4 ms·
The model is being trained to use claude code. i.e. the agentic patterns are reinforced using reinforcement learning. thats why it works so well. you cannot bui
by mikert89 8mo ago
The model is being trained to use claude code. i.e. the agentic patterns are reinforced using reinforcement learning. thats why it works so well. you cannot build this on your own, it will perform far worse