I've said it before but if LLM assisted coding tools are going to retain significant usage over time (as the cost of those initial, apparent short term gains becomes clearer) it'll be dependent on better supports for people to read and understand codebases, not for LLMs to read and understand them, for human beings to do that.. they'll need to let go of the delusion that all of this can be automated
I guess part of the reason that reading code can be harder than writing it is that what you really need to understand isn't necessarily *in* the code, you're kind of inferring the understanding you need *from* the code.
Again, we keep conflating the work with the artifact of it, and the value of code comprehension illustrates why that's a mistake.
If you have a fediverse account, you can quote this note from your own instance. Search https://glasgow.social/users/sue/statuses/115601035968686856 on your instance and quote it. (Note that quoting is not supported in Mastodon.)