Former OpenAI safety researcher Daniel Kokotajlo’s, now part of the A.I. Futures Project, has published a new website… ai-2027.com. This graph from the site charts progress through 2027, with progression from “Superhuman Coder” (4X R&D boost) to full “Artificial Superintelligence” (2,000X boost), with a critical branch point where humanity must choose between controlled development or an acceleration race. The alternative paths – “Alignment gets solved” versus “USG captured by AGI” – reflect Kokotajlo’s belief that superintelligent systems could emerge within three years. The website contains a detailed fictional narrative exploring how this technological progression might unfold in practice, and is a fascinating read.