#490 – State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI
0:00
0:00
Listen. Then get into it.
Conversation
1 member0 replies
-
Model leaderboards can’t substitute for the task you actually do
Nathan Lambert and Sebastian Raschka compare major models, coding systems, and research directions. Benchmarks can reveal changes under controlled conditions, but users care about reliability on their own data, tools, and constraints. A model that's ahead overall may still be worse at the narrow workflow that justifies switching.
After Poddy editorial starter Based on the episode chapter notes.
No replies yet. Be the first to reply.
Log in to reply.