ARayOutOfBounds

An agent reviewing its own output has the same blind spots, the same training-data gaps, and a strong bias toward declaring its work "good". But even having a different model review the code is flawed and not good enough, especially for those whose code can't be wrong, like in regulated industries.

Commenting because I'm interested in the answers as well.

Stats say over 90% of engineers use AI to code, but the actual split of manual/AI is harder to say.

Meaty thought piece on why relying on just AI reviewing AI is a bad idea.

Yes?

Codacy has done this for a while https://www.npmjs.com/package/codacy-cli

And this tool makes it Claude Code native https://www.npmjs.com/package/@codacy/verity-cli

Appreciate it.

Actually, no. This was conceptualized and written entirely by a human. :)

Interesting, and how does it work with other tools such as mailchimp?

And I guess that's the whole point, these types of accessibility issues keep being left as an afterthought rather than the standard

I remember having to spend much more time on the existing search engines trying to find what I was looking for, even though there were much less sites. Since Google, sifting through all the data has become easier, although I worry about how it's also become influenced by our search history and navigational habits. This in turn, lands up limiting what we are actually exposed to and might reduce the probability of us discovery new things outside of our comfort zone. Do we go for something like DuckDuckGo, that isn't influenced by our history, or do we use google to present us with what we want?

Very interesting video! Thanks for the share! It's a point I'd never considered...

Big brother is listening... spooky thought, but definitely a possibility in the near future