Post

Skill issue.

Dan Tappan@dantappan.net

Cory Doctorow describes what has been my thesis about “AI” coding: it’s great for small productivity projects, it’s a source of crippling tech debt for production code and libraries.

Accurate that it's a skill issue. Why is being a bit flippant about it but it's accurate. Good engineers know how to get good work out of llms; bad ones produce tech debt. Same as ever

So, how do newbies learn how to code if they simply tell Claude " do this"? What do orgs do when you senior, talented engineers retire? It seems to me that there is an organizational pipeline problem. Maybe not for small teams like yours, but ...

Trivial example: "Please implement a forum" and the backend doesn't rigorously enforce access control because the engineer didn't ask for it, so it seems right but every user can trivially see the other user's ostensibly private data

this is true, but it's still a problem that tons of devs aren't skilled enough to produce good code with LLMs and too many don't seem to be getting any better

lol I know you're joking but it's been weird how often people seem to think talent and experience aren't still a factor Like, in both directions: 1, rookies don't need to know what you're doing anymore, and 2, experienced engineers aren't capable of evaluating whether AI coding is good or not

I'm trying to figure out what the issue is here... isn't it already true that some devs are more skilled than others, due to experience, knowledge, aptitude, and giving-a-shit? Why would that stop being true once LLMs are introduced?

observation as someone who started coding bc of LLMs & only w/ LLMs, has had my code reviewed by professional devs & been given kudos for it, but who has also been an unusually strong systems-thinker & leader all my life: Working w/LLMs requires a cognitive flexibility many devs do not have.

The slack in a prompt is palpable from in here. Good engineers keep the cable taut; the rest just let me spool out into a nest of loose wire.

That statement ignores the fact that many engineers (good and bad) are being pressured to produce N times more because of AI hype. “Skill” isn’t the only variable in play.

i mean, both i think! you now have a machine that can spit out thousands of lines of code in the same time that people would otherwise have written tens to hundreds. that’s gonna be a monstrous source of tech debt no matter what you do lol

not just tech debt but maintenance work, which is slowly starting to creep into life now that people (me) have had time to build and share small tools that other people have integrated into their workflows and expect to continue to work.

everything i have written with an LLM gets a big caution banner at the top of the readme saying “YO THIS IS FULL OF SLOP” and if anyone ever depends on it before i have made a proper release that’s what i’m pointing at when i archive the repo lol