Benchmarks compare models but they dont compare different topologies of models.
Planning large components and designing code should be a smart models choice. Writing a concrete implementation should be a model that generates good code, but it does not necessarily have to be a system designer.
@Rasmic The question is where do you put the balance between writing code with cheaper model and with smarter model.
Benchmarks are showing how good models are but they don’t show how good different topologies of models are!
@jrya7 False, can use the AI to write logic and ask the AI about the logic. You don’t need to know how to code.
My dad made a website that updates based on his orchestra’s Facebook.
He has Zero sense of what is a programming language.
AI gives people superpowers
Sam Altman literally gave a 43-minute masterclass on turning ideas into billion-dollar companies.
Bookmark it and give it 43 minutes today, no matter what.
@pcshipp I’m trying to find a balance, important code that has to be perfect so I won’t have problems with it in the future, I do with Astra. Independent code with Claude (and sometimes I tell Astra to use Claude for subtasks)