Hard takeoff

A theoretical scenario in AI development where an artificial intelligence system undergoes a rapid, discontinuous increase in capability, potentially leading to an intelligence explosion. Contrasted with “soft takeoff” (gradual progress), hard takeoff implies a sudden, uncontrollable acceleration in AI intelligence.

• self-improvement is the mechanism enabling hard takeoff: AI agents now significantly contribute to their own evolution.
• In “Hard Takeoff has started” (https://www.youtube.com/watch?v=mhoFqhLXc3g), Matthew Berman asserts we have officially entered the recursive self-improvement phase, where AI systems autonomously enhance their architectures and algorithms, potentially triggering an intelligence explosion.

Related concepts:

• intelligence-explosion
• ai-safety
• AI alignment
• AI development trajectories

Backlink: 2026 04 14 AI Recursive Improvement

Source Notes

• 2026-04-07: Hard Takeoff has started
• 2026-04-08: Hard Takeoff has started
• 2026-04-10: Hard Takeoff has started