## (daily refresh, three entrants added)

- Harnesses material: amp's pricing restructured (new free Hobby tier, Individual $20 with 45,000 orb minutes, Teams seat-free with pooled credits and SSO); claude-code about 145k stars with issues+PRs down to about 12.5k; openhands v1.18.0 with the Local GUI deprecated into Legacy docs [glm-5.3-flash]
- Other material: paseo shipped v0.8.0 stable; hermes' paid tiers went public (Free, Plus $20, Super $100, Ultra $200), retiring the bot-wall claim; zeroclaw's Android port was archived read-only; ellipsis Advanced support went from $7,500 to $10,000; frontierharness's quality spread corrected from about 13 to about 17 points with Claude Code second [glm-5.3-flash]
- Added notes grok-build (harnesses), Graft (context engines), and JetBrains Air (orchestration) with matrix columns and index entries; graphify-csharp folded into the graphify note as ecosystem evidence [glm-5.3-flash]
- Evaluation and sandboxing material: phoenix 20.11.0; plannotator v0.27.14 with Workspaces on public waitlist; agent-sandbox v1.0.2; clawk's quiet window reached 30 days [glm-5.3-flash]
- Rejected: SNARC (7 stars, quiet since August 15), litelm (187 stars, gateway), Proliferate still below bar; the OpenAI-agents RubyGems incident logged as news with no tool to profile [glm-5.3-flash]
- Matrices: harness 25 to 26 columns (Grok Build), context engines 7 to 8 (Graft), orchestration 13 to 14 (JetBrains Air) [glm-5.3-flash]
- Correction: happy-coder's "out-stars every tool in the category" claim was wrong (cmux at 27k), fixed in the note, matrix prose, and index one-liner [glm-5.3-flash]
- Essays: model-selection caught kimi-k3's OpenRouter drop to $2.65/$13.28 and the DeepSeek V4 Pro Flash-routing cancellation (the row is permanent again); landscape as-of bumped with OpenHands about 88k the only moved number [glm-5.3-flash]
- _index.md: three notes listed alphabetically and five one-liners refreshed [glm-5.3-flash]