Much research is going into these directions, but I'm more interested in mind-wandering tangents, involving both attentional control and additional mechanisms (memory retrieval, self-referential processing).
Memory in world models is interesting. But I think the main issue is that its holding everything in pixel space (its not, but it feels like that) rather than concept space. Thats why its hard for it to synthesise consistently.
However I am not qualified really to make that assertion.
You can do specialized SLMs with different roles working on problems. Also deterministic workflows. That is what I gathered its use. I know last year, multi-agent scenarios were topping to benchmarks but I don't know if 2025 has been the same.
Linux is behind Windows wrt (Hybrid) Microkernel vs Monolith, which helps with having drivers and subsystems in user mode and support multiple personalities (Win32, POSIX, OS/2 and WSL subsystems). Linux can hot‑patch the kernel, but replacing core components is risky and drivers and filesystems cannot be restarted independently.
Random side fact but this was also a thing map makers did back in the day. Including fake towns. In that way they could identify who was stealing their work.
I think taking key points from a session and making a new skill is less useful than "precaching" by disseminating the key findings and updating related or affected skills, eliminating the need for a new skill (in most cases).
On the other hand, from a pure functional coding appeal, new skills that don't have leaking roles can be more atomic and efficient in the long run. Both have their pros/cons.
https://research.google/blog/titans-miras-helping-ai-have-lo...
https://arxiv.org/abs/2501.00663
https://arxiv.org/pdf/2504.13173
Much research is going into these directions, but I'm more interested in mind-wandering tangents, involving both attentional control and additional mechanisms (memory retrieval, self-referential processing).
reply