Don't trust your LLMs too much. An LLM must output a token at every step. Even if it is uncertain, the Softmax layer forces it to choose a winner. It is physically incapable of saying, "I don't know, let me re-calculate." It is programmed to be "confident."
Just checked out @NousResearch’s Hermes Agent man, it feels like living in the future, the system level integration is so great. One nice feature would be it automatically killing llama server and activating it on laptops to save power.
Hey @NoBoilerplate, after watching your NixOS video, I’ve been thinking about how NixOS compares to the direction Fedora is taking with its Atomic variants and projects like Universal Blue’s Bazzite.
Checked Universal Blue’s Aurora, An opinionated Fedora Atomic based desktop, featuring KDE Plasma
You get:
- Stability
- Rollbacks
- Latest updates
- Batteries included
Just good Linux experience, their take on AI’s cool too