Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-09-12 · high
A CLAUDE.md instruction against commit co-author lines used to lose to a built-in reminder anyway.
Why now: The 2.1.269 changelog release is covered here as of September 12, 2026.
Claude Code used to ignore a no-attribution rule in CLAUDE.md, adding Co-Authored-By and Generated with Claude Code lines to commits anyway. Version 2.1.269 fixes that. A hardcoded attribution reminder had been overriding that rule, per the 2.1.269 changelog.
It's a real gap between what users configure and what the tool does. Anyone who set that rule and still saw attribution lines in past commits wasn't making a mistake.
The new release gives user-level rules precedence over the built-in reminder. Rules set through managed settings, the kind an organization controls rather than an individual, still apply and aren't affected.
The same release changes how /btw handles side questions. Those questions were producing fabricated tool calls, invented function calls with invented output, printed as if they'd run. Now Claude Code flags them instead of printing them.
Both bugs share a shape. The tool stated something false about its own behavior with no visible seam. One overrode a config setting with no error or warning. The other invented evidence of work it never did, in a surface people read as authoritative.
The changelog doesn't say how long either bug existed before this release, or how many commits picked up attribution lines users had explicitly banned.
Each link below shares sources, entities, or timing with this story.
Eight percent of a monthly credit allowance gone in two hours. That's a real Copilot Pro+ user after metered token billing took effect June 1. Another spent over $6 on a single change request. A Claude 4.8 session reportedly ate 1,180 credits, roughly 16% of a Pro+ allowance,...
This Rust harness (+2,585 stars) competes on resource footprint rather than features: 27.8 MB PSS for a single session with local embedding disabled, claimed 13.9× less than Claude Code and 6× less than jcode's own embedding-enabled mode. Time-to-first-frame 14.0ms against a c...
A paper from Xiao Yu, Baolin Peng, and Ruize Xu makes a claim that seems obvious once stated and is genuinely new as a training methodology: modern agents are inseparable from their inference harnesses, so training them in stripped-down RL sandboxes produces a train/serve mism...
Rust, created May 14, at 2,643 stars (GitHub). Every run produces checkpoints linking a commit to the session that made it, including prompts, tool calls and reasoning. It runs Claude Code, Codex, its own agent and anything from the ACP registry side by side against one codeba...
wanshuiyin/HERO-Anti-OverDefense went from creation to 68 stars in a single day. HERO is Hashing, Edge cases, Rubrics, Overbuild, and the claim is that agent over-engineering isn't diffuse but falls into four recognizable shapes suppressible with a portable prompt contract acr...
It now dedupes local CLAUDE.md files against checked-in ones, proposes trimming content it can derive from the codebase, ranks unused skills, MCP servers, and plugins by context cost, and flags slow hooks. Background subagents run by default and, when launched from claude agen...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.