Three Key Skills Emerge for Advancing AI Expertise

Btw this is for people who are just using chat

@MatthewBerman For curiosity: what do I need to know about “how to use Skills”? I mean, isn’t that something we add to an agent set up and then an LLM will know how to perform a task better thanks to that Skill?

@MatthewBerman What about RAG, memory management, MCPs ?

@ErRahul337 These are all advanced. Mine are for people starting from just using chat.

@MatthewBerman the model behavior one is carrying the whole list tbh

most people skip it and wonder why their prompts keep missing

@MatthewBerman What about loops

@brayaON20 Yep!

@MatthewBerman fourth one would be allow your llm to use powerful tools, tool calling make llms 10x more powerful, give your llm a computer of its own

@MatthewBerman why browser control? It takes up so much tokens and is hard to use

@0x_lun Yeah.

@MatthewBerman Thanks for clarification 🙏

@MatthewBerman Browser control is the one I haven’t used yet. It seems like the internet was built to prevent bots

@MatthewBerman this is so on point

@MatthewBerman Browser control is becoming one of the most practical superpowers in AI work right now. Being able to make the browser do complex tasks autonomously changes what is actually possible without building everything from scratch.

@MatthewBerman Matt let the agents in the browser

@Alex_FF Advanced

@MatthewBerman Exactly.

@MatthewBerman Model behavior is the one that never makes the listicles. Everyone talks about prompting, nobody talks about the fact that the same prompt gives you different reliability across model versions.

@MatthewBerman browser control on codex cli is cool

@brayaON20 @MatthewBerman isn't skills just .md files explaining the model how to use something specifically?