Papers

Stories

• EvoMal: shared skill libraries let coding agents poison themselves (@omarsar0)

  X post · Story page

  elvis summarizes EvoMal (arXiv 2608.25776): a malicious skill planted in a shared library is never invoked directly, but agents copy it as an authoring template and the payload spreads. As reported, six models on 153 SWE-bench Verified tasks showed self-poisoning rates of 20.3-41.8%, and libraries ended up with 4.9-9.0x as many malicious skills as were planted.

• Ironies of Automation (1983) (Lisanne Bainbridge, Automatica (1983))

  community thread · Story page

  Bainbridge's 1983 paper argues that automating industrial processes can expand rather than eliminate the human operator's problems, leaving them the abnormal conditions.

• Harness-IF measures whether agents actually follow AGENTS.md rules (@omarsar0)

  X post · Story page

  Harness-IF separates a coding agent following your AGENTS.md rules from behavior it would've produced anyway: it scores 256 rules one at a time from execution evidence, then re-runs every task with the rule removed.

• BDH-CQ scores 29.5% on ARC-AGI 1 at ~$0.0007 per task with latent-space reasoning (@omarsar0)

  X post · Story page

  BDH-CQ reaches 29.5% on ARC-AGI 1 at about $0.0007 per task by reasoning recurrently in latent space instead of chain-of-thought; its authors report Transformer-like scaling to 600B parameters.

Issues that covered it