9 ms·
Not really. For example we still can’t get coding agents to work reliably, and I think it’s a memory problem, not a capabilities problem.
by koakuma-chan 1y ago
Not really. For example we still can’t get coding agents to work reliably, and I think it’s a memory problem, not a capabilities problem.
- atlex2 1y agoOn the other hand, test-time weight updates would make model interpretability much harder.