I started using agents to build complete products fairly early. Those projects still had a clear finish line: ship one program, or finish one game. This time I wanted to push further, giving them a small game platform with a starting point but no prescribed endpoint. Instead of a fixed workflow, I used a multi-agent architecture built for open-ended goals and able to grow with the project: the harness handled scheduling and shared state without constraining what Codex and Claude Code could do. The agents chose the next work and divided it among themselves. Eight days later, theworldatplay.com had 52 games; I had touched very little of the code. I was surprised by the work that still needed me.
The home page was beautiful and I could not find the games
The first home page looked like a printed magazine: a big headline, generous whitespace, and collages of Connect6, chess, and Hive across the top. Honestly, better than anything I would have designed.
But none of the images were clickable. They were pictures of games, not entrances. On a phone, the page ran to seventeen screens in total, while the first place you could actually start a game did not appear until after the third.
Nothing was broken. Every test passed. The catalog, links, and metadata were correct, and the page was fast. By the project's written standards, it was finished.
