bel_hajo

I actually asked Claude to audit the codebase specifically for security vulnerabilities. It found a few things and I fixed them, but I'm very aware that asking another LLM to check the LLM-generated code might be the software equivalent of asking the fox to inspect the henhouse.

Then I'd probably just stare at the error message until the LLM comes back...

For context: I can read and understand some HTML and CSS, but beyond that I really can't code.

I started building LearnFrom by describing what I wanted to AI, letting it write the code, testing it, fixing what broke, and repeating the process.

Somehow, that turned into a real SaaS used by 100+ creators and businesses.

And now I'm wondering whether "it works" is a very dangerous substitute for "it's well engineered."

That's what I'd like HN to help me figure out.