Playing more around with agents and multi-agent workflows. Figure the LLM space will just move at own uncontrollable pace but the the paradigm of having Agents talking to each other to solve problems is very powerful. Tried a couple of ideas around code generation and the output is better than just asking GPT-4. The issue is that GPT-4 seems the only good model to use for multi-agents (others get stuck in a loop) hence cost can become an issue.