It's been a while though, and the major providers still seem to rely on more powerful LLMs (System Two) for all tasks, regardless of complexity.
Similar to how chain of thought arose, I thought the System One/System Two Thinking Fast and Slow model was one of those human structures that fit really well onto training neural nets, and I thought that this new paradigm could really change things.
That doesn't seem to be the case so far though, and I'm kind of puzzled to why this hasn't broken the industry open by now. Of course my predictions are not infallible, and one shouldn't expect much accuracy or for progress to abide by rigid schedules, but nonetheless: how come Kahneman Computer Use models aren't everywhere by now?
This is a lot more work than basic zero or few shot prompting of a large general LLM. It's also a bit contrary to the challenge of an autonomous agent in an open ended environment.
It will take time for people to do that kind of work and apply these models in their domais. Similarly it will take time for someone to figure out how to potentially automate the decomposition process so they could be leveraged in more autonomous agents.