Summary
In a Y Combinator fireside-style session, Stan Hulu — founder of Dust and formerly an engineer at both Stripe and OpenAI — shares his perspective on building an enterprise AI platform in an era dominated by frontier labs. Hulu spent three years at OpenAI working on large language models and mathematics research before leaving to focus on product, founding Dust around the premise that LLMs were ready to fundamentally transform how people work.
The core thesis behind Dust is model agnosticism: rather than locking enterprise customers into a single lab’s tokens, Dust acts as an orchestration layer that can route work to whichever model is currently best suited for a given task. Hulu uses a vivid analogy — buying machines from the same company that supplies your electricity, where a single provider failure leaves you stranded — to argue that vendor independence is a structural advantage no individual lab can replicate. He notes that in practice, usage often concentrates toward one provider at any given moment, but the landscape shifts quickly enough that the flexibility is essential.
Hulu also speaks candidly about the funding environment, describing Dust’s seed and Series A as straightforward but the Series B as genuinely difficult, partly due to the company being based in France and partly because large venture dollars are increasingly absorbed by the labs themselves. He expresses cautious optimism that lab IPOs will eventually return liquidity to the broader startup ecosystem. The conversation closes with reflections on how work itself will evolve, with Hulu predicting that activities that feel like real work today will look as passive to future observers as knowledge work looks to a prior generation of physical laborers.
📺 Source: Y Combinator · Published July 23, 2026
🏷️ Format: Interview
