19 batch (find-references dropped from the health check as not universally supported), PyPI unchanged at 2.1.0, forks refreshed to 2,146
  • headroom: recorded main’s unreleased 2026-11-19 to 11-20 fix round, PyPI unchanged at 0.44.0, counts refreshed (6,277 forks, 326 open issues and pull requests)
  • repomix: corrected the post-v1.19.0 main summary again (2026-11-20 security-check worker-pool perf fix and handlebars removal), no release since v1.19.0
  • context-engines-feature-matrix: Headroom delivery cell wrap count corrected 15 to 16, TOON spec reference annotation to v4.4, no membership change
  • context-engines: category link pass over 220 external URLs, no dead sources (api.github.com 403s are rate limits, pypistats 429 rate limit, the scmagazine-to-scworld 403 stands on the 2026-11-20 ruling), front matter and internal links clean, matrix membership, sort, and cell counts verified
  • context-engines: entrant scan carried over, the Meirtz default branch is unchanged since 2026-05-28, no new candidates
  • control-planes: crowdstrike-falcon-guardian - rotation re-verification, every reference re-fetched and live (Reddit still 403 to scripted fetch, docs path still a client-rendered shell), added the Fiddler acquisition (2026-11-03, Falcon Observe for AI agent observability) and the Falcon AI Gateway explainer (2026-11-13) as references, the gateway’s routed-traffic-only boundary now a caution.
  • control-planes: code-atelier-governance-sdk - rotation re-verification, repository and PyPI unchanged (0 stars, 0 forks, v0.7.3 with 11 releases, dormant since 2026-07-23), production-branch changelog unchanged, Article 12 banner still frozen at its June reading, no drift.
  • control-planes: databricks-unity-gateway - rotation re-verification, all eight references re-fetched live, 2026-08-04 GA and service-policies Beta re-confirmed, ai_query rate-limit gap re-confirmed on the solved vendor-forum thread, no drift.
  • control-planes: mapped source awesome-ai-governance unchanged since 2026-11-20 (newest commit the already-resolved Trina AX addition), scan carried over; matrix verified matched (24 columns, case-insensitive sort, 25-cell row parity, Current status cells checked against work list metadata), category index entry list verified matched and alphabetical, all 24 notes pass front matter, tag, and internal-link checks.

(daily refresh, 2)
#

  • evaluation-review: crit v0.24.0 recorded (2026-11-20, Excalidraw renderer with comment-pinned canvas, an outside contribution) and the Arch AUR package added, rotation re-verified with every cited URL live
  • evaluation-review: reviewbench rotation re-verified and the live board’s shift recorded: the page now returns a ranked table topped by third-party OhMyBug at 83.0 grounded F1 (run dated 2026-10-25) with GitHub’s Copilot entries at #4 and #6, and the inaugural-board and publication-gate framing rewritten around it
  • evaluation-review: langfuse rotation re-verified (pricing page unchanged tier for tier), activity line to 2026-11-21, v4.83.0 still newest
  • evaluation-review: phoenix rotation re-verified, activity line to 2026-11-21, 22.0.0 still newest on PyPI
  • evaluation-review: harbor v0.27.0 recorded (2026-11-21, eight days after v0.26.0: Pi endpoint controls, cancelled-trial recording, Harbor Server announced) and the release-train wording moved to quickening
  • evaluation-review: verbatim-citation-gate v0.3.3 recorded (operator-signs gate fix ships) and the package is now on PyPI (0.3.2 and 0.3.3), ending the git-URL-only distribution recorded since creation
  • evaluation-review: frontierharness-eval re-verified out of rotation after its repository was rewritten on 2026-11-20: v1.0 tasks, results, npx CLI, and skill removed, v1.1 shipped as Harbor artifacts (22 configurations, 5 models, 62 tasks), still unlicensed
  • evaluation-review: tau2-bench activity line to 2026-11-21 (community leaderboard submissions merging), plannotator to 2026-11-21
  • evaluation-review: repaired the six notes whose 2026-11-20 Status edits carried no Changes bullets (agent-qa, harbor, langfuse, phoenix, plannotator, reviewbench), bullets appended dated 2026-11-20
  • evaluation-review: matrix maturity cells refreshed (crit v0.24.0, harbor v0.27.0, verbatim-citation-gate v0.3.3) with stale Jevals and slop-grader ages corrected; no membership change
  • evaluation-review: the mapped source carries its scan over unchanged (newest commit 2026-11-19T14:02Z, already resolved by the 2026-11-20 digest line)
  • executions: dedicated-list re-check again finds no credible dedicated executions source; StefanYasin/awesome-agent-ops (1 star, pushed 2026-11-19) and ipanalytics/Awesome-Agent-Ops (1 star, pushed 2026-11-16) both unchanged
  • copilot-automations: fully re-verified in rotation, all five sources re-fetched, claims re-confirmed unchanged; MCP-server tools recorded beside GitHub tools
  • claude-code-routines: fully re-verified in rotation, all five sources re-fetched; corrected trigger-management claim (GitHub triggers addable from the CLI at v2.1.225 or later), docs’ run-history section now carries the green-status caveat itself
  • claude-code: recorded the train reaching v2.1.330 (published November 20: Enterprise Admin Console org sign-in, the easel chart mod, full sign-in link flag, self-hosted runner start dir)
  • juggler: recorded the v0.8.2 release (published November 20, the preannounced version-bump release now on the feed: the Git pin grown into a full git client, ordered multi-provider web search, prompt history)
  • kilo-code: recorded v7.8.32 as the newest stable (published November 20, promoting the v7.8.31 telemetry change and adding the Agent Manager worktree-deletion confirmation guard)
  • openhands: recorded the v1.43.0 release (published November 20: setup-wizard automations, experimental feature flags with Conversation reactions, decision cards, message-edit views)
  • pi: recorded the v1.6.0 release (published November 20: GPT-6.2 Astra and Sol with Sol the default on three providers, a breaking render() array change, Qwen Token Plan model additions)
  • kimi-code: recorded npm 2.7.2 and desktop 1.4.2 (both published November 20, the global-thinking-effort revert fix mirrored across the two trains)
  • junie: recorded the installer train continuing through 4191.1 (published November 20, again without release notes attached)
  • qwen-code: recorded v0.28.3 as the latest stable (published November 21: GitHub as the third managed channel, a Generic Webhook Channel, WeCom and Teams adapters, Mods in ACP, headless background workflows)
  • copilot-cli: recorded the prerelease train reaching 1.0.115-5 (1.0.115-0 published November 19) behind the unchanged 1.0.114 stable
  • amp: rotation re-verification, every cited reference live, recorded the November 17 Amp for Windows release, tier page re-verified unchanged
  • bullet: rotation re-verification, every cited reference live and the benchmark page re-read unchanged, download windows refreshed (monthly 142, weekly 33)
  • fx: rotation re-verification, every cited reference live, the site’s binary-size claim moved from 6.17 to 6.33 MiB
  • grok-build: rotation re-verification, every cited reference live, the stable dist-tag promoted v1.0.92 with the alpha at v1.0.93 (both November 20), API prices re-verified unchanged
  • mirror: rotation re-verification, every cited reference live, PyPI at 0.1.2 (November 17), Beam’s OpenRouter price dipped again to $0.25 with $0.05 cache reads, Price history row appended
  • cline: rotation re-verification, every cited reference live, recorded the v4.1.27 release (published November 21) and the marketplace install count crossing six million (6,000,523)
  • harnesses: matrix verified against membership (45 columns, title-case-insensitive sort, cell counts, one distinctive cell per column against its note), front matter and category index clean
  • laya: 0.5.1 released 2026-11-21 (questions.json required in a policy’s manifest, laya-onnx can skip the ONNX Runtime download), pushed date moved
  • laya: JevBench open-weights board carries the first vega rows, Vega 0.8B rank 225 by capability at 32.995 and Vega 4B rank 221 at 36.579 (read 2026-11-21)
  • baml: stable train moved to baml-language 0.26.0 (2026-11-21), the 0.25.3 nightly line closing at its 20261119.a build, pushed date moved
  • cua-s1: two 2026-11-20 dependency-security commits on the component (mcp for GHSA-qx49-fqc8-xw99 and a python sweep, no scoring change), host pushed 2026-11-21
  • jeeves: rotation re-verification, all references holding, forks 26, weights 95 downloads, stale age claims corrected to seven weeks
  • jevbench: rotation re-verification, boards stable at 261 ranked open-weights and 54 API rows with tops unchanged, first vega rows noted
  • jevlike: rotation re-verification, all references holding, repository still frozen at 2026-09-16, forks 122, open issues 7
  • semif: rotation re-verification, all references holding, repository unchanged since 2026-09-23, forks 356
  • system1-arena: rotation re-verification, all references holding, leaderboard rows unchanged, dataset 498 downloads (window swing, no trend read)
  • hybrid-execution: matrix maintenance cells moved for baml, cua-s1, and laya; no membership change
  • memoryfields: rotation re-verification, every cited URL live, no facts changed (spec still v0.3, launch thread still 191 points, tool repo quiet since 2026-11-08 with the two August tags).
  • memvid: rotation re-verification, every cited URL live, no facts changed (no push since 2026-07-14, both registries frozen at 2.0.160, the source list still carrying Memvid as disputed).
  • metaclaw: rotation re-verification, every cited URL live, no facts changed (no push since 2026-06-07, v0.4.1 still the latest release).
  • letta: rotation re-verification, pricing ladder unchanged, landing-page repo still quiet since 2026-10-18, letta-code latest tagged release refreshed to v0.34.68 (2026-11-21).
  • agentmemory: rotation re-verification, every cited URL live, no facts changed (npm in sync at 0.9.33, v0.9.33 still latest, no dedicated HN thread), trailing-month downloads refreshed to 95,193 (2026-10-21 to 11-19).
  • memory: TeleAI source carries over unchanged since the 2026-11-20 check (no content commits after 2026-11-19T22:16Z), no new entrant candidates.
  • memory: funes, memos, and supermemory default branches re-checked on the day-behind-push ruling (CI dep bump, side-branch push, docs sync respectively), nothing material, no note edits.

(daily refresh, 3)
#

  • bifrost: release train transports v2.6.0 (2026-11-19, Model Routers SIMPLE/MEDIUM/COMPLEX classification with semantic, decision-model, or LLM classifiers on a Go 1.27.4 rebuild) with the 11-18 plugin patches, ent-v2.6.0-base, and helm-chart-v2.1.49; repo pushed 2026-11-21.
  • litellm: v1.110 line went stable (v1.110.1 2026-11-19 GPT-6.2 prompt-caching fix, v1.110.2 2026-11-20) with v1.109.2/3, v1.108.12-14, v1.106.11/12 patches, v1.111 rc.2/rc.3, and a v1.112.0-dev.1 train head; PyPI latest 1.110.2; repo pushed 2026-11-21.
  • ollama: v0.40.17 stable per the releases feed (published 2026-11-10, correcting this note’s rc readings), shipping Inkling 2 Mini (Thinking Machines, 26B, 5.4B active, Apple silicon only this release) and launch improvements; default branch moved 11-18 to 11-21; repo pushed 2026-11-21.
  • localai: v5.0.0 released 2026-11-18 (desktop app, Projects, server-side chats, Files, Office documents, Linux Python sandbox, team sharing, the add-on library) with v5.0.1 on 2026-11-20; repo pushed 2026-11-21.
  • magnitude: only QA-build prereleases past CLI 0.2.17 (npm 0.2.97-0.2.99 and cliqa-r3-r0-0.2.17, 2026-11-20/21); repo pushed 2026-11-21.
  • experiential: exp-code 0.1.8 (2026-11-17) and the gateway line’s v0.7.256 through v0.7.261 (2026-11-19 to 11-21) with a preview-input-outcomes prerelease; repo pushed 2026-11-21.
  • llm-gateway: default-branch model-catalog additions (Kimi K3.1 Pro Prime, Hy4, GPT-6 Luna Decisions, Pixtral Large, Lyria, Novita MiniMax rows retired) with no release past v1.26.0; repo pushed 2026-11-21.
  • kimi-code: twenty-fourth recorded OpenRouter move: kimi-k3 input floor eased to $0.1805/$11.0055 on Relace (cache $0.18) with the listing, Wafer $0.153, Chutes $1.85/$9.50, Moonshot $3/$15, and the 29-endpoint roster held; the k3.1-pro listing held $0.2530/$8.80 with the board at 28 and Morph re-priced to $1.306/$8.55; kimi-k2.7-code fell to $0.60/$3.00 (cache $0.12).
  • openrouter: fee state re-verified unchanged (5.5%/$0.80 Standard, crypto 5%, Business 8%, BYOK $25,000 then 5%); models listing reads 476 entries with 24 -pro ids.
  • glm-coding-plan: rotation re-verification, every cited figure held on all seven sources; the digitalapplied analysis’s 2026-11-17 correction recorded without moving cited claims.
  • qwen-coding-plan: rotation caught the Token Plan campaign promo reversion (Lite $6, Essential $10 from $3/$5 on 2026-11-04) and the Token Plan docs re-dated (guide 11-13, overview 11-17, FAQ 11-21); price history row appended.
  • chutes: rotation re-verification with GLM-5.3 PAYG cut to $0.80/$2.50 (cached $0.16) from the $1.225/$3.85 of 2026-11-07, the featured table at 19 models with GLM-5.2-Fast new, every other cited rate and the GPU rate held; price history row appended.
  • claude-plans: rotation re-verification, prices and multiplier rows held on all five sources; the pricing page’s Fable row now lists 50% of weekly limits on Ultra alongside Max; price history row appended.
  • model-access-feature-matrix: Qwen cheapest-paid-entry cell re-priced to Token Plan Lite $6 with the promo reversion.
  • model-access: category index Qwen one-liner re-priced; links, front matter, no-type checks, matrix 27-cell parity, and sorted membership verified across all 26 members.
  • ag2: recorded v1.3.3 (2026-11-19, google-genai cap lifted to <4), PyPI tracking at 1.3.3.
  • agent-swarm: recorded v2.14.1 (2026-11-20) and v2.14.2 (2026-11-21) on the train, npm matching at 2.14.2.
  • atlas: recorded alpha-0.5.0 (2026-11-20, agent identity resolution) as the latest release.
  • ax: recorded v0.4.0-rc.1 (2026-11-20) opening a release-candidate line toward a 0.4 stable.
  • autogpt: recorded the platform beta advancing to v0.12.0 (2026-11-19, breaking TRUST_X_FORWARDED_FOR removal).
  • autogen: rotation re-verification; archive declaration stands; successor microsoft/agent-framework went 404 with no redirect (satellites live, Go port still naming it upstream), recorded in Status, reference annotated.
  • bernstein: rotation re-verification; recorded v4.0.0 (2026-11-20, governance-tightening major with a migration guide) on GitHub and PyPI.
  • cezar: recorded desktop-v0.1.12 (2026-11-20) on the desktop track, v0.20.0 still the newest stable.
  • chorus: rotation re-verification; claims hold, v0.24.1 still latest.
  • claude-squad: rotation re-verification; claims hold, v1.0.20 still latest, push still 2026-11-13.
  • clawteam: rotation re-verification; dormancy stands, the HN search’s 26 hits are all other Claw-named projects.
  • conductor: rotation correction; the announcement now reads “Plans have changed! We’ll still be developing Conductor”, the dated wind-down schedule is gone from the page, downloads are advertised again, and the train continued past “final” 0.93.0 through 0.98.0 (2026-11-20); Status, Cautions, Pricing, Bottom line rewritten, matrix cells and index one-liner moved with it.
  • crewai: recorded 1.15.44 (2026-11-19, versioned shareable artifacts), PyPI tracking.
  • dmux: rotation re-verification; claims hold, npm window re-queried at the same 861, watch-list status stands.
  • emdash: recorded v1.2.13 (2026-11-19, empty body) at the head of the train.
  • frink: rotation re-verification with a site finding; frink.dev and its pricing page serve Vercel 402 DEPLOYMENT_DISABLED while the repository stays active (pushes through 2026-11-20), recorded in Status, Cautions, Pricing.
  • langgraph: recorded 1.2.24 (2026-11-20, REMOVE_ALL_MESSAGES fix), PyPI matching.
  • multica: recorded v0.11.1 (2026-11-19, doc-deletion restricted to creators and admins).
  • omnara: recorded the v0.2.26 pair (2026-11-19) and v0.2.27 pair (2026-11-20) on the omnarad and cluster lines.
  • open-swarm: recorded the shell build train moving to v1.8.1-exp.4/exp.5 (2026-11-17/18), update-channel builds.
  • openharness: recorded v0.6.79_cli (2026-11-21, Harness Pro ring counts, dial sessions on Windows/Linux) and v1.4.8_web (2026-11-21, website-only).
  • openrig: recorded v0.7.20 (2026-11-20, consent-first handoff and delivery holds) as the latest release.
  • orca: recorded v1.4.266 and v1.4.267 (2026-11-20) and v1.4.268 (2026-11-21) as the newest desktop releases.
  • pandacode: recorded v0.3.133 (2026-11-19), the withdrawn v0.3.134, and v0.3.135 (2026-11-20, re-issue), the second withdrawn-then-republished pair on the train.
  • parallel-code: rotation re-verification; claims hold, v5.1.0 still latest, push refreshed to 2026-11-21.
  • paseo: recorded v0.17.0-beta.1 (2026-11-19, existing-worktree workspaces, Antigravity approvals) opening the v0.17 beta line.
  • praisonai: recorded v4.8.0 (2026-11-19, security-hardening release) and praisonaiagents 1.8.1 on PyPI.
  • ruflo: recorded v3.79.1 (2026-11-20, npm README fix) and v3.80.0 (2026-11-21, signed SIM handoff ingestion).
  • scion: rotation re-verification; recorded v0.4.0-preview.3 (2026-11-20) on the preview line, push refreshed to 2026-11-21.
  • squad: recorded v2.0.0 going stable (2026-11-19, Fact Checker auto-scaffold, approve-once Copilot CLI contract).
  • superset: recorded desktop-v2.0.11 (2026-11-20) as the newest desktop release.
  • t3code: nightly tags recorded moved onto v0.1.12-nightly, v0.1.10 still the newest stable.
  • orchestration: matrix Current-status row synced (17 cells) with conductor and frink corrections, star figures untouched; membership, sort, cell counts, front matter, and internal links all verified clean.
  • lars-faye: rotation re-verification, all cited URLs live, feed still ends at the July 22 essay, course still $199 with the $100 early-bird, no claim moved.
  • chip-huyen: rotation re-verification, all cited URLs live, Amazon listing served 200 clearing the 2026-11-05 outstanding re-fetch, no claim moved.
  • boris-cherny: rotation re-verification, GitHub followers moved to 12k as of 2026-11-21, blog still ends at the 2026-09-19 post.
  • latent-space: rotation re-verification, newest issue moved to the formal-verification piece (2026-11-20) with AINews through 2026-11-21, subscriber count still absent and wiki figures unchanged.
  • ray-amjad: rotation re-verification, two new uploads (2026-11-04 and 2026-11-16, now 21 uploads April 9 to November 16), the Code Mode video at 221,495 views became his biggest, cohort page moved to the December run (185 lessons, 2026-12-07 to 12-18) and claims 5,000+ taught, 50K figure keeps its 2026-10-04 as-of.
  • people-and-publications: matrix ray-amjad cadence cell moved to 21 uploads through November 16; membership, sort order, cell-count parity, front matter, and internal links verified across all 34 notes.
  • langchain-agent-protocol: rotation re-verified with the second-implementation caution corrected (Aegra, independent Apache-2.0 Agent Protocol server, 1,329 stars as of 2026-11-21, four named production users) and the streaming bindings moved to langchain-protocol 0.0.22 (finish-reason fields, 2026-11-19)
  • acp: v2 moved from Draft to Preview (2026-11-20, schema-v2.0.0-beta.1) and the stable train to v1.20.0 with schema-v1.34.0 (subagent sessions and plan operations stabilized)
  • webmcp: spec-text revision moved to 2026-11-19, the Tool interface (getTools and executeTool now trade identity-stable Tool objects)
  • mcp: Events extension SEP-3415 merged (spec final, extension beta, poll, push, and signed-webhook delivery with cursor resumption); spec revision stays 2026-07-28
  • agents-md: rotation re-verified, every cited URL live, newest Claude Code release 2.1.330
  • a2ui: implementation wave on the default branch 2026-11-20, tagged train and spec table unchanged
  • ag-ui: client fixes on the default branch 2026-11-20, train stays release/2026-11-17
  • agent-host-protocol: dependency-security wave 2026-11-19 (CVE-2026-43798 swift-nio-ssh fix), train stays v1.1.0
  • uhp: implementing product’s train moved to v0.41.24, spec unchanged at its 2026-10-19 fifth version
  • protocols: feature matrix ACP and WebMCP maturity cells moved and the Agent Protocol prose gained the Aegra record
  • protocols: a2a, anp, pic-standard, and personal-agent-protocol re-checked on their pushed-at dates with nothing new behind them
  • chonkie: pushed 2026-11-19, trailing-month downloads 1,429,687 (as of 2026-11-21), PyPI 1.7.0 unchanged
  • code-graph-rag: pushed 2026-11-21 with the default branch’s latest commit a C++ lambda-creation-reference PR merge, open issues 370 to 304, PyPI 0.7.1 unchanged
  • docling: PyPI 2.151.1 (2026-11-19, via 2.151.0, 243 releases total), pushed 2026-11-21, trailing-month downloads 4,367,922
  • graphrag: pushed 2026-11-19 at the repository level with the default branch still at the 2026-11-02 workflow pinning, releases/latest still v3.3.0
  • langchain: pushed 2026-11-21, latest tagged release langchain-core 1.8.4 (2026-11-19)
  • lightrag: pushed 2026-11-21, open issues 295 to 235 aligned across sections, PyPI 1.5.7 stable unchanged
  • llamaindex: pushed 2026-11-20 at the repository level with the default branch still 2026-11-14, 1,512 open issues counting PRs, v0.14.26 unchanged
  • knowhere: release v1.2.64 (2026-11-21, retrieval evidence downloads moved to worker threads), pushed 2026-11-21
  • pageindex: PyPI SDK 0.7.4 (2026-11-19, 63 releases total, published after the prior run’s 0.7.3 fetch)
  • open-codebase-index: pushed 2026-11-20 with the default branch’s latest commit the startup and incremental-refresh perf PR #500, installs 5,536 to 2026-11-21 roughly flat, npm 0.40.1 unchanged
  • ragflow: pushed 2026-11-20, open issues to 1,297, v1.0.0 and the agent-native-database description unchanged
  • unstructured: pushed 2026-11-19, open issues 456 to 474, PyPI 0.27.37 unchanged
  • r2r: rotation re-verification re-fetched every cited URL, all claims unchanged (repo parked at 2025-11-07, PyPI 3.6.6, docs 404, sciphi.ai unreachable, footprint unchanged)
  • orama: rotation re-verification re-fetched every cited URL, trailing-month installs 8,507,702 (fresh high), release stall, quiet main branch, and HN footprint unchanged
  • retrieval: feature matrix maintenance row refreshed to 2026-11-21, reading figures aligned, no membership change
  • retrieval: category index entry list and matrix membership re-checked against all sixteen notes, no membership change
  • retrieval: awesome-rag-production (last push 2026-11-14) and Danielskry/Awesome-RAG (last push 2026-10-06) both unchanged since their 2026-11-19 checks, scans carried over

(daily refresh, coordinator)
#

  • cross-category: the ReviewBench board’s third-party top entries (OhMyBug, intent-review-agent) stay evidence inside the reviewbench note, no repository resolvable for either
  • code-review: shippie (mattzcarey/shippie, 2,531 stars) held report-only on the non-credible-source sighting line, reopen on a mapped-source listing or independent coverage
  • harnesses: q22 wiring evidence refreshed, walkinglabs/awesome-harness-engineering woke (pushed 2026-11-09) and ai-boost/awesome-harness-engineering stays active (pushed 2026-11-20), both canonical paths verified

(daily refresh, 4)
#

  • agent-sandbox: recorded v1.0.12 (2026-11-20, suspend/resume and archive extraction across the SDKs, sandbox-scoped router v2 tokens), release count to thirty-two; rotation re-verification, all cited URLs alive
  • agent-substrate: recorded v1.0.1 (2026-11-20, egress-gateway fix refusing resolved loopback, link-local, and cloud-metadata destinations); rotation re-verification, all cited URLs alive
  • aigate: matrix maturity cell corrected from the stale v1.0.0 to v1.1.2 (2026-11-10 ssh-agent-leak security wave), matching the note’s 2026-11-12 correction
  • aicontainer: recorded 0.14.3 (2026-11-20, npm package ships the agent-setup guide, Claude Code plugin removed); trailing-month downloads 717 to 713
  • brig: recorded v0.6.0 (2026-11-20, breaking upstream-following git rewrites in policy, Brigfile targets with approval-gated boots, credential-pattern redaction, Linux hull work)
  • opensandbox: recorded release-1.3.0-rc.2 (2026-11-19, rootfs-only snapshots extended to native S3 sandboxes, BatchSandbox Completed terminal phase)
  • drop: volatile refresh, firewall train through 2026-11-20 (persisted allow/deny rules, connection-time enforcement, auto host-port removal), v0.3.2 still latest
  • nemoclaw: volatile refresh, train through 2026-11-21 (v0.0.145 prep, OpenShell 0.1.9 recognition); open issues 876 to 318 on a 781-item bulk closure wave since 11-15, most without a completion reason
  • nono: volatile refresh, signed-policy train through 2026-11-20 (signed policy enforced at startup, signed file bundles verified); advisories re-checked at fifteen; open issues 304 to 308
  • microsandbox: volatile refresh, 0.8 line cut on main unreleased (breaking snapshot import limits, DNS resolver isolation), v0.7.16 still newest published
  • openshell: volatile refresh, post-v0.1.9 train through 2026-11-21 (GitHub App refresh strategy, subtree-path credential templates, Helm capacity sizing); open issues 741 to 781
  • e2b: Status PyPI figure aligned to 2.64.0 (2026-11-21), matching the matrix’s 2.64.x maturity cell
  • cubesandbox: rotation re-verification, all cited URLs alive (repository, license, latest release, PyPI at 0.8.0, site, three HN threads holding at 7, 5, and 3); post-v0.8.0 train through 2026-11-21
  • artifact-fs: rotation re-verification; Artifacts billing start moved to 2026-11-19 (rates unchanged) and a repository tier announced as coming soon, two Price history rows added
  • sandboxing: matrix maturity cells refreshed for the release-landed members and the aigate cell corrected; 25 columns re-verified sorted, aligned, and count-matched; membership unchanged
  • session-analytics: agentsview - volatile: docs schema contract back to version 5, the fourteenth recorded flip; counts refreshed
  • session-analytics: ccusage - volatile: unreleased Rust memory and binary-size work with –last window fixes; DeepSeek Harness twentieth source still pending release; trailing-month downloads 1,059,766
  • session-analytics: agentmeasure - recorded v0.15.0 through v0.15.4 (2026-11-20, PyPI verovo 0.15.4): buyer identity and post-signoff snapshot, bill-source integrity with fail-closed locking, Growovo handoff premises and shared error families, task-bound normalization
  • session-analytics: coucou - recorded Mac v0.6.1 (2026-11-19): ProMotion frame-rate fix, Botan founders list pointing at iPhone-back hardware, crash and Hermes fixes
  • session-analytics: token-monitor - recorded v0.94.0 and v0.95.0 (2026-11-19/20); README coverage moved to 55 named tools plus the aggregate (headline 56-plus), Junie and OpenCodeReview the newest rows
  • session-analytics: claude-code-agent-monitor - rotation: two community PRs unreleased since 2026-11-19 (editor-folder launcher, Windows test gate); maintainer concentration refreshed at 1,101 of 1,313 commits
  • session-analytics: codeburn - rotation: three fixes unreleased since v0.9.28 (Cursor Keychain Deny stop, Kimi Code managed workspaces); trailing-month downloads 36,573
  • session-analytics: claude-replay - rotation: still quiet since mid-September, trailing-month downloads 8,102
  • session-analytics: matrix - Token Monitor cells and prose counts to 55 named, the Kind cell’s stale 51 fixed; schema line to version 5, the fourteenth flip
  • session-analytics: index one-liner moved to the widget’s 55 named tools
  • agent-native: volatile refresh, the pack’s 11-19 push #127 packages turn-into-app as an Anthropic plugin (.claude-plugin manifest, icon, mcp.json) with the catalog held at 26, npm held 0.4.10 with 5,780 pinned-window downloads (11-15 to 11-21), framework pushes through 11-21 remain app development, HN re-run zero hits
  • agent-plugins: rotation re-verification, every cited URL re-fetched with all v1.0.0, draft, and critical-guide claims held, pushed_at 11-20 a side-branch push with main still at 11-14
  • opencode-skills-and-plugins: rotation re-verification, release train moved to v2.3.2 (11-18) via v2.3.1 (11-17), the 1.x line released v1.19.6 (11-17) ending its quiet stretch, docs unchanged in substance, ecosystem counts held 31/10/2
  • skills: matrix Agent-Native no-coverage date re-run to 11-21, no cells moved
  • loopforge: note created from the varun1505 2026-11-14-through-11-19 sweep admission (Tencent/LoopForge, 5,352 stars as of 2026-11-21, MIT, active, zero HN) on the SuperPlane platform line, the stored-SOP loop the differentiator.
  • software-factory-feature-matrix: thirteen columns with LoopForge inserted between Loki Mode and Machinist; loop-owner, isolation, and self-improvement readings and the choosing bullet updated; the long-tail paragraph records the 2026-11-14-through-11-19 sweep resolutions.
  • software-factory: LoopForge added to the category index.
  • fabro: volatile refresh, pushed 2026-11-20, nightly train at v0.423.0-nightly.0 (2026-11-20); no material change.
  • har: recorded the v1.25.2-through-v1.30.0 run (nine releases, 2026-11-15 to 11-20: usage invariants and a usage-integrity line, work-unit binding on a running slot, an onboarding telemetry-level choice, a personal worktree on/off setting); HAR HQ pricing re-verified unchanged.
  • loki-mode: recorded the train to v11.174.0 (2026-11-21, npm in sync): whole-issue runs the default road on every model, parallel issue runs, per-role model picks, the ALREADY_SATISFIED seal, measured-only conflict labels, and a label-calibration replay suite, with v11.154.0 fixing a false-VERIFIED window; Docker pulls 338,759.
  • machinist: recorded the PR-tracking merge (PR #614), 290 commits past v0.4.0 with v0.4.0 still newest; machinist.sh re-fetched and still serving the pre-redesign app description; zero coverage holding.
  • open-inspect: volatile refresh, pushed 2026-11-21, the release list still empty and the zero-coverage search holding; no material change.
  • open-swe: recorded Robin Desktop v0.3.19 (2026-11-17) as the newest stable desktop tag, grounded through the by-tag endpoint while the releases list stays a nightly wall.
  • ouroboros: rotation re-verification, every cited URL re-fetched and all claims confirmed; recorded the train to v0.64.0 (the v0.61.0 model-passthrough minor, the v0.62.0 Auto Seed-review-stage removal, the v0.63.0 evaluation fixes).
  • super-simple-software-factory: rotation re-verification, every cited URL re-fetched and all claims confirmed (still a single commit on main, no push since 2026-08-04, no releases); fork count refreshed.
  • superplane: volatile refresh, pushed 2026-11-21, newest main commit the 2026-11-20 subscription-models fix (unreleased), fleet-manager v0.1.7 still newest, pricing page re-fetched unchanged.
  • zeroshot: recorded the train to v16.2.0 (v16.1.0 removed the map overflow group control, v16.2.0 advertised selectable gateway wire protocols), pushed refreshed.
  • spec-driven-development: spec-kit - rotation re-verification of every cited source (all alive): recorded the v1.4.1-v1.4.3 releases (2026-11-16 to 19: community-catalog additions, drift analysis in tasks, 30/30-day stale lifecycle, extension-install fixes) and re-grounded the integrations claim after the docs restructure stopped carrying a counted page
  • spec-driven-development: spec-kitty - rotation re-verification of every cited source (all alive, the Hysen review now 301ing to its canonical path): corrected the stale rc2 prerelease-pin claim, recorded the default branch’s unreleased stale-lane sibling-recovery work, PyPI figure 2,576 last month
  • spec-driven-development: pdd-kit - rotation re-verification of every cited source (all alive): the default branch still sits at the v0.40.0 release commit with pushed_at 2026-11-20 on side branches, npm figure 2,258 last month
  • spec-driven-development: openspec - recorded the v2.1.0 and v2.1.1 releases (2026-11-17 and 19: global skill installs through Vercel’s skills installer, then skills shipping inside the CLI package with no Git or network needed, archive refusing to half-finish) and refreshed the download count (5,669,156 last month)
  • spec-driven-development: gsd - recorded the v1.22.0 release (2026-11-19: STATE.md size warnings, executor_runtime declaration with the opt-in cross-runtime gate registered ahead of the dispatch work, the resolve_model_ids omit tier fix, lazy-loaded gsd-tools, executor-routing questions at new-project time) and refreshed the npm figures (91,281 successor, 100,911 legacy)
  • spec-driven-development: arc-kit - recorded the v6.22.1 through v6.24.0 releases (2026-11-17 to 20: section-link fixes and a Claude Code plugin page on arckit.org, a new overlay pack with eight document types and generated answer worksheets, closed questions through Claude Code’s question selector) and moved the community-plugin count to nineteen
  • spec-driven-development: archcore - recorded the v0.15.6 through v0.15.9 releases (2026-11-17 to 19: nearest-store resolution, the new archcore check command, proof-tracked plans whose checked tasks carry pinned proofs)
  • spec-driven-development: ai-unified-process - recorded the eighth release wave (2026-11-16 to 17: aiup-core v3.19.1-v3.20.1 parenthesized actors, diagram-color warnings, and relative-link fixes) with the default branch’s unreleased open-questions and PlantUML work beside
  • spec-driven-development: ai-dlc - volatile refresh: pushed 2026-11-21 with 355 open issues and the default branch carrying unreleased traceability and journey-test work, v2.11.4 still the newest stable
  • spec-driven-development: mcp-probe-kit - volatile refresh: pushed_at 2026-11-15 on side branches with the default branch quiet since 2026-10-28, npm figure 797 last month
  • spec-driven-development: cc-sdd - volatile refresh: npm figure 75,683 downloads last month, the repository still pushed 2026-10-22 at v3.1.0
  • spec-driven-development: vibe-coding-prompt-template - recorded the v3.2.0 release (2026-11-17, The Fresh-Chat Release, shipping vibeworkflow 0.4.0 with fresh-chat prompts, CI drift-checked templates, and a Quick/Guided/Deep router) and refreshed the npm figure (328 last month)
  • spec-driven-development: bmad-method - volatile check only, no edit (the 2026-11-20 pushed_at a side branch, default branch quiet since 2026-11-12, v6.12.4 still the newest release)
  • spec-driven-development: the feature matrix refreshed the adoption row to 2026-11-21 as-of dates (cc-sdd 76k npm, GSD 91k npm, OpenSpec 5.7M npm, MCP Probe Kit 797 npm, PDD Kit 2,258 npm, Spec Kitty 2.6k PyPI) and corrected the OpenSpec workflow cell to /openspec-propose (the command v2.0.0 removed)
  • spec-driven-development: the category index moved the OpenSpec one-liner’s npm figure to about 5.7M a month

(daily refresh, 5)
#

  • surfaces: cc-haha volatile re-verification on the released signal, recorded v0.7.9 (November 20, Agent Teams pause and resume with member-model switching while paused, retryable team resume, Claude subscription usage display, app-icon badge, session Markdown export, neutral dark theme and resizable workspace preview, output-token escalation, image compression, mobile app 0.1.5) as the new latest release; no matrix cells moved
  • surfaces: codeg volatile re-verification on the released signal, recorded v0.47.0 (November 20, pi-ai DeepSeek provider routes, Codex 2.8.0 with GPT-6.2, Claude Code adapter 0.109.0, plan-checklist persistence, OpenCode startup-wait handling, memory reductions) as the new latest release, and cleaned the Status release chain’s stale v0.43.0 tail the November 14 to 19 passes left behind; no matrix cells moved
  • surfaces: jetbrains rotation re-verification, all cited URLs re-fetched live and standing; junie.jetbrains.com rendered its marketing shell without the plan figures again, so the $8.33/$25.00 ladder stands dated 2026-11-08, the AGENTS.md docs (redated November 19) and Air Governance GA post stand, the three threads read 596, 317, and 78 points, and helgilibrary.com’s site-wide 403 wall is confirmed standing
  • surfaces: px0 rotation re-verification, all five cited URLs re-fetched live and standing (repository pushed November 6 and not archived, v0.1.21 still the latest release with the train paused seventeen days, Show HN still 7 points and 0 comments, the site still review-workbench positioned)
  • surfaces: roo-code rotation re-verification, all cited URLs re-fetched live and standing (archived repository still pushed May 15, v3.54.0 the final release, marketplace listing still 404, roomote.dev still pitching self-hosted source-available single-tenant model-agnostic, thenewstack.io still 403 to curl and full-text to webfetch)
  • surfaces: trae rotation re-verification, all cited URLs re-fetched live and standing ($20/$60/$200 ladder unchanged with 5,000 free completions and 2 cloud tasks, trae-agent still stalled at its February 5, 2026 push with no tagged release ever, HN threads at 954 and 85 points)
  • surfaces: OpenBitFun’s Cowork-style general-workspace rejection holds on the fourteenth q#14 re-read (the openBitFun/OpenBitFun README re-fetched 2026-11-21 still pitches a general-purpose agent workspace with four Agent Harness modes, cross-device remote control, and a Mini App marketplace, 4,102 stars, pushed 2026-11-20), the question again reads safe to close
  • surfaces: no credible surfaces-dedicated source has appeared; the GitHub search re-run returned member-ecosystem lists (ZhangYu-zjut/awesome-Antigravity 160 dormant since January, MGrin/awesome-bb-plugins 71, MichaelZelbel/awesome-antigravity 40, jojomensah89/awesome-trae 22 dormant), one broad ecosystem list (0xWelt/Awesome-Agent-Ecosystem 124), broad tool lists (PierrunoYT/awesome-ai-dev-tools 78, dalisoft/awesome-ai-coding 47, QAInsights/awesome-ai-tools 34, ishandutta2007/Awesome-AI-Code-Editor 18, ai-solutions-sk/awesome-ai-coding-tools 4), and the known agent-clients list (1shiharat/awesome-agent-clients 17 pushed 11-17), none a category list
  • surfaces: zed’s v1.29.3 (November 20, two bugfixes), bb (desktop-v0.50.1), openchamber (v2.7.3), vscode-copilot (1.147.0), whiteboard (v0.5.2), codenomad (v1.1.0), and puppyone (v0.4.4) releases all match their recorded latest; continue’s main branch is quiet since July 21 with pushes on side branches only, so no notes moved
  • surfaces: matrix verification pass, no changes needed - 25 columns case-insensitively sorted by member title, every row at the header’s 26-cell count, the Open source row’s license cells verified against every member note, the index list matched to membership and alphabetical
  • task-management: agentrq - recorded the v0.16.0 through v0.16.5 train (WebAssembly-sandboxed hooks with built-in generic and GitHub hooks, hook URLs, the server DDoS guard and rate limiter, optional Sentry and Prometheus, workspace presets, SQLite write serialization) and refreshed counts (85 forks, 3 open issues and pull requests, pushed 2026-11-21); v0.16.5 now the newest release
  • task-management: backlog-md - refreshed counts (109 open issues and pull requests, pushed 2026-11-21, 161,829 npm downloads over the trailing month to 2026-11-21); v1.55.3 still the newest release
  • task-management: beads - recorded the v1.4.0-rc train advancing to rc.3 (2026-11-18) and rc.4 (2026-11-20), binary swaps with no new migration, and refreshed counts (1,949 forks, 1,781 open issues, pushed 2026-11-21, 51,856 npm); v1.3.4 still the newest stable release
  • task-management: epiq - recorded v1.18.0 (2026-11-18: standalone setup and init commands, the :q TUI quit family, oldest-first GUI comments contributed by zz-simbe, filter persistence across board switches) and the unreleased performance wave on main, counts refreshed (25 forks, 2 open issues and pull requests, pushed 2026-11-21, 4,825 npm); v1.18.0 now the newest release
  • task-management: its-a-plan - recorded v1.12.0 (2026-11-19: Pro Asana import and the wider audit log, Community color themes and comment reactions, background agent runs through the API with get_agent_run, Gitea and Forgejo pull-request links, Kie and DataForSEO agent tools) and refreshed counts (324 forks, 82 open issues and pull requests, 32 releases, pushed 2026-11-21); v1.12.0 now the newest release
  • task-management: kanzo - recorded main’s unreleased community work since v0.4.2 (a sandbox-friendly build, MCP task and project moves, an x86-64 SQLite baseline fix, the shared-boards sidebar) and refreshed counts (112 forks, 16 open issues, pushed 2026-11-20); v0.4.2 still the newest release nine days on
  • task-management: kata - refreshed counts (56 forks, 120 open issues and pull requests, pushed 2026-11-21); v0.23.0 still the newest release, unreleased main work noted
  • task-management: ordewell - recorded the v0.13.4 through v0.14.4 line with the desktop app shipped in installers for macOS, Windows, and Linux (v0.14.1) and the stability passes, and refreshed counts (20 open issues and pull requests, down from 39, pushed 2026-11-21, about 9,500 npm); main carries an unreleased Pi runner, v0.14.4 now the newest release
  • task-management: pad - recorded v0.21.0 (2026-11-18: Agenda week and month views, resolution undo, occurrence acknowledge, private ICS calendar feeds, item-page commitments, linked time gates, forecasts) and v0.21.1 (the OAuth form-body cap and JWT client-assertion refusal) and refreshed counts (18 open issues and pull requests, pushed 2026-11-21); v0.21.1 now the newest release
  • task-management: veritas-kanban - refreshed counts (99 forks) with the default branch checked directly: the 2026-11-02, 11-09, and 11-16 pushes all landed off main; v6.3.0 still the newest release
  • task-management: agent-kanban - rotation re-verification with every cited URL re-fetched live: nothing material changed, with v1.13.4 still the newest tag and the last push still 2026-11-13
  • task-management: ccpm - rotation re-verification with every cited URL re-fetched live: all claims confirmed, pushed still 2026-11-04 with no tagged releases, forks refreshed to 839
  • task-management: operon - rotation re-verification with every cited URL re-fetched live and recorded 4.8.0 (2026-11-18: embedded filter dates, the palette timer command, the mini calendar sidebar, the shared Task and Property Editor section layout), counts refreshed (pushed 2026-11-18, 1,095 npm); 4.8.0 now the newest release
  • task-management: task-management-feature-matrix - refreshed the release and activity cells and the Ordewell kind cell, the reading prose moved to the v0.14.x line
  • task-management: the dedicated-list sweep found no credible task-management-specific list for the seventh consecutive run; its incidental candidate TechDufus/openkanban rejected on the thin-record below-band tail
  • 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: v0.15.0 published 2026-11-15 after the last run’s fetch (homepage search index, benchmark and score additions ARC through NAF-Bench, Cohere Labs talk deck), 11-20 pushes title the talk slide Benchmarks’ Last Exam and merge SpatialBench, scBench, ClawBench, and AI-professions imports, badge 54,092 records over 116 days as of 2026-11-21
  • trackers-and-leaderboards: bestvaluemodel - Volatile re-verification: 11-16 to 11-21 diff runs add Qwen4 Max, ten GPT-6.2 Astra/Sol rows, four Gemini 4 Neon rows, Inkling 2 Mini and three Inkling/LFM3 rows, remove Gemma 4 E4B two days after adding it, and double the Carbon and Argon price lines back to blended $8
  • trackers-and-leaderboards: margin-lab - Rotation re-verification: the 11-18 commits pin agent CLI versions per daily run, run the trackers independently, and resolve pinned versions the way a task container does, both trackers still paused with baselines from runs beginning September 24, 2026, tracker figures now client-rendered, HN thread 760 points unchanged, v0.5.6 still newest release
  • trackers-and-leaderboards: vectara-hallucination-leaderboard - Volatile re-verification: the 11-17 pushed_at movement is a side-branch push, default branch still at the 10-23 leaderboard merge, October 22 stamp and 1.8 percent leader unchanged, refresh gap four weeks
  • trackers-and-leaderboards: artificial-analysis - Rotation re-verification: coverage 763 models, Intelligence Index v5.0.1, new Progress Index v1, Decision Index v2.1, API Performance Benchmarking v3, Endpoint Accuracy Index v1.1, Google DeepMind MKA partnership, Coding Agent Index v2.1, November release-wave evaluations recorded (Fable 5.2 Max 65 leads, GPT-6.2 Astra 62 the #2 model)
  • trackers-and-leaderboards: benchlm - Rotation re-verification: coverage 1,034 profiles, 828 benchmark pages, 215 ranked with 145 Supported, leader Claude Fable 5.2 at 93.30, scoring now BenchAlign v6.0 (two-parameter IRT, frozen 92-model cohort), and the failed-checks disclosure is gone from the methodology page, note corrected
  • trackers-and-leaderboards: trackers-and-leaderboards-feature-matrix - Cells refreshed from the six re-verifications, BenchLM methodology and verification cells corrected for the v6.0 rebuild
  • trackers-and-leaderboards: trackers-and-leaderboards - Dedicated-list check re-ran 2026-11-21: no new candidate above the bar, SAILResearch/awesome-ai-leaderboard (383 stars, pushed 11-16) still the pending proposal awaiting the owner’s wiring call
  • section-root: model-provider-feature-matrix: delta-pass price check, the OpenAI coding-specialized cell moved on gpt-5.3-codex’s deprecated marker (price unchanged $1.75/$14) and the Moonshot workhorse cell on kimi-k2.7-code’s aggregator listing re-pricing onto its floor row ($0.60/$3.50 to $0.60/$3.00, cache reads $0.16 to $0.12, CoreWeave’s int4 row), the K3 floors held both legs (Wafer $0.153/$16.00, Chutes $1.85/$9.50, listing $0.192/$16.00, roster 29, Relace output to $11.01) and the K3.1 Pro anchor held at $0.2403/$8.4023 on its fourth read with the listing holding at $0.253/$8.80 and the board widened 27 to 28, every other cell unchanged against all seven pricing pages, the -pro set re-confirmed, models.dev showing no new covered-provider releases after the 2026-11-18 Neon rung [glm-5.3-flash]
  • section-root: model-selection-for-coding-tasks: synced with the same sweep, gpt-5.3-codex’s deprecated marker recorded in the Notes row, kimi-k2.7-code’s listing folded into the floor paragraph, the K3 and K3.1 Pro floor sentences and the fourth clock re-dated, and the gpt-5.6 re-check sentence extended with the 2026-11-19 and 2026-11-21 reads [glm-5.3-flash]
  • section-root: model-benchmark-matrix: link re-verification, all 35 cited URLs fetched 200, the OpenAI PaperBench page readable again for the first time since 2026-10-11 with the four claims re-read on the page itself, Terminal-Bench still 4.1, as-of re-dated [glm-5.3-flash]
  • section-root: agentic-coding-tools-landscape: link re-verification with no fact moves, eighteen of nineteen cited links 200 (Bloomberg again 403), DeepSeek Harness newest prerelease still dsh-v0.2.2-rc.5, Cursor models docs and pricing unchanged, Gemini CLI enterprise banner unchanged at its 2026-11-10 stamp [glm-5.3-flash]
  • section-root: context-management-patterns: link check, all nine cited pages 200, the compaction-survival, skill-cap, v2.1.304, /autocompact, and idle-compaction claims holding on the live docs, as-of held [glm-5.3-flash]
  • section-root: the-agentic-development-environment-landscape: rotation re-verification with no fact moves, all 17 cited URLs 200, Air’s supported-agent and surface claims holding, the eleven referenced repositories live and unarchived (Paradev MIT, Orca MIT, Paseo NOASSERTION Apache-2.0 wrapper, Emdash Apache-2.0, agent-orchestrator still reading 35-plus) [glm-5.3-flash]
  • section-root: the-tells-are-structural: re-fetched all eight cited sources 200, study figures and the no-ai-slop ruleset’s six structural rule ids unchanged, star figures untouched [glm-5.3-flash]
  • cross-category: the andyrewlee 2026-11-20 addition (Trina AX, b1ccdf6d) is a second listing and the 2026-11-20 control-planes thin-tail rejection carries by mechanism (trina-app/trina-ax, 94 stars as of 2026-11-21, AGPL-3.0, same subject the awesome-ai-governance 11-20 pass rejected at 92 stars with zero coverage)
  • cross-category: ResearchForge (forger-labs-hq/researchforge, 12 stars, 0 forks, Apache-2.0, pushed 2026-11-20, PyPI 0.4.8 at 943 downloads/month as of 2026-11-21) rejects on the thin-record tail, automated-research kind, footprint entirely maker-authored, reopen on independent coverage or traction
  • cross-category: SciΠvot (AutoResearch-Factory/SciPIvot, 15 stars, 1 fork, dormant since 2026-10-19, arXiv 2610.21082) rejects on the zero-adoption floor with the benchmark routing to evaluation-review recorded and member Agon’s org-sibling relationship named, reopen on third-party adoption or coverage
  • cross-category: the 2026-11-21 stars scan resolves one rejection, abhinavxd/libredesk (3,187 stars, starred 2026-11-21T04:48:22Z) a self-hosted customer support desk with no agent surface, the no-agent-story line; checkpoint advances to 2026-11-21T04:48:22Z
  • cross-category: AMO, ai-for-developers, hesreallyhim, e2b-dev, and slavakurilyak carry over unchanged (AMO and hesreallyhim ticker-only commits since 11-18, the other three with no commits in the window)

(daily refresh, coordinator, wave 5)
#

  • kimi-code: recorded the K3.1 Pro host-floor anchor’s fourth read ($0.2403/$8.4023 on the same Relace fp4 row, roster 28, the main listing $0.2530/$8.80 now matching Wafer’s mxfp4 row) and kimi-k2.7-code’s listing cutting its output leg to $0.60/$3.00 (cache hits $0.12) onto its host floor; Price history row appended
  • model-selection-for-coding-tasks: the two gpt-5.6 Pricing-windows cells refreshed from the 2026-11-14 re-check to 2026-11-21