Post

seems conspicuous that programmer commentary on genAI is split between "oh yeah i use claude all the time now. the 10× programmer is real and he's me. i am a god" and "yeah my coworkers use claude all the time. they keep committing crap that doesn't work and now i spend half my time fixing it"

287 reposts1K likes

can't wait for the stories about the latter folks getting laid off first because they're not as Productive™

also conspicuous that artists and writers almost universally revile this junk, but there's a subset of programmers going "thank god, i don't have to write code any more" no one was making you do this?? start a food truck idk

I don't hate anybody for doing a job for the money, but also, like... do the job then??? honestly I still love programming despite it all so I just don't get it

The difference of quote between code i made for money and for love are there. “Yeah I could fix this properly but would require me hyper focusing for 3 days straight or do yet another bandaid fix that would be great for the deadline we have and im aware will only require testing in a few places”

I saw a post this morning saying they have a 1000 line prompt that then takes a few hours to generate all their code.... that sounds awful to work with oh and it generates 30,000 unit tests. As if anyone needs more than 8

I’ve watched a lot of ppl lose their love of programming bc of the incentives pushed by the corp they work for. Writing code is a very different thing than grabbing a Jira ticket for a 10 yr old, million line monolith sans unit tests, and seeing the deadline is 1/5 the time it’ll take in reality.

I'll admit I'm partially one of those, at least in corporate. it's mostly that it is good money and I already had the degree and the job before I realized (or let myself admit) I didn't like it that much, so now that's where my career is. I still have too much pride to use AI, I'm dumb not a coward.

What I don't get is, basically all programmers used to hate reading and verifying and testing code even more. And now they do basically only that.

Back in school I was following an embedded systems track. During mentor time, someone asked our mentor whether you could get your diploma if you couldn't program. His answer boiled down to 'I did and here I am now'. To my complete bafflement.

Someone on Mastodon said we'll get Schrodinger's jobs: you can't do the job remotely and you must go to the office, but simultaneously it can be replaced by AI. (They ended up deleting the post because they got overwhelmed with the >1000 reposts)

You were never getting fired for being "not as productive" They fire people they don't like, that's about it In my former company, they recruited some dude without experience and without a degree because of it's last name Race, name, religion, where you live: those are the criteria

Any sort of amateur coding community like game modding is getting unbearable, I can't work with anyone else's code. If you want something "that just works", it now means Claude instead of brute forcing copy-pasted snippets until you accidentally learn something

oh my god i cant imagine how bad it must be in those communities. they were already pretty toxic with a slight distaste for programming and just wanting it to work, but now it must be a whole other world of hurt aaaaaahhh im so glad i only barely do MC modding now

I think the biggest differentiator is how robust your tooling and dev practices were before GenAI. If you have strong test harnesses and suites, aggressive linters, etc, then LLMs are constrained enough that they help. But most companies don't have that, so the code rot builds up FAST

i keep trying this AI coding shit and I keep wasting my goddamn time Even for code I don't care about it wastes my time triangulating a non-solution My favorite is when I told mistral to use algorithm A; it wrote comments saying it was implementing A; what did it implement? Not A!

This is unironically why I keep my AI Usage strictly to "find that stray comma I somehow put there that is fucking up the entire site", or more discussionary shit like "how *should* I go about implementing this". yk, stuff where it isn't doing stuff for me.

The way I see it, If it can't count X amount of Y, or realize that a certain unicode just DOESN'T exist, why should I trust it to create code that makes any sense on real projects? It's helpful within it's limits. just too many people abuse the shit out of it.

almost like bad programmers were already bad at realizing that they were bad programmers and this has just made them worse

One of my coworkers wrote some Python code in a weird, roundabout way. I asked him why he didn't do it the obvious way. "ChatGPT said it wouldn't work." The obvious way works in the Python 3 version we're using. It works in Python 2.7. Apparently, "Just try it and see what happens." is a lost art.

Not only did I have to try it for him, but I ended up distracted by the question of how the hell this misinformation got synthesized, because I couldn't find any instance of anyone or anything making such a claim by just searching the web.

The task in question was "define a property on a NamedTuple". Somehow the Statistically Average Coder had generated the assertion that the property has to be patched onto the class after it's created.

lmao what technically it has to be on a subclass, but basically everyone just inline-subclasses namedtuples anyway