I try to write whenever I have time.

• 2026-07-07: Superhuman Articulacy as an LLM Safety Target
• 2026-07-05: Pangram as a model for AI safety general managers
• 2026-06-28: A reading list for generalists
• 2026-05-08: Notes on the current LLM research paradigm
• 2026-05-06: Blind deep-deployment evals for control & sabotage
• 2026-05-02: Dev tools in 2026
• 2026-04-30: Goblin Mode, 24 Hours Later

Personal

• 2026-06-29: More about me
• 2026-01-02: The best year of my life