@petrogustavo How many lines of cocaine do you think it'll take you to finish this book? And how many more for its only likely reader (you) to actually get through it?
@thenanyu I see this as an extremely productive paradigm, but the compute is currently a limitation. I would trust Fable to create and execute the quatility improvement runs, but instead I have to act as the bridge to Opus and Sonnet.
@Claude1o Feed Claude files about you (e.g., past projects) and ask it to interview you to build a personalized setup. It can read/write files on your computer. Once you create a solid workflow for a task, you just give it the core facts and it handles the rest automatically.
@Claude1o Claude Cowork or Code. Cowork is the Natural Language Programming version. LLMs lack memory, so the current paradigm uses RAG: it reads text files with your instructions, success criteria, and preferences for each task.
@TomWatkins1994 If it keeps advancing at this rate, will programming skills still matter? Beyond fundamentals, clear success criteria, and a few past examples, what else will we contribute? Once it can reliably plan, execute, and troubleshoot, I’m struggling to think of anything else.
@YashHustle_22 It's essentially become a battle between the smartest model and the most compute, but I like Anthropic's chances of catching up on compute.
@ryanbrewer We’re still limited to our human brains. Also, the fastest interface is text or speech. That’s why Elon Musk is banking of Neuralink as the future paradigm.