I understand AI is polarizing, but it's just a tool. You don't have to try to vibe code or give a coding agent unrestricted access to a system. And you can make software better using models to multiply force if systems are properly constrained.

Yes, there are a lot of controversial takes on "AI" and many have merit. LLMs may not even be the most useful AI tool, at least it's unlikely the best to simulate consciousness. I find LLMs to be useful tools and try to learn to use them more efficiently and appropriately. They're just fancy. token calculators after all.

If you're into systems here's how this one is configured: https://github.com/verificationdesign/verificationdesign