Vera Yan

And claims like 'thousands of coding agents collaborating to build software'? Another layer of BS.

Anyone who’s experimented with multi-agent systems knows that as the number of agents increases, coordination complexity scales exponentially. More agents often mean more chaos, not efficiency.

Contrast that with AI companies making grandiose claims, like 'describe a feature and we’ll auto-generate the PRD, architecture, and code.' That’s the kind of idea that sounds amazing but is almost certainly BS.

Now when I evaluate an AI startup, I look for those solving small but real problems through AI automation. If the solution integrates seamlessly into existing workflows, it’s often a sign of an idea that can sell.