Tweet: LLMs Automate Script Writing for Large Search Spaces
The post reframes how LLMs actually contribute to hard problems: not as direct reasoners but as generators of throwaway automation scripts. Researchers and engineers doing combinatorial or exploratory work can apply this by delegating tedious search scripting to models. The observation about in-context learning through oracles offers a practical pattern for anyone building evaluation or verification loops.
A post on X argues that current models do not solve problems by reasoning over a large search space directly, nor do they work well when the search space is small. Instead, models write many scripts to search large spaces that would be tedious to handle manually. The author adds that this approach works beyond research mathematics. They also note that models perform well at in-context learning via oracles.
> They don't work in a small search space, and they aren't using a gigabrain directly on a large search space. They're just writing lots of scripts to search large spaces that would be annoying to do.
This works beyond research math btw Models are good at ICL via oracles
