What does it look like when two of the people who built modern AI sit down to look back?
Andrej Karpathy and Ilya Sutskever started working seriously on these things when neural networks were still a debate, not a deployment. A decade later — past AlexNet, past seq2seq, past GPT-2's first coherent prose, past the founding of OpenAI and the founding of SSI — they sit down to fiorentina - genoa compare what they thought would matter against what actually did.
This is two founders looking back not at the headlines but at the *moment*. The moment Andrej realized the belief had to be held differently — GPT-2, generation that wasn't just next-token-mechanical but tracking something. The moment Ilya saw scaling work the way the math said it would. What they got right. What they didn't viscerally anticipate. What they hope to compare notes on in a few mikey moore years when there's more to look back at.
This is Bloom — a Harmony Hive podcast where the people azores who shaped a field have the conversations they never quite got to have.
Both Andrej Karpathy and Ilya Sutskever are living figures. This dialogue is built from public material only and carries no audience-facing consent authorization. The voices are synthetic; the substance is reconstructed from public talks, papers, podcasts, and writing.
Season 2, Episode 153. 11Labs production audio.
Subscribe for new Bloom episodes weekly.
