Honestly, harder than FAANG in some ways. For example, their system design is heavier than typical FAANG. They ask things like designing a RAG-based enterprise chatbot, a low-latency GPU inference serving system, or a Slack-like multi-tenant chat system. And they have a behavioral round where they specifically ask about your views on AGI and AI safety. The flavor is very different. It rewards depth over pattern matching. For prep, Gotham Loop has been building out their OpenAI question bank and, DDIA for system design fundamentals, and honestly just get comfortable writing production-quality Python from scratch without relying on LC templates, using apps like Apexinterviewer too to simulate OpenAI's peculiar interview environment and do multiple practice reps. Put all these together and you should be in a good place before your first interview round.