Months of chatting with ChatGPT, and somehow I'm still the one doing all the learning.
Continual-ARC raises an interesting question: Can AI actually learn from experience instead of starting from scratch every time?
OpenClaw, Hermes Agent and Prime Agent don't meaningfully learn from experience.
On Continual-ARC, our new benchmark at @letsunifyai, none of them were able to memorize anything of value and none were able to generalize to new tasks. Training a small 9B model's weights did both.