Post

Here's my enormous round-up of everything we learned about LLMs in 2025 - the third in my annual series of reviews of the past twelve months simonwillison.net/2025/Dec/31/... This year it's divided into 26 sections! This is the table of contents:

Don’t quite agree with the first topic. These are 3 related things, not the same thing. Reasoning, RLVR and inference scaling

Reasoning can be used to any and every topic. RLVR only to very specific topics that can be verified by external means. I can’t think anything else besides math, coding and puzzles like Andrej mentioned

And when they take place. RLVR is during training, reasoning at inference time

Sure, but RLVR is the thing that enables that reasoning process, right? Otherwise we are back to the old "think step by step" prompting hacks