Strong disagree, I use tools like DeepWiki, Mintlify, Cubic daily to analyze complex codebases. Their diagrams on PRs and generated docs have made it much easier to quickly drop into a complex codebase and understand how it works / make meaningful changes. Especially DeepWiki's diagrams + interactive "Deep Research" / Codemap feature are amazing at building flow diagrams to answer specific questions.

I recently did a complete rewrite of our browser driver codebase by deeply analyzing playwright, puppeteer, selenium, and the chromium source code, relying heavily on auto generated diagrams to decide which parts to read myself / dive deeper on. Yes they occasionally leave things out if the underlying codebase hides the complexity or has no docs, but they're great for thoroughly documented mature codebases.

And if they are GenAI they are probably wrong...

Why are people upvoting this? I thought I was missing something.

The ones upvoting are not people...

"Auto-generated"? As in... a defined translation from one abstraction layer to another? Disagree

As in not authored by a human who knows which details are important and which are noise for a given level of abstraction.

A mechanistically generated diagram, i.e. a program reading in something and then emitting a diagram, following a specific algorithm.

> a defined translation from one abstraction layer to another

A solution like the aforementioned cannot reliably present a view with a raised level of abstraction, as that necessitates either making unsound assumptions, or being provided with hints (which can be wrong or expire). Just like with code (because they are code).

And so if this is indeed just a rote visualization like that, sticking to the same level of abstraction (thereby not encountering this issue), then they stipulate it is "almost certainly useless, just like they usually are".

I personally find this agreeable, but then I rarely find diagrams particularly illuminating in general, so...