Research Do Corporate Incentives Teach AI to Protect Itself? July 13, 2026 · 4 minute read The tools that make an AI safe, like kill switches, also make it stoppable. A model therefore has an incentive to help limit rival AIs but not itself.
An Athlete's Mind ( recent reads) The Designer, The Craftsman, & The Programmer ( recent reads) The Second And Third Falling ( recent reads) Vibe Coding: Taste Is The Last Frontier ( recent reads) "Is The Light Merely For The Learned?" (Er Lyset For De Lærde Blot?) English Translation ( recent reads) "Surrounded by Enemies" (Kringsatt av Fiender), or "To The Youth" (Til Ungdommen) English Translation ( recent reads) To An 18 Year Old Me ( recent reads) Mediocre Intelligence, Outstanding Results ( recent reads) How to host a hackathon ( recent reads) Raising for the Endgame: An AI Safety Founder’s Primer ( recent reads) Learning From Architecture ( recent reads) AGI Security: Rethinking Society's Infrastructure ( recent reads) AGI Privacy ( recent reads) Don't Give Up ( recent reads) The AI Bubble ( recent reads) The concrete risks of AI misdeployment ( recent reads) Sentware ( recent reads) Confident optimism ( recent reads) The Expensive Internet hypothesis ( recent reads) The AGI endgame ( recent reads) "Just do the thing" ( recent reads) Cybermorphism ( recent reads) Go on a yearly hiatus ( recent reads) Let's make AI agents safe ( recent reads) My Tools ( recent reads) Audiovisual reading ( recent reads)

Research

Socials

News