@nate_512CS329A Self-Improving AI Agents is the new Stanford Online course and the syllabus is not playing around. Self-correction runs through the whole thing instead of basic API wrappers, and frontier reasoning mechanics get the main spotlight. The stack they teach is the one people are actually using: verifiers and test-time compute, RL and Constitutional AI, then planning, memory, tool execution, plus architectures for deep research agents. Project-based from start to finish. You build alongside the researchers while their lectures run, and the end result is a functional self-improving architecture you can point at. Full playlist is in the thread 🧵

@nate_512CS329A Self-Improving AI Agents is the new Stanford Online course and the syllabus is not playing around. Self-correction runs through the whole thing instead of basic API wrappers, and frontier reasoning mechanics get the main spotlight. The stack they teach is the one people are actually using: verifiers and test-time compute, RL and Constitutional AI, then planning, memory, tool execution, plus architectures for deep research agents. Project-based from start to finish. You build alongside the researchers while their lectures run, and the end result is a functional self-improving architecture you can point at. Full playlist is in the thread 🧵