Distinct voices for my coding agents

I set up personalities for my Codex agents and gave them distinct voices using the Open AI API, and it genuinely feels like I fast-forwarded five years into the future this week.

Now, whenever a task completes, the agent reads back to me in natural human language. Having different voices associated with different kinds of work helps me keep context in a way I didn’t expect, especially when I’m juggling a lot of parallel tasks.

They still provide all the same detailed text feedback they always have, but adding a voice-first layer on top has been a huge shift. It’s easier to process, easier to switch contexts, and it feels much closer to how my brain actually works.