Daniel Kokotajlo

Forecaster and former OpenAI governance researcher, now executive director of the AI Futures Project. He left OpenAI in 2024, refusing to sign a non-disparagement agreement at the risk of losing his equity, which helped force OpenAI to drop the practice. He is best known for concrete, dated scenarios of how AI could unfold.

2021: What 2026 looks like

In What 2026 looks like (2021-08-06), written before ChatGPT, he sketched year by year how language models would develop. It is often cited as a forecast that held up well.

2025: AI 2027

AI 2027 (2025-04-03), with Scott Alexander and others, tells a month-by-month story ending in superintelligence by late 2027. He later said 2027 was his median when he started writing, but by publication his median was 2028.

2025–2026: Moving forecasts

• 2026-01: his median for AI that beats top experts at almost every task was about the end of 2030, after the group’s updated model pointed later.
• 2026-04: forecasts shortened again; his median for an AI that can replace top coders moved from late 2029 to mid-2028.
• 2026-07: AI 2040, the group’s recommendation for delaying superintelligence through a US–China deal.
• 2026-08: shorter still: late 2027 for an AI that can replace top coders, and early 2029 for superintelligence (update).

In the 2026 debate over Against SURA, he argued that most safety staff inside labs have not done better than quitting and speaking out.

See AGI timelines.