On ToTheMoon, Ilya Sutskever is the researcher the hosts cite whenever the talk turns to the limits of the current way of building models. He first came up in 2024 through the story of his departure from OpenAI, and even then the hosts noted the gap between research standing and public fame: one of the creators of AlexNet became widely known only after ChatGPT and the Altman firing, and joke insider tips about him joining a Russian bank showed that the tech world discusses people unknown outside the bubble (episode 012).
In late 2024 Sutskever became the channel's main reference in the scaling-ceiling debate. The hosts relayed his claim that scaling laws had run out and drew a conclusion: if models sit at the same level for six months, a public safety benchmark is due (episode 033). He was placed next to LeCun as someone who is not a pessimist but doubts the current approach can reach AGI (episode 032), and by December the channel was quoting his framing of the fundamental task ahead: catching the moments when a model deceives (episode 038).
In 2025 attention shifted to his new company. The exhaustion of training data was tied to his search for different fundamental approaches (episode 041); in spring the hosts discussed a two-billion-dollar round (episode 054) and noted that the company's website offered only a few lines about its plans (episode 056). By year's end his point that the current architecture has hit a ceiling was cited in passing (episode 089), and in 2026 the hosts recalled that, brought in by Musk, he later refused to let Musk take sole control of OpenAI (episode 109).
Author's trends
the author's thoughts whose episodes feature Ilya Sutskever · blue marks the months they meet