The way most people leverage test-time compute today is via a form of test-time NL reasoning that is computationally equivalent to test-time…

The way most people leverage test-time compute today is via a form of test-time NL reasoning that is computationally equivalent to test-time search (sometimes with a verifier / grader in the loop). The other major avenue to leverage test-time compute is test-time training. Gradients are a precious signal, there's no reason not to use it at test time (other than the fact that it would be difficult / expensive from an engineering standpoint). https://x.com/fchollet/status/1869053929314722019 Anyway, glad to see that the whole "let's just pretrain a bigger LLM" paradigm is dead. Model size is stagnating or even decreasing, while researchers are now looking at the right problems -- either test-time training or neurosymbolic approaches like test-time search, program syn…

Source: Francois Chollet (X) | 2026-08-13