No single document has shaped the conversation around 2027 more than AI 2027, the scenario published in April 2025 by the AI Futures Project. Founded by former OpenAI governance researcher Daniel Kokotajlo, the nonprofit lays out a month-by-month story in which the automation of coding and AI research accelerates sharply, culminating in fully autonomous agents that — by roughly the end of 2027 — outperform humans at "everything".
Whether you read it as a warning or a roadmap, the forecast has become a reference point for policymakers, labs and investors. Its core claim is that recursive self-improvement in AI research could compress years of progress into months. It is exactly the kind of prediction 2027 will either validate or quietly retire.
What would actually have to happen
- AI systems reliably automating large parts of their own research and engineering.
- Agents holding up on long, open-ended tasks — not just short demos.
- Independent benchmarks, not vendor charts, confirming the jump.
Read alongside our look at AI agents going to work and the compute race powering it all.