@varunsrin Code generation via agents is just a faster way to get to computation…you still need to understand if your computation is doing the right thing, is scalable and cost efficient in terms of your business. Us old experienced tech people get this shit through years of learning 😁
@karpathy Having been involved with various initiatives to deliver personalization over the years..I do believe LLMs + agents + context is genuinely going to take personalization to the next level…
@iruletheworldmo I’ve been using gpt 5.4 for helping create narratives for complex commercial legal cases..it seems to be more critical than Opus 4.6 when going back and forth to refine the narrative. Opus 4.6 was a lot better for some finance flows…e.g putting together a complex DCF model…