LLMs are now sufficiently smart to do at least some basic ML research tasks autonomously (going from idea to report). And quite likely by the end of the year they might be capable enough to reproduce ML papers with some variations, thus, possibly, accelerating AI research.

Example of autonomous work (artifacts): https://github.com/killerstorm/auto-ml-runner-artifacts/tree...

Runner script: https://github.com/killerstorm/auto-ml-runner