Rosa Del Mar

Daily Brief

Issue 39 2026-02-08

Vulnerability Research As An Llm-Suitable, Search-Like Task With Iterative Tooling Gains

  • Assuming LLMs will not become effective at vulnerability research is risky.
  • Frontier AI labs include vulnerability research outcomes in their model cards.
  • Some public commenters are dismissing the underlying vulnerability-research claim as mere advertising.

Artifact Provenance And Social Context

  • Karen James, described as the speaker's friend and neighbour, made the speaker a Kākāpō-themed mug.
  • The speaker says they love the mug very much.