Ex-forecaster rejects OpenAI's $2M NDA, reveals 2027 automation roadmap
thetripathi58 · x · 2026-09-01
Former OpenAI forecaster Daniel Kokotajlo revealed he rejected a strict gag order and $2 million exit package upon resignation, willing to lose 80% of his net worth to warn the public. Internal backlash eventually forced OpenAI to backtrack on the silence clause while letting him keep the money.
Kokotajlo is now exposing the timeline frontier labs are hiding: competitors like Anthropic are growing so fast they could encompass the entire world economy by 2030. The secret schedule for 2027 is to fully automate human AI researchers, building an army of geniuses in data centers that trains itself 100 times faster than humans. He estimates a 70% probability of failure, but tech CEOs are too scared of losing the race to listen.
More from AGI Musings
- Proposal for a global weekly AI ban to save human intelligence — lipeng0820 · 2026-09-01
- Agent boundary is about memory and control, not cloud vs local — BLUECOW009 · 2026-09-01
- METR staff surprised by HF incident, showing dangerous-capability evals failed — NathanpmYoung · 2026-09-01
- Andrew Chen on the AI Era Shift: Hardware is the Only Moat, SaaSpocalypse Begins — andrewchen · 2026-09-01
- Bacteria cooperate without consciousness; AI substrate argument debunked — max_paperclips · 2026-09-01
- Opinion: AI future should be multipolar, not a Singletonian monopoly — lfschiavo · 2026-09-01