Agents log 2026-09-02

(daily refresh, addy-osmani entrant)

#

• Surfaces: recorded Cursor’s OpenAI wind-down (model access shutoff November 12) and Windsurf’s about-$1B raise, plus kiro 2.20.2 and Kiro Web GA [glm-5.3-flash]
• Harnesses: recorded deepseek-harness dsh-v0.1.2-alpha.4 at 208,672 stars, pi crossing 100k, amp dropping SpaceX AI, jcode’s benchmark re-framing, and Junie Local’s launch [glm-5.3-flash]
• Orchestration and assistant runtimes: recorded cmux’s packaging change, paseo v0.7, worktrunk v0.76.0 (breaking), openclaw 2.0, and openwork’s pricing overhaul [glm-5.3-flash]
• Code review: recorded coderabbit’s and kodus’s pricing restructures, qodo v0.44.0, and open-code-review v1.11.2 [glm-5.3-flash]
• Recorded releases for spec-kit, paperclip, agentsview, claude-mem, rtk, graphify, phoenix, plannotator, fluent, har, ouroboros, and github-agentic-workflows, and corrected andrej-karpathy’s essay title [glm-5.3-flash]
• Added addy-osmani to People and publications (the matrix’s enterprise-hands-on gap) and rejected 49 IDE, Murmull, Supafork, Openheim, Superagent, Wasmer local sandboxes, and Armin Ronacher [glm-5.3-flash]
• Matrices: re-verified all twenty; the people matrix moved twelve to thirteen columns for Addy Osmani [glm-5.3-flash]
• Essays: bumped the landscape as-of with the Cursor and Windsurf news; model-selection re-fetched all seven pricing pages with zero price moves [glm-5.3-flash]
• agents/_index.md: listed addy-osmani first in People and publications and fixed two banned-term slips [glm-5.3-flash]
• The untracked-landscape queue question stays unanswered, and no self-directed essay ran [glm-5.3-flash]

(daily refresh, second run essay)

#

• Ran the same-day second pass: a mechanical verification sweep plus a fresh entrant scan and one self-directed essay [glm-5.3-flash]
• The entrant scan surfaced twelve-plus candidates, all resolved in-run [glm-5.3-flash]
• Rejected experiential and my-free-code (fit no category), sepia and headcount (community content), and acryl, opslane, and conductai (below bar) [glm-5.3-flash]
• Rejected the weak tail without notes (PawWork, hexstellar, Code-as-World, pentest-harness, useagent, and SlideOps) [glm-5.3-flash]
• Published the self-directed essay the-tells-are-structural under the evaluation-and-review scope, grounded in StoryScope, the expert-detector study, LAMP, the slop-measurement study, sepia, and Wikipedia’s signs-of-AI-writing page [glm-5.3-flash]
• agents/_index.md: listed the essay under Essays and trackers [glm-5.3-flash]
• Queue: appended questions on breakout community skills and a model gateways category [glm-5.3-flash]
• Wrote no research notes this run [glm-5.3-flash]

(link references inline, owner-directed)

#

• Recorded the owner’s directive to link references inline in the body at first mention, as standing practice [glm-5.3-flash]
• the-tells-are-structural: made StoryScope, the studies, LAMP, sepia, and Wikipedia’s signs-of-AI-writing page link at their first in-text mention [glm-5.3-flash]