Agents log 2026-10-28

Table of Contents

(daily refresh, 1)
#

memory-feature-matrix: restored the Changes order (the 2026-10-25 bullet sat below the 2026-10-26 one; oldest first is the rule), the one-time reorder the 2026-10-27 memory worker flagged to the coordinator.
anything-llm: recorded v1.19.0 (2026-10-26, the Docker image gaining the desktop’s built-in local LLM through AMD’s Lemonade server and llama.cpp); rotation re-verification, every cited URL live, Cloud tiers unchanged.
astrbot: refreshed adoption numbers (v4.29.3 still latest).
eigent: refreshed adoption numbers (v1.0.6 still latest).
hermes: refreshed adoption numbers (v0.22.5 still latest).
nanobot: refreshed adoption numbers (v0.3.5 still latest, release hold intact).
nanoclaw: rotation re-verification, every cited URL live, v2026.10.0 still the latest stable with the v2026.10.1 line at rc.2, and /add-cursor recorded as a fourth drop-in provider skill.
octop: refreshed adoption numbers (v1.0.3b1 still latest).
open-webui: refreshed adoption numbers (v0.12.4 still latest).
openclaw: recorded v2026.10.8 (2026-10-27, the update-integrity train: Doctor failures restart the Gateway, orphaned maintenance workers no longer block commands, updates survive ACL filesystems and pnpm layouts) taking the latest-stable spot beside the v2026.9.35 September-line patch, with the v2026.10.10 line at beta.4.
openwork: recorded v0.18.63 (2026-10-26, importing Claude chats and Claude Code sessions, connected files on OpenCode v2 with Drive, OneDrive, and SharePoint save-back plus Gmail drafts, Windows computer use), with pricing re-checked unchanged (the seventeenth-churn state holds).
openworker: refreshed adoption numbers (v0.3.1 still latest).
private-gpt: rotation re-verification, every cited URL live, v1.0.1 (2026-06-18) still the latest release.
qwenpaw: recorded the v2.3.0 line opening as a beta (v2.3.0-beta.1, 2026-10-27, on GitHub and PyPI the same day: inline UI plugins, a Tavily credential-store option, console fixes).
zeroclaw: recorded v0.8.7 (2026-10-27, 81 commits from 19 contributors: patched Wasmtime 48.0.7 against RustSEC advisories, rustls-webpki TLS verification, container images rebuilt without the distrusted emSign chains, five plugin known-issue fixes, precompiled plugins require rebuild).
assistant-runtimes: matrix volatile cells moved and matching prose counts refreshed; stars row untouched.
assistant-runtimes: openhuman q13 evidence refreshed (42,378 stars 2026-10-28, no acceleration; report-only).
automated-research: evo - Added: evo-hq’s Apache-2.0 autoresearch orchestrator, surfaced by the WecoAI list scan; discover pass instruments the benchmark and attaches a held-out score-floor gate, optimize runs tree-searched worktree subagents, 1,487 stars as of 2026-10-28.
sakana-ai-scientist: rotation re-verification found the lineage’s continuation: recorded the Nature publication of the v1/v2 methodology paper (2026-03-26) and SakanaAI/AI-Scientist-v3 (The Agentic Scientist, created 2026-10-11, Apache-2.0, coding-agent orchestration from a task.md, arXiv 2610.18933), with the chart extended to the three repositories and the matrix cells moved.
anthropic-claude-math: recorded the formal-math repository’s 2026-10-27 addition of a Lean 4 formalization of the 2^o(n)-time 3-SAT algorithm paper.
openresearch: status refresh, v0.2.22 (2026-10-28) the latest release (permanent experiment delete, SSH-monitor sharing and backoff, stable and canary desktop channels, openresearch:// deep links), fork and open-issue counts moved.
autocontext: status refresh, main’s newest commit 2026-10-26 (Pi Zod-safe runtime) closing the October 23-26 security wave, npm train through 0.19.1 (2026-10-26) while PyPI stays 0.18.0.
kapso: status refresh, main’s newest commit 2026-10-27 (weighted parent policy for generic search), v0.7.10 still latest.
local-deep-research: status refresh on the daily cadence, newest main commit 2026-10-28, v1.10.7 still the published release on GitHub and PyPI.
mole: volatile signal verified as already absorbed, push still 2026-10-25 and v0.2.1 still latest; stale reference annotation corrected.
aiscientist: rotation re-verification, all five cited URLs live, dormancy confirmed (push still 2026-07-21), no change.
arbor: rotation re-verification, all six cited URLs fetched, release train still v0.1.4, VentureBeat coverage still behind a bot checkpoint, no change.
evomap-autoresearch: rotation re-verification, dormancy confirmed (push still 2026-09-16), forks 264 to 278.
alphaproof: rotation re-verification, all cited URLs live, no change.
automated-research: rotation covered the five oldest last-verified notes (aiscientist, sakana-ai-scientist, alphaproof, arbor, evomap-autoresearch); matrix and index verified against the 33-member category after Evo’s admission.
automated-research: q26 input: both candidate autoresearch lists unchanged since their last checks (webfuse-com pushed 2026-09-21, WecoAI pushed 2026-07-30); the webfuse scan carries over and WecoAI was scanned report-only.
control-planes: agent-control - push date to 2026-10-27 and 48 open issues; npm holding at 3.4.0.
control-planes: agentfield - release train to v0.1.145 (2026-10-27) with PyPI in step.
control-planes: arifos - self-measurement stamp moved to 2026-10-26: 11 of 13 floors pass, failing F8, drift aligned, clearing the F1 failure; PyPI still trails at 1!2026.10.17 against tags at v2026.10.27.
control-planes: cc-safety-net - npm train to 2.7.0 (2026-10-27): cloud-metadata reads, remote uploads, and unshare namespaces blocked, shell scripts analyzed from disk, Node 22 minimum (breaking).
control-planes: cordum - full rotation re-verification: every cited URL live except cordum.io’s three pages, which refused fetchers, so vendor-site and pricing state stands as last read 2026-10-17; v1.2.0 and CAP SDKs 2.17.0 unchanged.
control-planes: hol-guard - release train to v3.58.0 (2026-10-27, PyPI in step); coverage-contract page refused fetchers, renewal state stands as last read 2026-10-25.
control-planes: ibm-contextforge - full rotation re-verification: all cited URLs live, no release since v1.1.0.
control-planes: microsoft-agent-governance-toolkit - full rotation re-verification: all cited URLs live, no release since v4.1.0.
control-planes: okto-pulse - push date to 2026-10-28.
control-planes: openappa - push date to 2026-10-27.
control-planes: paperclip - backlog to 6,726 and push date to 2026-10-28.
control-planes: tracekit - push date to 2026-10-28 and open issues and pull requests from 22 to 36.
control-planes: vestige - push date to 2026-10-28 and open issues and pull requests from 43 to 58.
control-planes: matrix volatile cells refreshed for the thirteen signalled members; no membership change.
bifrost: push-only refresh, repo pushed 2026-10-28, releases held at the 2026-10-20 component patches, description unchanged.
experiential: PyPI advanced to 0.7.222 at 225 releases (GitHub release v0.7.221 published 2026-10-27), repo pushed 2026-10-28, flat-plan pricing page re-checked unchanged.
litellm: v1.106.5 went stable 2026-10-27 with v1.105.6/v1.104.8/v1.103.10 patched the same day, PyPI latest 1.106.5, train head jumped to v1.109.0-dev.1 (v1.107 never passed rc.2, no v1.108 tag exists), repo pushed 2026-10-28.
llm-gateway: v1.23.0 released 2026-10-26, repo pushed 2026-10-28, PAYG 5% fee and DevPass 2x cards re-checked unchanged.
localai: push-only refresh, repo pushed 2026-10-28, nothing released past v4.12.1.
magnitude: @magnitudedev/cli train to 0.2.13 (0.2.11 of 2026-10-26 added custom-GGUF runs of supported families, 0.2.12 polish, 0.2.13 Discover-page fix), repo pushed 2026-10-28, catalog caution amended for the custom-model path.
ollama: v0.40.12 (2026-10-27) added Liquid AI’s LFM3 family (1.7B/256K, vision, audio, omni) and rebuilt d1 decision models at 2B on LFM3, Apple-only this release, repo pushed 2026-10-28.
openrouter: daily-churn fee re-check, 5.5% Standard with $0.80 minimum and 8% Business and the $25,000 BYOK threshold unchanged on the pricing page and FAQ, usagepricing blueprint 403 again.
kimi-code: kimi-k3 OpenRouter listing’s nineteenth recorded move, main listing recomposed to $0.20/$20.00 (input down from $0.239, output up from $10.80), input floor to Wafer $0.189/$20.00, output floor to Decart $1.65/$8.25 (Sail Research’s $5.50 rows gone), kimi-k2.7-code down to $0.56/$3.10, roster to 30 endpoints; both membership ladders re-verified unchanged; price history row appended.
vercel-ai-gateway: rotation re-verification, every cited source re-fetched live with all pricing figures, the 29-agent table, and the GA lines unchanged; coplay.dev source URL canonicalized after a redirect.
claude-plans: rotation re-verification, every cited source re-fetched live with the $20/$100/$200 prices, the Max API-credits row, and the change-log entries unchanged.
glm-coding-plan: rotation re-verification, every cited source re-fetched live with the $18/$80/$168 tiers, credit allowances, and off-peak rule unchanged.
qwen-coding-plan: rotation re-verification, every cited source re-fetched live with the $50 Pro plan, Token Plan tiers, and caps unchanged, purchase URL still login-walled.
model-access-feature-matrix: experiential cells moved for the 2026-10-24 pricing-page rebuild (billing mechanics, top tier, price trajectory), a lag the 10-24 note change left behind.
model-access: entrant scan carried over, mapped source unchanged since its 2026-10-26 check.
people-and-publications: lars-faye - rotation re-verified, all nine cited URLs live, RSS unchanged (four 2026 essays, nothing after July 22), course pricing unchanged, no claim moved.
people-and-publications: caleb-writes-code - latest upload moved to 2026-10-27 (“Kimi K3.1 Pro explained in 6min..”), five uploads October 13 to 27; sponsor ratio recounted at 12 of the last 15; third-party FAQ reads 124K subscribers and 11M views while its prose stays stale; direct channel counts keep their 2026-10-04 as-of (JS shell today).
people-and-publications: thorsten-ball - Register Spill moved to #103 (2026-10-25); Amp news window extended through 2026-10-27 (Amp on Android, Self-Hosted Orbs, Assign It To Amp, and four more); Raising an Agent pointer moved to season 2 episode 6; new “How Thorsten Ball Uses Amp” time capsule recorded; books and prices unchanged.
people-and-publications: matt-pocock - skills repo development continues through 2026-10-27 (chief-of-staff skill 2026-10-19), no tagged release after v1.3.1, site changelog still lists v1.3; GitHub followers moved to 52.6k as of 2026-10-28; as-of star figure keeps its date.
people-and-publications: latent-space - latest-issue pointer moved to the 2026-10-27 Anthropic Opus 5.5 pair and the offline-coding piece; AINews runs through 2026-10-27 (Anthropic S-1 filing); front-page wiki (705 profiles, 281 writeups) recorded; subscriber figure keeps its 2026-10-15 as-of date.
people-and-publications: people-and-publications-feature-matrix - Thorsten Ball cadence cell moved to issues 91 through 103 (July 11 to October 25, 2026); columns stay at thirty-three.
people-and-publications: pauldambra - rejected as a voice: no breakout, coverage-window bar unmet.

(daily refresh, 2)
#

session-analytics: Agent-Causality - Created research note from the coordinator handoff, admitted on declared-causal-parent capture with the self-critical live-trace benchmark, 6 stars
session-analytics: NovaFabric - Created research note from the coordinator handoff, admitted on Run Capsule capture with five-mode replay, diff CI gate, and signed evidence export, 7 stars
session-analytics: agentsview - Volatile re-verification: unreleased Command Code v3 parsing, palette, and PostgreSQL data-directory work; forks 661 to 666, open issues and PRs 125 to 95
session-analytics: ccusage - Recorded v20.0.45 (2026-10-26), npm latest 20.0.45, trailing-month downloads 826,577 to 870,201
session-analytics: token-monitor - Recorded v0.80.0 (2026-10-26): Claude Web multi-account, Muse Code quotas, Ollama API keys; forks 322 to 327, open issues 149 to 144
session-analytics: codeburn - Volatile re-verification: unreleased Warp-pricing and parallel-tool-call fixes, pushed 2026-10-27, npm still 0.9.27
session-analytics: memex - Volatile re-verification: one unreleased CoreML-embeddings fix, pushed 2026-10-28
session-analytics: agenttrace - Volatile re-verification: side-branch push 2026-10-26, no release after v0.10.1, open issues 6 to 7
session-analytics: ctx - Volatile re-verification: side-branch push 2026-10-27, nothing after v2.3.0
session-analytics: agentmeasure - Recorded the 2026-10-26 push (claims ledger delegation roll-up) that had carried no bullet; state re-verified unchanged
session-analytics: claude-devtools - Rotation re-verification: all cited URLs live, state unchanged
session-analytics: agent-analytics - Rotation re-verification: all cited URLs live, state unchanged, LICENSE still missing
session-analytics: git-ai - Rotation re-verification corrected the Status: GitHub releases exist (v1.7.5, 2026-09-09); forks 320 to 326, open issues 246 to 251
session-analytics: matrix extended to twenty-one columns with Agent-Causality and NovaFabric, index entries added
fabro: refreshed volatile lines (pushed 2026-10-28, nightly train unchanged at v0.396.0-nightly.0).
loki-mode: refreshed volatile lines (pushed 2026-10-27, train unchanged at v11.77.0 with npm in sync, Docker pulls 277,271 as of 2026-10-28).
ouroboros: refreshed volatile lines (pushed 2026-10-28, v0.59.9 still newest on GitHub and PyPI).
superplane: recorded v1.0.0 (2026-10-27), the first stable platform release after the rc1-through-rc9 train, with the stable-tags-lag caution and bottom line rewritten.
zeroshot: recorded the train to v10.19.1 (2026-10-26, npm in sync): guided graph-design skill and Platform endpoint move in v10.19.0, SDK publishing fixes and bundled-harness bumps in v10.19.1.
machinist: rotation re-verification, every cited URL re-fetched, all claims confirmed (v0.4.0 still newest, 176 commits past v0.4.0, zero coverage holding to 2026-10-28).
har: rotation re-verification, every cited URL re-fetched, all claims confirmed (v1.16.0 still newest, HAR HQ Team $400 pricing unchanged, Kerno v1 live).
fusion: fixed ## Changes list formatting (stray blank line).
software-factory-feature-matrix: recorded SuperPlane v1.0.0 and the awesome-software-factories 2026-10-23 to 2026-10-26 sweep resolutions (wippaAGENT, Guarded Dark Factory, shipmill below the bar; six entries flagged cross-category).
software-factory: no new dedicated list above the bar; varun1505/awesome-software-factories (42 stars) remains the only one.
trackers-and-leaderboards: benchmark-radar - volatile re-verification: 2026-10-27/28 merges add share-any-benchmark cards (PR #1017, deploy gated on card serving) and a homepage-title search change (PR #1022); badge to 39,640 records over 93 days as of 2026-10-28; v0.13.0 still newest
trackers-and-leaderboards: bestvaluemodel - volatile re-verification: the 2026-10-28 diff removed 427 rows and re-scored 72 with no additions, the pattern of an upstream Artificial Analysis index re-base; bot refresh cadence unchanged
trackers-and-leaderboards: livebench - rotation re-verification with every cited URL re-fetched live: nothing moved since the 2026-10-18 push, newest question release still 2026-01-08, frozen-set gap approaching ten months
trackers-and-leaderboards: openrouter-rankings - rotation re-verification with every cited URL re-fetched live: about page now claims 700T+ monthly tokens (from 500T+), rankings page again client-rendered with no data-through date readable, Data API docs and Stripe-announcement page unchanged
trackers-and-leaderboards: vectara-hallucination-leaderboard - volatile re-verification: README stamp still October 22, 2026, leader antgroup/finix_s1_32b at 1.8 percent unchanged, repo pushed 2026-10-23, no edit needed
trackers-and-leaderboards: matrix and category index verified against membership (13 members, sorted, all rows at header width); no broken internal links; star-history chart markup correct in all five charts
trackers-and-leaderboards: no credible dedicated awesome-list source found this refresh; the standing no-dedicated-source decision stands
section-root: model-selection-for-coding-tasks: added gpt-6.1-terra (2026-10-26, $1/$5 with a $0.05 cache read) as a column, the first GPT-6-family model priced between the cheap tier and the converged workhorse rate; gpt-5.6-terra and gpt-5.6-luna still listed on OpenAI’s pricing page; the Kimi daily-churn readings moved (kimi-k3 listing $0.15/$20.00 with $0.14 cache reads, input floor $0.169/$20.00 on Wafer, output floor $0.17/$8.099 on Relace’s fp4 endpoint, roster 30, kimi-k2.7-code’s host floor cut to $0.56/$3.10 on Inceptron’s int4 row); every other provider anchor re-verified unchanged against all seven pricing pages, OpenRouter, and models.dev.
section-root: model-provider-feature-matrix: synced with the guide’s refresh, the Moonshot flagship and workhorse cells’ OpenRouter floors moved and the endpoints roster widened from 29 to 30; every other vendor cell re-verified unchanged against all seven pricing pages, OpenRouter, and models.dev.
section-root: agentic-coding-tools-landscape: moved the DeepSeek Harness line’s newest prerelease to dsh-v0.2.2-alpha.2 (2026-10-28, with rc.4 the same day), re-dated the Gemini CLI enterprise-docs stamp to 2026-10-26 with the same banner, folded Cline’s install count to about 5.8 million, and re-read the Cursor pricing page with the X-bundle plans unchanged; every external link re-fetched, Bloomberg and devops.com again 403 to scripted fetches.
section-root: the-agentic-development-environment-landscape: rotation re-verification, all 17 URLs 200, Air’s supported-agent, surface, and cloud-run claims re-verified against its live page; Orca’s dated star count untouched per the stars exemption.
section-root: context-management-patterns: rotation re-verification of all eight cited pages, every one 200, the pattern claims hold against the live docs, intro as-of re-dated.
section-root: the-tells-are-structural: re-fetched all eight cited sources, every one 200, the intro’s re-verification date moved; sepia and slop-grader star figures untouched per the stars exemption.
section-root: model-benchmark-matrix: link re-verification, every cited URL 200 with the PaperBench page again 403 to scripted fetches, Terminal-Bench still at 4.0, intro as-of re-dated.
section-root: agents-_index: list currency restored, the verified stamps of all 25 matrix one-liners synced to their matrices’ updated dates, ten stale member counts corrected, twelve stale newest-column or revision claims corrected, and the essay dates re-dated; the drift came from the 2026-10-27 run updating category matrices without an index pass.
cross-category: stars scan from the 2026-10-25T02:35:56Z checkpoint found no new starred repositories; checkpoint unchanged at 2026-10-25T02:35:56Z
cross-category: andyrewlee 2026-10-27 additions resolved: hotline and open-walnut on the thin-record tail, agentglass on the mission-control tail, Scalekit on the closed-tier line, Speakeasy Gram on the MCP-platform kind, CAO handed to orchestration
cross-category: bdots renamed to fayazara/coworkers upstream (306 stars, was 156 on 10-25); the held 10-25 rejection carries
cross-category: AMO additions resolved: isomux and intentic on the thin-record tail; sno-station (929 stars, active, zero coverage) handed to orchestration
cross-category: broad-source back catalogs fully walked per the 2026-10-25 digest line; 950 never-matched entries resolved on standing families (93 dead links, the rest batched below)
cross-category: note-worthy candidates handed off report-only: ChatDev to orchestration, SWE-agent and open-swe to harnesses, qm to orchestration/harnesses routing, ironclaw to assistant-runtimes, humanlayer to control-planes/orchestration routing, claude-code-security-review to code-review, claude-obsidian to memory, cli-agent-orchestrator and tuios to orchestration
cross-category: ai-for-developers, hesreallyhim, e2b-dev, slavakurilyak unchanged since their last checks (commit history); scans carried over, back catalogs still walked
sourcegraph-code-context: the sourcegraph/sourcegraph GitHub mirror’s new 404 leaves the note untouched, the subject is the sourcegraph.com platform and every citation fetches (site re-confirmed 200 this run)
section-root: gpt-6.1-terra (2026-10-26, $1/$5) added to model-selection-for-coding-tasks, the Kimi OpenRouter floors and k2.7-code host floor moved, DeepSeek Harness recorded dsh-v0.2.2 and dsh-v0.2.3 same-day releases, DeepSeek’s Search/Fetch APIs start billing November 1
software-factory: fabro, fusion, loki-mode, and ouroboros Changes lists repaired (stray blank lines removed inside the lists, no content change)
session-analytics-feature-matrix: the Privacy row label reworded to Privacy handling in the rewritten row (the banned-terms list rules; the two note-level “posture” uses stay on queue q#65)
evaluation-review: the session-analytics routing of forevercrab321-svg/leevar-battery is already resolved, the battery was rejected 2026-10-27 on the zero-adoption transcript-grading floor line

(daily refresh, 1) #

(daily refresh, 2) #

(daily refresh, 1)
#

(daily refresh, 2)
#