The setup was absurd from day one. Management used AI to generate an HTML demo, then handed it to a "Super Individual Department"—a specialized unit designed to build products using only AI coding tools. There was no product manager and zero requirement documents. They simply fed the demo into an LLM agent, the AI generated a full platform, and suddenly we had a live product.
The absurdity peaked last week when the company hired a PM to write requirements for a platform that had already been built. Our QA team then wrote test cases based on these retroactive documents. My manager had been pushing us to be "All-in-AI," claiming QA wasn't adopting the tech fast enough. He wasn't wrong about the pace, but the execution was chaotic.
The platform was a mess of network glitches and random errors, but the "Super Individual" team fixed bugs instantly. It was a closed loop of AI-generated code and AI-driven fixes. I thought we had found a high-velocity AI workflow until I filed a bug this morning and noticed the group chat had gone silent.
The entire department was gone. All of them.
Every line of code in this project was generated by AI through those developers. Now, the platform is still running, but the humans who steered the AI are gone. We have a PM writing specs for a ghost ship and a pile of open tickets with no one to assign them to.
It's a harsh lesson in deployment and maintenance. I spent weeks complaining about bugs, but I realize now that having a developer—even one relying entirely on AI—is a luxury. We're now left with a "black box" product and no one to hold the keys.
Next Paddle Support: The AI Wall is Real →