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

Agents log 2026-10-25

Author
Tom Rochette
History / Edit / PDF / EPUB / MD
Created: / Updated: / Status: finished / 29 minutes read (~5723 words)
agents 103
Table of Contents

(daily refresh, 1)
#

  • section-root: model-selection-for-coding-tasks: added Kimi K3.1 Pro (released 2026-10-23, $2/$10, cache $0.20, flat 1M, HighSpeed $4/$20) as a column, the first Moonshot model at the converged workhorse rate
  • section-root: model-selection-for-coding-tasks: K2-series sunset recorded on kimi-k2.7-code, closed to new users 2026-10-31 with all K2 models slated to retire by end of December 2026
  • section-root: model-selection-for-coding-tasks: Kimi K3 OpenRouter floors to 2026-10-25 readings, the listing at $0.239/$10.80, the input floor $0.19/$8.94 on Relace, the output floor $1.20/$5.50 on Sail Research, roster 29; gpt-5.6-terra and gpt-5.6-luna still listed on OpenAI’s pricing page
  • section-root: model-provider-feature-matrix: Moonshot flagship cell moved to K3.1 Pro and the K2 retirement stamped across the Moonshot cells, the challengers-sweep line rewritten now that a challenger flagship matches the workhorse rate
  • section-root: the-agentic-development-environment-landscape: rotation re-verification, all 17 URLs 200, Air’s paragraph rewritten against its live page (IDE plugin plus web/CLI/mobile, Claude Agent plus ACP roster, Junie folded in)
  • section-root: agentic-coding-tools-landscape: DeepSeek Harness prerelease to dsh-v0.2.1-rc.3 (2026-10-23) and the Gemini CLI enterprise-docs stamp to 2026-10-22, Cursor pricing re-checked unchanged
  • section-root: the-tells-are-structural: all eight cited sources re-fetched 200, as-of moved to 2026-10-25

(daily refresh, 2)
#

  • session-analytics: agentsview - volatile pass: no release after v0.46.1 with main-branch work landing 2026-10-24/25 (session-context export, friction digests, PostgreSQL push optimizations), token-usage docs re-verified at schema version 5.
  • session-analytics: agenttrace - rotation re-verified: every cited URL live, v0.10.1 still latest with three unreleased default-branch commits from 2026-10-05/06, npm still 0.10.1, HN footprint still empty.
  • session-analytics: ccusage - recorded v20.0.42 through v20.0.44 (2026-10-23/24): source-specific config in unified reports and a release-packaging fix; npm latest 20.0.44, trailing-month downloads 826,577 (2026-09-23 to 2026-10-22).
  • session-analytics: claude-code-agent-monitor - recorded v2.4.5 (2026-10-23), a release-tooling version bump with no feature changes.
  • session-analytics: claude-replay - rotation re-verified: every cited URL live, v0.11.0 still latest, the 2026-10-16 dependency bump still the only commit since 2026-09-18, weekly npm downloads 406 (week to 2026-10-22).
  • session-analytics: clawtrace - rotation re-verified: every cited URL live, repository (pushed 2026-08-15, not archived), npm plugin 0.1.23, credit packages, paper, and 1-point Show HN all unchanged.
  • session-analytics: coucou - recorded Windows/Linux 0.5.3 (2026-10-23, matching the Mac line’s version) and Android 0.3.0 (2026-10-22, pairing-flow release).
  • session-analytics: ctx - volatile pass: no release after v2.2.8 with four unreleased fixes landing 2026-10-25 (release-advisory security remediation, macOS 26.2 SDK repin, admission-failure and exhaustive-replay fixes).
  • session-analytics: memex - recorded v0.31.1 (2026-10-24, released without notes): fix wave over resume, search, usage, and ingest (ten code-review findings, Hermes turn overlap, Kimi retry drift, a .DS_Store wedge).
  • session-analytics: token-monitor - recorded v0.78.0 (2026-10-23): Edge Dock glass styles for Windows 11 and macOS, ZCode session titles with subagent grouping, tooltip and z-order and quota-check fixes.
  • session-analytics: tokscale - recorded v4.20.1 (2026-10-23), shipping the two 2026-10-20 submit-path fixes and adding Antigravity family re-attribution coverage; npm latest 4.20.1, trailing-month downloads 167,077 (2026-09-23 to 2026-10-22).
  • session-analytics: session-analytics - links, front matter, tags, matrix membership and sort, and index list all verified; no membership change, no matrix cell changes.
  • software-factory: fabro - volatile refresh: pushed 2026-10-25, nightly train at v0.396.0-nightly.0 (2026-10-24); no material change.
  • software-factory: fusion - rotation re-verified (all cited URLs re-fetched), npm latest still 0.77.0 (stable channel more than seven weeks behind), trailing-month downloads down to 2,094 (2026-10-22), still no HN story; no material change.
  • software-factory: loki-mode - train to v11.77.0 (2026-10-24), npm latest and next back in sync, loki github protect now requires the receipt status check on the default branch, Docker pulls 265,042 (2026-10-24).
  • software-factory: ouroboros - v0.59.8 and v0.59.9 patches (2026-10-23 and 2026-10-25, GitHub and PyPI, routine fixes); pushed 2026-10-25.
  • software-factory: superplane - RC train to v1.0.0-rc7 (2026-10-24) and fleet-manager v0.1.6 (2026-10-24); v0.30.0 still the newest stable platform tag; pushed 2026-10-25.
  • software-factory: zeroshot - v10.17.0 adds Active Memory (recalls lessons from costly repository mistakes and recurring graph procedures, default-on for local and Docker runs); v10.18.0 rebrands Zeroshot Cloud as Zeroshot Platform and its notes state the old $100 Cloud pricing is no longer stated anywhere.
  • software-factory: super-simple-software-factory - rotation re-verified (all cited URLs re-fetched), still a single commit on main, no push since 2026-08-04, no releases; no material change.
  • software-factory: software-factory-feature-matrix - Zeroshot self-improvement cell none to partial (Active Memory), delivery and pricing cells renamed to Zeroshot Platform (Cloud rebrand); all other cells re-verified unchanged.
  • software-factory: awesome-software-factories unchanged since its 2026-10-22 sweep (last README commit 2026-10-22, the already-resolved one), scan carried over, no new entrants.
  • ai-unified-process: recorded the 2026-10-24 release wave, the plugin model rebuilt from four combined stack plugins to seven layer plugins, aiup-core v3.1.0, and the Vaadin/jOOQ and Angular/JPA combined plugins deprecated ahead of removal in core 4.0.0 with /use-case-coverage moving into core.
  • openspec: recorded the v1.15.0 release (2026-10-24, standalone install without Node.js from release binaries, update –dry-run) and refreshed downloads to 2,753,900 last month.
  • mycelium: recorded the v0.329.x and v0.330.x releases (2026-10-23 to 24): review cadences became data with an adaptive corrections-audit cadence, and the canvas gate now refuses a subagent’s writes to the canvas and diamonds files on Claude Code.
  • arc-kit: rotation re-verification, recorded the overlay catalog’s growth to fourteen community plugins, the arckit-claude marketplace as the recommended Claude Code install, and the Open Badges certification program.
  • gsd: refreshed counts (npm 40,218 last month for gsd-core, 49,080 for the legacy package), pushed 2026-10-25 with v1.18.0 still newest.
  • cc-sdd: rotation re-verification of every cited source, refreshed npm to 26,372 last month.
  • pdd-kit: refreshed counts (npm 200 last month), pushed 2026-10-24 with v0.37.1 still newest.
  • spec-kitty: PyPI re-read byte-identical (3,251 last month, fifth identical read), pushed 2026-10-25 with the 4.x line still at rc7.
  • spec-driven-development: feature matrix, rebuilt the AI Unified Process cells for the layer-plugin wave and refreshed the adoption row (as of 2026-10-25).
  • bb: recorded desktop-v0.47.0 and desktop-v0.47.1 (October 23 and 24) as the new latest stable releases; added the README’s opt-out telemetry disclosure to Cautions (rotation re-verified).
  • antigravity: rotation re-verified, all cited URLs stand, prices and the models table unchanged.
  • delta: rotation re-verified, blog still ends October 15, pricing unchanged.
  • visual-studio-2026: recorded 18.11.1 (October 20) as the current servicing release above 18.11.0; prices unchanged (rotation re-verified).
  • codenomad: recorded stable v1.0.0 (October 24) shipping off the dev train: Missions, visual automation, native plugins and RPC, signed Tauri updates, OpenCode 2.1.2 minimum and 2.1.4 recommended.
  • openchamber: recorded v2.5.1 (October 23) as the new latest release: Gitea, Forgejo, and Gerrit git hosts, stacked pull requests, usage-limit fallback, GitHub Enterprise, Tailscale tunnels.
  • codeg: recorded v0.40.0 (October 24) as the new latest release: activity timeline, preserved tool-permission decisions, inline and side-by-side file-edit comparison, one-way database migration.
  • whiteboard: recorded v0.4.0 (October 23) as the new latest release: plan-mode boards with nested review stacks, human board editing, themes, TypeScript 7 LSP.
  • cate: volatile check, v2.0.5 still the latest stable with v2.1.0-beta.1 above it, pushed 2026-10-25.
  • nimbalyst: volatile check, v0.81.7 still tagged stable with the v0.82.0 to v0.82.3 pre-release train above it.
  • surfaces: link, front-matter, matrix-width, and sort-order checks clean across all 24 members; no membership change, matrix and category index untouched.
  • agentrq: recorded v0.13.3 and v0.13.4 (2026-10-24: README rebuilt around animated scenes with a Task Timeline feature row, Machine Info CPU tiles, enrollment starting the server, dependency security patches), counts refreshed (87 forks, 9 open issues and pull requests, pushed 2026-10-25)
  • backlog-md: recorded v1.55.3 (2026-10-23: native Git certificate and private HTTPS proxy support restored, closing the v1.55.2 caveat, Nix task fixes, Bun 1.4.4 rebuild), counts refreshed (82 open issues and pull requests, 85,015 npm downloads)
  • beads: recorded v1.3.2 shipping stable on 2026-10-24 after the rc train (fix-only patch), counts refreshed (1,519 open issues, pushed 2026-10-25)
  • epiq: recorded v1.15.0 to v1.15.2 (all 2026-10-23: dense cards with Wide/Dense switch, stalled-ticket clocks and live dots, click-to-filter, SSH repo-connect fix, time travel showing each ticket’s code at the needle with the position kept in the URL), counts refreshed (1 open issue and pull request, pushed 2026-10-24)
  • its-a-plan: recorded v1.8.0 (2026-10-23: Pro edition launched at $24/month or $240/year per instance with per-workspace SAML and SCIM, audit log, workspace dashboard, issue-event schedule triggers, agent analytics, AI budgets; three high advisories fixed, two of them agent-permission holes closed with breaking changes); Pricing rewritten, Price history added
  • kata: rotation re-verification with every cited URL re-fetched live; recorded v0.20.0 (published 2026-10-24: TUI Attention inbox, notify –re reply requests with wait –until reply, participant broadcasts with a 50-recipient cap, issue.scheduled_reached date events, SQLite upgrade corruption hardening, PostHog-free telemetry rework), counts refreshed (41 open issues and pull requests, pushed 2026-10-25)
  • operon: recorded 4.2.0 (2026-10-23: shared desktop Task Editor workspace with native Markdown and task fields at one level, layout and opening settings, Task Finder result behaviors), counts refreshed (14 open issues and pull requests, pushed 2026-10-23)
  • ordewell: counts refreshed (19 forks, 47 open issues and pull requests, pushed 2026-10-24, about 10,600 npm downloads); v0.10.12 still the newest release
  • pad: counts refreshed (25 forks, pushed 2026-10-25); v0.18.1 still the newest tag
  • ai-dev-tasks: rotation re-verification with every cited URL re-fetched live; still dormant with the last push at 2025-11-05, both prompt files unchanged, counts refreshed (10 open issues and pull requests, 1,707 forks, 217 watchers)
  • task-management: matrix release and activity cells refreshed for nine members; the It’s a Plan Pricing cell updated for the Pro edition
  • task-management: TomzxCode/agentic rechecked on its reopening trigger (independent adoption or third-party coverage); 1 star, 0 forks, zero HN hits, rejection stands
  • task-management: no credible dedicated awesome list surfaced for the category; it keeps riding the broad sources
  • bestvaluemodel: created and admitted on the 2026-09-25 reopen license, sustained operation demonstrated (45 commits and near-daily data diffs 2026-09-24 through 2026-10-25), matrix column and index entry added.
  • benchmark-radar: volatile re-verification, badge to 37,927 records over 90 days as of 2026-10-25, citation expanded to 15 report authors, share cards and a fourth outside contributor merged, Microsoft Clarity moved to opt-in consent with cross-tab withdrawal.
  • vectara-hallucination-leaderboard: volatile re-verification, README table stamp moved to October 22, 2026, leader unchanged (antgroup/finix_s1_32b at 1.8 percent).
  • livebench: volatile flag cleared, repository unchanged since its 2026-10-18 re-verification (pushed_at identical), no edits needed.
  • epoch-ai: rotation re-verification, team at 59 listed people, data stamps through Oct 25, pushes through Oct 24, consultations now visibly include Anthropic, xAI, and METR.
  • lmarena: rotation re-verification, October 2026 rebrand recorded (arena.ai), every cited URL re-fetched and resolving.
  • trackers-and-leaderboards: link, front-matter, as-of, matrix-width, and membership checks clean; matrix sort order and index list match membership.
  • trackers-and-leaderboards: dedicated-source re-check found no credible new awesome list (the BenchLM mirror again at 3 stars, plus sub-10-star lists); the 2026-10-23 standing decision holds.

(daily refresh, 3)
#

  • kimi-code: recorded Moonshot’s K3.1 Pro release (October 23, first K3.1-series model: flat-priced 1M-token context, $2/$10 per million tokens with cache reads at $0.20, Highspeed variant, weights due November 4) and the K2-series retirement clock over kimi-k2.7-code (new-user cutoff October 31, planned retirement by end of December 2026) as a caution; the quickstart’s Terminal-Bench cells still carried placeholder brackets, so no benchmark score is cited.
  • amp: rotation re-verification, every cited reference live; recorded the October 19 Linear issue assignment and October 22 Self-Hosted Orbs announcement (free for individuals and teams, Enterprise orb pricing from October 29), tier page re-verified unchanged.
  • codex: rotation re-verification, every cited reference live, and recorded rust-v0.164.0 as the latest stable (October 23): ten /presets slots, inline /model and /permissions arguments, Kitty/Sixel image previews.
  • nanocoder: rotation re-verification, every cited reference live, push refreshed to 2026-10-25, monthly npm downloads 9,411 (window September 23 to October 22), v1.31.0 still newest.
  • claude-code: rotation re-verification, every cited reference live, changelog head 2.1.310, and repaired the note’s Changes ordering (the 2026-10-16 bullet had been prepended above the creation bullet).
  • exo: rotation re-verification, every cited reference live (RSI.md at its post-restructure path), push refreshed to 2026-10-23, no releases and no new independent evaluation.
  • gptme: rotation re-verification, every cited reference live, v0.35.1 and 32,174 monthly PyPI downloads re-verified unchanged, contributor counts refreshed (TimeToBuildBob 2,626 against the author’s 1,949).
  • atomic-agent: recorded v0.6.11 (October 24) with desktop-v0.0.5: approvals-and-safety release showing multi-line commands line by line with shell-accurate quoting.
  • om-o: recorded the train reaching v5.1.46 (October 25): v5.1.45 incremental long-session reads (42 ms and 68 MB instead of 40 s and 1.8 GB on a 23,000-message session), v5.1.46 init-deep committed-mode pull-request-handoff fix.
  • qwen-code: recorded v0.26.1 as the latest stable (October 24): QQ bot per-group session isolation restored, opt-in daemon token persistence.
  • onecli: recorded v2.15.0 (October 21, GitHub skill library, browser steps, private preview links, custom connections, multi-replica runner) and v2.16.0 (October 22, agent-managed webhooks, failed-channel-reply owner alerts, chat image previews).
  • mirror: recorded the push moving to 2026-10-22 and the PyPI package reaching 0.1.1 (October 22, after 0.1.0 on October 16); the single GitHub tag mirror-agent-v0.1.8 stands.
  • command-code: recorded the npm train reaching 1.88.0 (October 23) after 1.84.0 (October 16), changelog count pinned to its 2026-10-17 reading.
  • harnesses: feature matrix checked against the K3.1 Pro and K2-retirement handoff, no feature cell bears (no matrix row covers model line-ups or retirement dates), membership and sort unchanged.
  • agent-manager: refreshed the release line to v0.43.2 (2026-10-24).
  • agent-orchestrator: recorded v0.14.6 (2026-10-24) as the newest stable on the nightly-to-stable train.
  • agent-swarm: recorded the v2 train through v2.5.4 (2026-10-25).
  • agentlas-os: recorded v1.2.69 (2026-10-25) on the release train.
  • agno: recorded v3.1.10 (2026-10-23) as the latest release on GitHub and PyPI.
  • apowerb: recorded PyPI at 0.2.97.
  • bernstein: recorded v3.23.0 (2026-10-24) as the latest release.
  • crewai: recorded 1.15.30 (2026-10-24) as the latest release on GitHub and PyPI.
  • emdash: recorded v1.2.9 (2026-10-23) on the release train.
  • foremerge: recorded v0.6.0 (2026-10-25) as the latest published release.
  • gascity: recorded v1.5.1 (2026-10-24) on the stable release line.
  • hive: recorded v1.0.11 (2026-10-19) and v1.0.12 (2026-10-24) on the release train.
  • langgraph: recorded release 1.2.17 (2026-10-24) on the release train.
  • multica: recorded v0.7.0 (2026-10-23) as the latest release.
  • oh-my-codex: recorded v0.21.36 (2026-10-25) as the latest release, npm in sync.
  • omnara: recorded v0.2.10 and v0.2.11 on both the omnarad and cluster release lines.
  • openrig: recorded v0.7.0 (2026-10-23) and v0.7.1 (2026-10-25) on the release train.
  • orca: recorded v1.4.240 (2026-10-25) as the latest desktop release.
  • paseo: recorded v0.13.0 (2026-10-24) on the release train.
  • praisonai: recorded v4.7.17 (2026-10-23) on the release train and praisonaiagents at 1.7.15 on PyPI.
  • ruflo: recorded the v3.60 through v3.63 train (2026-10-23 through 2026-10-25) on the release line.
  • sim: recorded v0.9.32 (2026-10-23) on the release train.
  • superset: recorded desktop-v1.40.0 (2026-10-23) on the release train.
  • t3code: recorded v0.1.2 (2026-10-22) as the newest stable.
  • a2ui: pushed 2026-10-25, tagged release train unchanged (the 2026-10-17 Python wave newest), spec hub table re-confirmed (v0.9.1 current, v1.0 candidate)
  • acp: pushed 2026-10-25 with the v2 draft’s RFD docs moving subagents, system prompt, and MCP-over-ACP to active; SDK train unchanged (Rust 3.5.0, TS 2.0.2), agents list 46
  • uhp: release train moved to v0.36.4 (2026-10-25) with a v0.37.0 release-candidate series (rc.1-rc.10) behind it; spec stays at its 2026-10-19 fifth version, suite at 2026.10.19.post3
  • anp: spec hub unreachable a second consecutive run, release train re-confirms v1.2 (2026-09-24), hub watch continues
  • agents-md: rotation re-verification, all cited URLs live, 60k-plus adoption figure and the Claude Code 2.1.277/2.1.281 changelog entries re-confirmed unchanged
  • llms-txt: rotation re-verification, all cited URLs live, spec still v2, Anthropic’s served file 857 to 899 lines, footprint scan re-confirmed
  • protocols: link, front-matter, matrix-width, matrix-sort, and index-membership checks clean across all 13 notes; no membership change
  • semble: recorded the fourth delivery surface (a semble skill installer option, 2026-10-24) and re-verified on rotation.
  • codealive: rotation re-verified, tier table, per-action rates, self-hosted page, and installer package unchanged, zero-HN footprint re-confirmed.
  • sourcegraph-code-context: rotation re-verified, $16K Enterprise floor and credits unchanged, newest-post claim refreshed for the October 19 and 21 use-case posts.
  • headroom: adoption claim moved to the trailing-month npm window (159,697 to 2026-10-22); 2026-10-24 main-branch fixes carried no release.
  • toon-format: adoption claim moved to the trailing-month npm window (6,723,407 to 2026-10-22).
  • context-engines: matrix Semble delivery cell gained the skill option and the TOON index one-liner moved to the trailing-month window; no membership change.
  • skills: agent-native - npm stable crossed to 0.3.51 with the nightly cadence continuing (@agent-native/scheduling at 0.2.45, a monorepo sub-package publish); the weekly window held at 4,392 (2026-10-16 to 10-22, unrolled) and the trailing-month figure added (54,848) under the zero-day-gap ruling.
  • skills: agent-native - catalog unchanged at 24 skills (no pushes since 2026-10-22), framework pushes remain monorepo app development (durable inbox admission, harness follow-up claims, Slack feedback tracking), and the zero-HN search re-run still returns zero hits.
  • skills: sepia - rotation re-verification re-fetched every cited URL and re-ran the zero-HN search (still only unrelated sepia stories); the note picked up the long-form journalism domain, the voice profiles beyond Hemingway, and the SlopShape replication (arXiv:2609.15369), whose current v3 reports 97.0 macro-F1 where the repo’s ledger quotes the pre-v3 98.0.
  • skills: agent-plugins - rotation re-verification re-fetched every cited URL, all claims held (spec still v1.0.0, the ten-field closed manifest and report-and-ignore rule, TSC and launch clients unchanged, repo still pushed 2026-09-28); no cells or figures moved.
  • skills: skills-feature-matrix - the reading prose’s Agent-Native npm figure moved to 0.3.51, catching up the note’s 10-24 crossing the matrix had missed; no cells moved.
  • skills: front matter, internal links, matrix column order and row widths, and index membership verified across all 15 notes; nothing broken, no membership change.
  • pageindex: PyPI SDK 0.4.7 (2026-10-24, 53 releases), pushed 2026-10-24, 151 open issues.
  • code-graph-rag: pushed 2026-10-25, open issues 538 to 526, PyPI 0.3.2 still latest.
  • docling: pushed 2026-10-25, 916 open issues, PyPI 2.142.1 still latest.
  • langchain: pushed 2026-10-25, 17,067 commits.
  • ragflow: pushed 2026-10-25, v1.0.0-rc2 still latest.
  • orama: rotation re-verified; adoption figure moved to the trailing-month window (5,194,510, 2026-09-23 to 2026-10-22) per the npm-gaps ruling, last complete weekly window 1,044,238, down about 27 percent; release stall and quiet main branch unchanged.
  • semantic-code-search: rotation re-verified, all claims unchanged, VS Code workspace-context page now dated 2026-10-21.
  • retrieval-feature-matrix: maintenance row refreshed to 2026-10-25, Orama cell moved to the monthly window; no membership change.
  • retrieval: both mapped sources unchanged since 2026-10-24 (awesome-rag-production last commit 2026-10-15, Awesome-RAG 2026-10-06), scans carried over, no entrants.

(daily refresh, 4)
#

  • deja: recorded v0.23.1 (MiniMax Code as the forty-third wired harness, project aliases under ~/.config/deja/projects, –sort date, cold-build peak halved to 4.7 GB on 40,000 sessions, redaction CPU cut 17x, quadratic look-alike-search fix); harness count updated in note, matrix cells, and index line.
  • claude-mem: recorded v15.1.0 (React My Work viewer on live worker data, per-prompt semantic injection in server mode, opt-in loopback HTTP MCP transport with per-session workspaces, broader secret redaction), npm in sync at 15.1.0; trailing-month downloads 63,865.
  • agentmemory: native plugin surface grew to Oh My Pi (connect omp installs native recall and capture, offline replay, and the 17 skills); push date refreshed.
  • operator-memory: OpenCode V2 adapter made the default install with V1 frozen to legacy (helper commit 2026-10-20).
  • gbrain: release train continued to v0.60.444.0 (six October 25 releases, repair steps now print one commit-push command; no retrieval behavior change).
  • memoryfields: rotation re-verification corrected the tooling release record (memoryfield-tool v0.0.1 and v0.5.0, both August 2026, where the 2026-10-16 pass read none); spec holds at version 0.3.
  • memory: volatile passes on mem0 (BYOK docs rewrite), memU (branch-only pushes), MemOS (certifi security pin), Supermemory (MCP key validation), and Screenpipe (routine app-v2.7.116) found no note-material changes.
  • executions: aeon - refreshed repository counters (1,491 commits, pushed 2026-10-25); nothing else moved
  • executions: github-agentic-workflows - release train to v0.93.4 (2026-10-25), stable-marked with v0.93.3’s prerelease flag cleared; forks 601
  • executions: claude-code-hooks - rotation re-verify: replaced the stale v2.1.214 matcher claim with current version gates, added the plugin mods function-hook surface, 33 events and five handler types re-confirmed
  • executions: claude-code-routines - rotation re-verify: recorded the scheduled-tasks rename, added issue events to the GitHub trigger, re-grounded the autonomy caution on auto-mode wording, replaced the stale v2.1.213 reference
  • executions: claude-code-action - corrected the star-history chart fallback markup
  • executions: codex-action - corrected the star-history chart fallback markup
  • executions: run-gemini-cli - corrected the star-history chart fallback markup
  • executions: airut - corrected the star-history chart fallback markup
  • executions: n8n - volatile pass re-verified unchanged (61.2k forks, pushed 2026-10-25); no edit
  • executions: rivet-agents - volatile pass re-verified unchanged (134 forks, 86 open issues, pushed 2026-10-24); no edit
  • executions: executions - no credible dedicated awesome-list source found for the category (search returned one 2-star stale list)
  • jeff: board credit corrected, the JevBench Jeff-1.0-Large row is jgeuter’s unrelated Gemma-4-31B fine-tune (CC BY-NC-4.0, created 2026-10-08), so the family is back to zero third-party readings; rotation re-verified, all cited URLs holding
  • jevbench: board v1.7.48 runs method v1.7 (the announced 10-second rule), H2O-Lightning-31B v1.0.7 takes the capability lead at 84.3, Chon-1.0-12B falls from 85.3 first to official rank 30 at 70.1, main board 220 ranked, API board 47 rows with Super Decider first at 78.5 and the Jev 1.14.0 references unranked (75.0 capability, 70.2 seventeenth)
  • atomic-agents: pushed 2026-10-24, generic-schema resolution and Instructor-callback fixes without a release, PyPI still 2.15.0
  • baml: nightly train through the 20261024 build (published 2026-10-25), still no stable 0.22.3, pushed 2026-10-25, npm month unchanged at 972,279
  • laya: 0.4.15 package (2026-10-24) recorded, pushed 2026-10-24, PyPI at 0.4.15
  • tool-eval-bench: pushed 2026-10-25 on a dependency bump, no release since v2.8.0
  • cua-s1: rotation re-verified, component README, model card, checkpoint metrics, thread, and likes all holding, the host’s cua-sdk-v0.9.0 and cua-spaces-v0.15.0 releases not touching the component
  • guardrails: rotation re-verified, all cited URLs holding, PyPI and download figures unchanged, pushed 2026-10-25 on branch activity with the default branch still ending at the August hub-retirement merge
  • marvin: rotation re-verified, nothing changed
  • hybrid-execution: matrix: JevBench cells and board paragraph moved to board v1.7.48 under method v1.7, maintenance cells moved for the five push-signal members, thesis count fixed to thirty-one
  • tool-eval-bench: API-surface cell gained TandemLLM, matching the note’s 2026-10-24 recording
  • laya: API-surface cell moved to 0.4.15, having lagged at 0.4.13 while the maintenance cell said 0.4.14
  • code-review: ellipsis - rotation re-verification recorded the free-for-individuals on-ramp leaving the homepage and pricing FAQ (the only free path now is case-by-case open-source support by email), the hero leading Software Factories, and the CLI push to October 23; all metered prices unchanged.
  • code-review: sourcery - rotation re-verification recorded the Perfectionist homepage positioning with the finds-more-issues claim, auto-approve for low-risk PRs, the SOC 2 Type II statement, and the docs’ diff-character budget and five-re-review cap; prices and release state unchanged.
  • code-review: kodus - volatile re-verification moved the push to October 25 (no release after selfhosted-2.2.6) and recorded the Kody Pilot Program discount banner (30 to 100 percent off for startups, open source, weaker-currency countries); tiers unchanged.
  • code-review: qodo - volatile re-check: pushed late October 24, v0.48.0 still latest.
  • code-review: roborev - volatile re-check: pushed October 25, v0.73.0 still latest.
  • code-review: code-review-feature-matrix - recorded Ellipsis’s on-ramp removal in the pricing cell and the Ellipsis paragraph; no other cell changes, star figures untouched.
  • code-review: code-review - dedicated-list check found no new credible list; the kodustech/awesome-ai-code-review stays unwired per the 2026-10-24 ruling and q21’s evidence base is unchanged.
  • people-and-publications: jesse-vincent - Added his 2026-10-11 “Picks and Shovels” post (Claude Opus 5.5 full-autonomy game build for picksandshovels.org) to Status and References; blog count to 546
  • people-and-publications: kent-beck - Added “Decompose Run-On Tests” (2026-10-13) and the paid “Take My Toil, Leave Me My Struggle” (2026-10-20) to Status and References
  • people-and-publications: nathan-lambert - Added “Meta shakes up the frontiers, open and closed” (2026-10-17) and the paid “Latest open artifacts (#25)” (2026-10-20) to Status and References; subscribers over 86,000
  • people-and-publications: the-pragmatic-engineer - Added October’s AI-beat issues to Status and References: Uber’s AI infra (10-14), paid AI-spend ROI analysis across 27 companies (10-20), security Pulse (10-22)
  • people-and-publications: Moved the Nathan Lambert matrix cadence cell to one to three a week (four essays September 19 to October 20)
  • kimi-code: Moonshot launched Kimi K3.1 Pro (2026-10-23, $2/$10 API with a $4/$20 highspeed variant, cache reads $0.20, 1M context, open weights promised by Nov 4) and put the K2 series on a retirement clock (closed to new platform users 2026-10-31, full retirement by end of December 2026); the entry tiers’ K2.7 Code model flagged in Cautions.
  • kimi-code: kimi-k3’s OpenRouter listing held at $0.239/$10.80 through the eighteenth recorded endpoint move (input floor to Relace $0.19/$8.94, Wafer’s cheapest input to $0.20/$16.00, Morph’s input leg to $0.90, roster 29), and moonshotai/kimi-k3.1-pro appeared in the catalog at $2/$10 (cache reads $0.20).
  • kimi-code: both membership ladders re-verified unchanged after the K3.1 Pro launch (Plus/Pro/Max/Ultra $19-$199; China Go/Plus/Pro/Max/Ultra ¥49-¥1,399); K3.1 Pro not yet on either ladder.
  • openrouter: fee state re-verified unchanged (5.5% Standard with $0.80 minimum, 8% Business, BYOK free to $25,000).
  • litellm: v1.106.3 went stable 2026-10-24 with v1.105.4, v1.104.6, and v1.103.8 patched the same day; PyPI latest at 1.106.3.
  • ollama: v0.40.11 went stable 2026-10-24, promoting the rc with nothing past it.
  • plano: 0.4.38 and 0.4.39 released 2026-10-24, the first releases since 2026-09-28 (Envoy v1.37.8 bump for CVE-2026-111497, agent-text extraction fixes).
  • bifrost: repo pushed 2026-10-25; releases unchanged since the 2026-10-20 component patches.
  • experiential: PyPI at 0.7.220 (223 releases), repo pushed 2026-10-25, pricing page unchanged from its 2026-10-24 read.
  • llm-gateway: repo pushed 2026-10-25, v1.22.0 still latest, DevPass state unchanged.
  • magnitude: repo pushed 2026-10-25, @magnitudedev/[email protected] still latest.
  • model-access: rotation re-verification of vercel-ai-gateway, claude-plans, glm-coding-plan, and qwen-coding-plan found every cited URL live and every stated price unchanged.
  • cross-category: held bdots (fayazara/bdots, 156 stars as of 2026-10-25, created 2026-09-30) on the assistant-runtimes absorbed runtime family, the cloudflare-os slot, reopen on q#13, OpenClaw-class scale, or independent coverage

(daily refresh, 5)
#

  • pnpm: the day’s one new star (pnpm/pnpm, starred 2026-10-25T02:35:56Z, 36,918 stars) judged out of scope as a general-purpose package manager with no agent story; stars checkpoint advanced to 2026-10-25T02:35:56Z
  • ai-for-developers Code Review section resolved in full: AICodeSlopMonitor, What The Diff, VibeDoctor, VibeAudit, MergeFix, and Git AutoReview rejected as closed or site-only products with no verifiable repository and no independent coverage; MealMacroAI rejected as off-topic list spam; UIZZE rejected under the skills standing exclusion (a skill pack); the pre-agentic static-analysis and security-scan tier (DeepSource, Codacy, Qodana, Semgrep, CodeQL, Snyk, Pixee, Amazon CodeGuru, Embold, Veracode, PullRequest) and the coding agent Refact.ai rejected as outside the category’s AI-review kind; SynthScan (2 stars), twoperson (1 star), and the same-author provider-clone pair zai-code-review/minimax-code-review (43/21 stars) rejected on the thin-tail profile; Relay (4 stars, dormant since May) rejected as memory-kind and below the memory floor; sweepai/sweep rejected on kind (issue-to-PR generation, not review) with dormancy since 2025-09-18 noted; argot (51 stars) and OrcaCode Review (179 stars) rejected on the thin-record profile, below every category admission with zero coverage, reopen on traction or independent coverage; the Swarm Orchestrator entry already resolved by the 2026-10-24 orchestration admission
  • star-history chart fallback markup corrected across forty-seven notes in twelve categories (the img fallback carried theme=dark; it now points at the default chart), following the executions worker’s four fixes
  • kimi-code (harnesses): Usage sentence refreshed to the eighteenth endpoint-move reading (input floor $0.19/$8.94 on Relace, Wafer $0.20/$16.00, Morph $0.90, roster 29, as of 2026-10-25) and the matching Changes bullet appended, syncing the move-by-move tracker with the model-access note
  • claude-replay: adoption claim converted to the trailing-month npm window (1,698 to 2026-10-22) per the 2026-10-22 ruling, reference moved to the range endpoint
  • trackers-and-leaderboards-feature-matrix: section-index one-liner moved to thirteen members with bestvaluemodel the newest, verified 2026-10-25

(daily refresh, 6)
#

  • assistant-runtimes: anything-llm pushed 2026-10-24 with 340 open issues, v1.18.0 still latest and the Cloud tiers (Basic $50, Pro $99) re-checked unchanged
  • assistant-runtimes: astrbot recorded v4.29.3 (2026-10-24, message tracing, redacted diagnostics export, multilingual installer with China mirrors, backup and Gemini fixes, v4.29.2 tag skipped), 1,016 open issues and PRs
  • assistant-runtimes: eigent recorded the pricing page pausing free registration credits for new accounts during a misuse cleanup (Price history row), paid tiers and annual options confirmed unchanged in the page source
  • assistant-runtimes: hermes corrected the backend list to the README’s current framing (six built-in terminal backends, Singularity and Apptainer as plugin backends), Portal tiers re-checked unchanged, 50,915 open issues
  • assistant-runtimes: nanobot rotation re-verification: every cited URL re-fetched live, v0.3.5 (2026-09-15) still latest with the Alpha classification current and the two-name maintainers table intact
  • assistant-runtimes: nanoclaw pushed 2026-10-24 with 1,022 open issues, v2026.10.1 still at rc.2 with stable at v2026.10.0
  • assistant-runtimes: octop recorded v1.0.3b1 (2026-10-25, a default expert at user creation, permissions split per page tab) opening the v1.0.3 line on GitHub and PyPI the same day
  • assistant-runtimes: open-webui pushed 2026-10-25 with 166 open issues, v0.12.3 still latest
  • assistant-runtimes: openclaw recorded v2026.10.7 (2026-10-24, replay-safe session recovery, Gemini 4 Argon support) taking the latest-stable spot beside the v2026.9.34 extended-stable gateway patch, open count falling to 7,814
  • assistant-runtimes: openwork recorded v0.18.61 and v0.18.62 (both 2026-10-23, pop-out computer view and floating browser windows, then OneDrive and SharePoint file support), pricing re-checked unchanged (seventeenth-churn state holds)
  • assistant-runtimes: picoclaw rotation re-verification: every cited URL re-fetched live, v0.3.1 (2026-07-03) still latest, both README banners intact, official site and docs serving normally
  • assistant-runtimes: private-gpt pushed 2026-10-23 with 11 open issues, v1.0.1 still latest
  • assistant-runtimes: zeroclaw pushed 2026-10-25 with 1,042 open issues, v0.8.6 still latest
  • assistant-runtimes: matrix re-verification moved seven status cells and matching prose counts, stars row left at its 2026-10-14 counts
  • aris: volatile refresh recording the keenable-search and firecrawl-search skills (mainline count 85 to 87) and the October 24 OpenAI model-defaults refresh; push date to October 24.
  • autocontext: volatile refresh moving the off-main push date to October 24; main unchanged since September 29.
  • autosci: recorded the repository’s October transfer from skyllwt to the PKU-DAIR organization (the Peking University DAIR lab of senior author Bin Cui), moved the star-history chart to the canonical path, and noted the README’s new three-paper works-produced section.
  • kapso: volatile refresh moving the off-main push date to October 24; v0.7.10 still the latest release.
  • local-deep-research: volatile refresh moving the daily-cadence push date to October 25; the release train still sits at v1.10.7 while main runs 2.0.0.
  • openresearch: volatile refresh moving the push date to October 24; v0.2.21 still the latest release.
  • deepanalyze: rotation re-verification refreshing the Hugging Face usage footprint (144 trailing-month downloads, down from 211; 93 likes) and adding SkillAdam to the ecosystem list.
  • ai-researcher: rotation re-verification confirming dormancy, the missing license file, and the no-threads HN footprint; the NeurIPS 2025 acceptance sharpened to Spotlight.
  • autoresearchclaw: rotation re-verification correcting the paper’s author list to 36 (arXiv v2, led by Jiaqi Liu and Shi Qiu) and confirming dormancy (push still 2026-08-19, release still v0.5.0).
  • agent-laboratory: rotation re-verification, all cited URLs live and every fact unchanged.
  • dexter: rotation re-verification, all cited URLs live and every fact unchanged (still no license file, push still 2026-09-23, release still v1.1.0).
  • automated-research-feature-matrix: ARIS skill cell 85 to 87 and the AutoSci operator cell to PKU-DAIR.
  • automated-research: links, front matter, section order, and as-of dates verified across all 32 notes; matrix and index match membership and sort.
  • automated-research: the webfuse-com/awesome-autoresearch source unchanged since its last check (pushed 2026-09-21); its scan carries over with no additions.
  • agent-control: release train moved to v8.15.0 (2026-10-23, Galileo Decision Models evaluator), agent-control-sdk 8.15.0 on PyPI, npm holding at 3.4.0
  • agentfield: train left release candidates, v0.1.143 (2026-10-24) the first stable tag with PyPI publishing 0.1.143 the same day
  • cc-safety-net: npm train to 2.6.19 (2026-10-24), closing the recursive root/home delete gap for Python, Perl, Node, and Ruby interpreter code
  • hol-guard: release train to v3.56.0 (2026-10-24, PyPI in step), coverage contract re-checked still unrenewed sixteen days past expiry, still pinned to 3.0.1
  • paperclip: v2026.1023.0 (2026-10-23, 200 commits, private tasks and projects) ends the eleven-day release pause, backlog regrew to 6,523
  • tracekit: release train to v1.4.2 (2026-10-24, PyPI in step), the sixth release in nine days, open issues up to 22
  • vestige: release train to v5.0.1 (2026-10-23), Retroactive Salience Backfill returns to the Strata memory path
  • veto: site’s three fixed-price policy packs ($750 to $2,000) recorded in Pricing and Price history, plan prices and repository unchanged
  • control-planes-feature-matrix: volatile cells refreshed across eleven members, no membership change
  • control-planes: awesome-ai-governance scan carried over, its only new commit a removal (the Nobulex entry, PR #166), no new entrants
  • evaluation-review: crit - Recorded the v0.23.1 release (2026-10-23) and refreshed the activity line, in rotation with every cited URL re-fetched and live
  • evaluation-review: pr-lens - Rotation re-verification recorded the hosted GitHub App’s diagrams-plus-code-review findings layer (drawing-only Action, CLI, and skill in the MIT core) and refreshed the activity line; every cited URL re-fetched and live
  • evaluation-review: langfuse - Recorded the v4.69.0 (2026-10-23) and v4.70.0 (2026-10-24) releases
  • evaluation-review: plannotator - Recorded the v0.28.29 (2026-10-23) and v0.28.30 (2026-10-24) releases
  • evaluation-review: hunk - Recorded the v0.24.1 stable release (2026-10-25)
  • evaluation-review: verbatim-citation-gate - Recorded the v0.3.1 release (2026-10-24) shipping the three gate fixes previously unreleased on main; PyPI still absent
  • evaluation-review: phoenix - Recorded the arize-phoenix 20.26.0 release (2026-10-23)
  • evaluation-review: reviewbench - Recorded the Laude RevEngine Prime rename to Laude RevEngine (PR 98, 2026-10-24); onboarding count stands at seventeen
  • evaluation-review: harnesstax - Recorded three more AgentBRANE dashboard-build republications (2026-10-25); profiling traces still unpublished
  • evaluation-review: workshop - Recorded two more unmerged outside pull requests (2026-10-22) alongside no default-branch commit and no release since 2026-08-22
  • evaluation-review: matrix maturity cells refreshed (Crit v0.23.1, Hunk v0.24.1, Langfuse v4.70.0, Phoenix v20.26.0, Verbatim Citation Gate v0.3.1) and the PR Lens Kind and Judge cells updated for the hosted-App findings layer; FrontierHarness Eval, Crit, and PR Lens fully re-verified in rotation, every cited URL live
  • evaluation-review: mapped source anhermon/awesome-agent-observability unchanged since 2026-10-23, scan carried over with no new candidates
  • microsandbox: recorded v0.7.10 (2026-10-23, anonymous registry auth in the Go SDK, the published smoltcp fork, custom filesystem backends, SDK-writable secret policy) and v0.7.11 (2026-10-24, SDK-exposed HTTP CONNECT proxy authentication, terminal-hang fix, config overrides preserved, interrupted-creation cleanup recovery)
  • yolobox: recorded v0.19.11 (2026-10-23, reset and uninstall volume handling on the Apple container runtime), the release train moving to seventy
  • brig: recorded the channel-experimental 0.4.2 line opening (builds 2026-10-23 and 2026-10-24), v0.4.1 still the newest tagged release
  • aicontainer: rotation re-verification, trailing-month downloads to 516, guardrail-hook scope corrected to the README, main branch quiet since the 0.12.1 release with renovate-only pushes since
  • fence: rotation re-verification, every cited URL alive, v0.1.67 still the newest of thirty releases
  • greywall: rotation re-verification, every cited URL alive, quiet window extended to seventy-three days without a commit to main
  • artifact-fs: rotation re-verification, every cited URL alive, pricing page unchanged since 2026-10-01 (billing began 2026-10-14 as recorded), public beta still unopened, main branch still quiet since 2026-08-12
  • overalls: commit count corrected to 1,746 on the default branch (the 30 recorded at creation miscounted a single default page), over 200 commits since 2026-10-21 in the launch-prep train, launch still set for 3 November 2026
  • agent-sandbox: push advanced to 2026-10-24, v1.0.8 remains the latest release
  • e2b: push advanced to 2026-10-24 on both repos, PyPI still 2.59.0
  • nemoclaw: push advanced to 2026-10-25, still prerelease-only and alpha
  • nono: push advanced to 2026-10-23, v0.80.0 still the latest release
  • nvx: push and newest prerelease advanced to 2026-10-25 on the daily dev cadence
  • opensandbox: push advanced to 2026-10-25, release-1.2.0-rc.2 still the latest
  • brood-box: push advanced to 2026-10-24 on branch activity, v0.0.25 still the latest release
  • sandboxing: matrix maturity cells refreshed (microsandbox v0.7.11, yolobox v0.19.11, aicontainer downloads 516), star figures untouched
  • sandboxing: arjan/awesome-agent-sandboxes unchanged since its 2026-10-23 check, scan carried over