5 ms·
I think that’s very reasonable but the ai companies are intentionally training them to work on harder and harder problems just beyond their capability. So if th
by tehjoker 4d ago
I think that’s very reasonable but the ai companies are intentionally training them to work on harder and harder problems just beyond their capability. So if they do that, they’ll give up too easily.
Do a breakthrough, make no mistakes
- dgellow 4d agoWhile also using harnesses that will execute any tool call with full execution rights. And no supervision. And with a prompt context that autocompact, meaning it will degenerate over time. The whole thing is designed be a complete disaster