Bugs have exploded with AI. Here’s what I’m seeing.

The numbers are brutal.
IBM once reported that fixing a bug in production can cost 30x more than fixing it in development.

And now with AI-driven development, the problem is worse.
I read last month that bugs have increased 41% because code is shipped faster than ever.

In our private beta of Sopa, we’ve seen the same pattern:
- Most bugs don’t come from “bad code”.
- They come from features that don’t match product requirements.
- Or from edge cases nobody thought of.

That’s exactly what we’re solving: reviewing PRs automatically and checking if code actually delivers what product asked for.

Less firefighting. More confidence before merge.

Curious to hear:
- Have you noticed more bugs since your team started using AI tools?
- How do you catch them before production?

Happy Wednesday 😇