LodeHQ
AI news for people who shipFree, by email

AI · lesswrong.com · Sunday, 8 November 2026

Corrigibility fund rejects agent-control proposals and sloppy applications

The Corrigibility Research Fund is filtering out recurring proposal patterns, including agent-control systems that try to keep AI in a box, so applicants need to show direct research value rather than another oversight layer. The evaluator says AI use is not an automatic rejection, but high Pangram scores, copy-pasted formatting, and applications that waste space defining corrigibility are warning signs; the strongest applications are clear, short, and dense.

Read the source: lesswrong.com ↗In the 8 November issue →