Sam Altman just said the quiet part out loud. The OpenAI CEO suggested the company could be less than six months away from achieving recursive self-improvement, a capability where AI systems can meaningfully enhance their own development, and that hitting this milestone could push back the company’s highly anticipated IPO.
For a company valued at roughly $852B in its latest funding round, the idea of voluntarily pumping the brakes on a public offering is, well, noteworthy.
What recursive self-improvement actually means
Think of recursive self-improvement like compound interest, but for intelligence. Instead of humans writing code to make AI better, the AI starts writing code to make itself better. Then that improved version makes an even better version. And so on.
Altman has previously described current AI progress as a “larval version of recursive self-improvement.” Earlier in 2025, he went further, stating that humanity had surpassed the “event horizon” for AI self-improvement.
