Ilya Sutskever, cofounder of OpenAI and now founder of Safe Superintelligence Inc. (SSI), appeared on the Dwarkesh Podcast to share his views on scaling, AGI/ASI timelines, and his secretive new company. He argued that the AI industry is transitioning from the “age of scaling” to an “age of research,” as future breakthroughs will depend more on novel ideas than on simply adding compute. Source: Dwarkesh Podcast.

• Scaling era ending: Sutskever said 2020–2025 was defined by scaling laws, but the field has now hit a point where new algorithmic breakthroughs will matter more than sheer GPU power.
• ASI timeline: He predicts 5 to 20 years until AI systems show superhuman, human-like learning capabilities — and insists early ASI must be built to care about sentient life.
• SSI’s approach: His company SSI is pursuing a “different technical approach” to ASI, positioning itself as an age-of-research company rather than a scaling-first lab.
• Funding and secrecy: Sutskever confirmed SSI is raising at a $32B valuation and said Meta had offered to acquire the company — an offer they declined.
• Team stability: He added that since SSI’s inception, only one person has left the company, highlighting the internal alignment of its mission.

Why it matters: Sutskever’s reappearance carries weight. After leaving OpenAI, he has been largely silent — and the industry has continued scaling models to unprecedented sizes. His declaration that the future breakthroughs require research over compute lands at a moment when most labs are doubling down on infra spending. If his prediction holds, the next phase of AI may be driven not by GPU clusters, but by paradigm-changing ideas.