This looks really amazing! I have to read it carefully and probably with some help from Claude, but the demos are 🤯 a computer inside an LLM that allows it to do long horizon computation reliably 🤯🤯🤯 what?!

Cuando veo todo lo que estamos construyendo encima de los modelos de lenguaje (MCPs, RAG, skills, agents, plugins, rules…) me acuerdo de esto que escribi ya hace casi una década: juanmirod.github.io/2017/09/13/d…