Comment by freetonik

> with the claim that we should move from “validating architecture” to “validating behavior.” In practice, this seems to mean: don’t look at the code; if tests and CI pass, ship it.

But... tests and CI are also code. It may be buggy, it may not cover enough, etc. It is also likely written by an LLM in this scenario. So, it's more like a move from “validating architecture” to “LLM-based self-validating”