Comment by bluefirebrand

This absolutely is intrinsic to the workflow

Using a package that hundreds of thousands of other people use is low risk, it is battle tested

It doesn't matter how good AI code gets, a unique solution that no one else has ever touched is always going to be more brittle and risky than an open source package with tons of deployments

And yes, if you are using an Open Source package that has low usage, you should be reviewing it very carefully before you embrace it

Treat AI code as if you were importing from a git repo with 5 installs, not a huge package with Mozilla funding