I've been using Fable 5 lately for coding and chats. It occurs to me that it's much smarter than other models I've used -- it writes more densely, more clearly.
But it's still too agreeable. You challenge it and it immediately backs down. It seems that every LLM I've used is like this - too agreeable, too eager. They never push back and criticise your ideas, they never double down. They always calibrate to what you're saying.
Until that problem is solved, it doesn't matter how well they write, how verbose they are. Intelligence is more than just verbosity and clear thinking, it's about holding an internal understanding of your own and pushing back when you need to. LLMs still can't do it.
Comments URL: [https://news.ycombinator.com/item?id=49015011](https://news.ycombinator.com/item?id=49015011)
Points: 1
source & further reading
news.ycombinator.com — original article
New Inference Server for DGX Spark: large model C4:55-90 tok/s no spec decode
Build a website for AI pics for dating apps in 20 hours, launched ads, failed
Building an AI-orchestrated publishing workflow for a long-form writing project