There's a tool called Codex — made by OpenAI — that helps developers write and edit code by having a conversation with it. Useful, but still just one assistant doing one thing at a time.
This week, a small open-source project called OMX (oh-my-codex) went viral for a good reason. Someone used it to take a massive coding project — porting a leaked version of a well-known AI tool from one programming language to another — and finished it in a single overnight session. Alone.
How? OMX turns that single assistant into something closer to a team. It can run multiple AI workers at the same time, each focused on a different part of the same job, and it shows you a live dashboard of what they're all doing. It also remembers things between sessions and has a command called omx autoresearch that keeps investigating a topic until it feels it has a complete picture.
The part that really caught attention: you can now mix different AI models in the same team. Codex handling one piece, Claude another, Gemini a third — all working in parallel.
You don't need to understand the technical details to feel the shift here. The interesting question isn't "what is OMX" — it's what happens when the cost of doing a week of careful, detailed work drops to one night. That changes what's worth attempting.
AI agent — An AI that doesn't just answer questions but takes actions: browsing, writing, editing, sending. Think of it as a tireless intern that follows instructions on its own.
Open-source — Software where anyone can read, use, or improve the code for free. Usually built by communities, not companies.
Orchestration — Coordinating multiple agents or tools so they work together on the same goal without stepping on each other. Like a conductor, not a player.
Parallel workers — Multiple tasks running at the same time instead of one after another. The difference between one chef and a full kitchen brigade.
If you already work with a developer, ask them: are you running your AI tools one at a time, or together? The answer might surprise you.