Skip to content

Person

Adam Kaufman

4 commentary pieces · December 2025 – July 2026

Commentary by Adam Kaufman

From the commentary rail — every link leaves the site for the original piece.

27 July 2026 · Redwood ResearchUntrusted advice for AI control: Short, strong advice significantly uplifts weak LLMsWhen a misaligned AI can only output tiny amounts of information, it may find sabotage very difficult
29 May 2026 · Redwood ResearchRetrying vs Resampling in AI ControlWe’ve just released a new paper: Retrying vs Resampling in AI Control. We revisit the resampling protocols introduced in Ctrl-Z with an up-to-date setting and much stronger models, and compare them against “retrying” protocols similar to Claude Code auto mode or Codex Auto-review.
30 March 2026 · Redwood ResearchBlocking live failures with synchronous monitorsA common element in many AI control schemes is monitoring – using some model to review actions taken by an untrusted model in order to catch dangerous actions if they occur.
18 December 2025 · Redwood ResearchBashArena and Control Setting DesignWe’ve just released BashArena, a new high-stakes control setting we think is a major improvement over the settings we’ve used in the past.