Fetching from the wire…
Tools2026-08-27 · source-backed
qwen-code 0.22.2 shipped three review changes on August 26 that deliberately pull opposite ways: a do-not-refute list and constructible rejection bar in the /review verifier to stop it invalidly rejecting speculative-but-real findings, test acceptance criteria attached to every finding plus an explicit non-convergence ruling to break review-fix loops, and a land-with-residual-risk advisory when critical findings survive across rounds (GitHub). Language-pitfall and wrapper/proxy checks moved from a general agent to dedicated high-effort roles. The shape to copy: a verifier needs a floor on rejections as much as a bar on acceptances. Most review harnesses only build the ceiling.
Each link below shares sources, entities, or timing with this story.
Three separate Anthropic changes over about two weeks point the same direction, and none of them announced themselves as a strategy. Claude Code 2.1.238 added claude self-hosted-runner --defer-shutdown-max-min, which keeps serving attached sessions on SIGTERM, parks whatever's...
A spec is a press release until someone who didn't write it implements it. GitHub made Agent Plugins 1.0 generally available on August 12 across VS Code, Copilot CLI, the Copilot SDK, and the Copilot app on all plans. The spec, published August 6, was co-authored by AWS, Anysp...
Qwen 3.5 is the first major model pretrained specifically for agentic multimodal workflows from the first training stage, not fine-tuned after the fact. 397B total / 17B active parameters (MoE architecture), 256K context window, 201 languages. Ships with Qwen Code (terminal ag...
PR #9708, merged August 22, adds a temporal-reachability lens asking whether a value exists at the moment its consumer needs it, with findings formatted as "produced at X, needed at Y, Y precedes X", and an incident-replay lens treating the failure story in a PR description as...
Frontier labs publish demos. This one published the thing they actually page. Anthropic's August 18 writeup describes Claude Tag running as the first responder for CI failures inside the company. Dedicated service account. MCP connectors to Datadog, Grafana, PagerDuty, GitHub...
Released today, v0.21.13 rebuilds the /review skill around dedicated platform subcommands (meta, fetch-diff) instead of raw gh commands issued through prompt prose, and caps posted suggestions to Critical findings after round 5 to stop review loops (GitHub). Operationally it a...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.