AI Endgame: researchers privately fear superhuman AI while racing to build it
thedealdirector · x · 2026-09-29
- Newsletter Infra Play #162 argues a viral post by Jacob Coxon (173M views) warning about superhuman AI kicked off the "endgame" stage — whatever happens, there's no returning to pre-2023.
- Builders earnestly believe AI could kill us all by decade's end; executives publicly soften it but privately express fear.
- At OpenAI, many haven't internalized the stakes; Anthropic understands them but believes no one else will act responsibly, so it must race ahead anyway.
- The author calls accepting this race from a private company's Slack a hubristic gamble: "speedrunning alignment" demands extraordinary confidence no better trajectory exists.
More from AGI Musings
- Frontier Lab Exec: AI Will Beat Humans in All Science in Months, But Now Is Wrong Time to Slow Down — geoffreyirving · 2026-09-30
- Gary Marcus: safety researchers finally admit we have no alignment guarantees — GaryMarcus · 2026-09-30
- Ex-DeepMind, OpenAI, Anthropic researcher Geoffrey Irving's interview called the most frank must-watch — geoffreyirving · 2026-09-30
- Researcher from Google Brain, OpenAI, DeepMind puts human extinction odds near a coin flip — geoffreyirving · 2026-09-30
- Debate: Is the Personal Agent the Final Product? Skeptic Says the Rich Never Use a Single Proxy — dansitu · 2026-09-30
- New Paper Analyzes How Cheaper AI-Driven Software Engineering Will Reshape the Economy — lihua_lei_stat · 2026-09-30