Hey Product Hunt! Building with AI is easy now. You type a prompt, one model does something in a black box, and you hope it comes out right. The part I wanted to change is that you never actually get to run the thing. pondas gives you real control over a team of AI agents and the loop they work in - the harness and orchestration engineers use to get quality out of AI, made simple enough that anyone can drive it, like a game. Here is what you actually do: - Tell the orchestrator what you want, in plain words, and it runs the whole team for you - or take the wheel and direct it yourself. - Build your own team. Add agents, give each a role, and set up the loop - who builds, who checks, who fixes, and how they hand off. - Watch it happen like a game. Every agent shows a live status, you see the token usage as it runs, and you get a notification the moment a step is done. - See it live and keep going. A live preview shows the real result, and you just say what to change - again and again - until it is exactly right. - Publish in a few clicks. When you are happy, it goes live at a link you can share, and the code is yours in your own repo. The point is not that you skip the code. It is that you finally control how the AI works - the agents, the checks, the loop - instead of trusting one prompt and hoping. No terminal, no setup, no technical background needed. Free credits to try it. I would love to see what you build, and hear where it breaks so I can fix it.

pondas

Run a team of AI agents like a game to build what you want

Run a team of AI agents like a game to build what you want