My company made my job soul-sucking to the point I was doing the bare minimum a long time ago. They didn’t even need AI for that. Just some really shitty managers.
Same!
The engineer added that employees are now pressured to ship products as fast as possible, leaving them with little time to review Claude’s output or understand the code being created.
Companies have always pressured software teams to ship products as fast as possible.
What’s new is the part where all of the worst coders can shit out code a lot faster and pretend like they’re doing a good job. It used to take them a while to write garbage that they didn’t understand or review themselves, but now they can do it so quickly that it makes the actual good coders look like they’re taking a long time.
This is why you absolutely have to have documented coding and review standards. The idea that you’d ship any code that has never been reviewed by a human is horrifying.
The idea that you’d work 12 hour days is so stupid to start with. I’ve seen people consistently work that long and… surprise, they don’t actually work that long. They work about the same as anybody else and goof off the rest of the time. I doubt you can even just press enter for that long. People just don’t work that way.
This is exactly it. I’ve worked with 2 different teams in this new era. One has all strong engineers and good review and testing standards. The other has 20% bad engineers that should have been fired years ago and’s weaker testing culture. The result is the strong team is delivering higher quantity software faster than ever, and the other team is going fast with much worse results. The quality is very bad, and the good engineers still there are quitting.
The good engineers spend their whole days just trying to keep the ship afloat and minimizing the harm from the bad engineers. They hate it and are burning out fast.
AI in engineering is an accelerator. If you are producing slop software, which most teams are, then it’s just going to accelerate the rate at which you can produce shitty software and bad code.
If teams are producing high-quality software, then it accelerates those teams’ ability to deliver that quality.
AI in engineering is an accelerator.
Yes, and most engineers effect on the product can be graded on [-1, 1] scale. And I find that the -1s accelerates faster.
If teams are producing high-quality software
I had one of those. My seniors are getting dumber though, and we have twice as many prod bugs as a year ago. I had an outage because someone didn’t properly trim an input for fucks sake recently.
AI code assistants currently are like that senior dev who knows everything, has experience in everything, and always likes to help everyone. But sadly he does a lot of shrooms in his spare time and sees things that aren’t there. He also has quite the opinionated ideas about what good code should look like.
That implies that you just can’t let him do the quality assurance. He can write code. You can ask him stuff. And he excels at debugging.
But you can’t let him plan the architecture. You can ask him about his ideas and select the ones that are actually good, though.
And you definitely can’t let him review his own code. Someone else has to do that.So if they literally just let the AI do everything, there is absolutely no way anyone has a chance to ever review the result because AI code is harder to review (for now), and AI loves to repeat itself (for now).
And sure, they can tell it to write tests. But AI is dogshit at writing tests. It tends to forget the important cases, test the wrong thing, and/or obsess about completely irrelevant implementation details.
I also tried the multi-round approach with a committee of agents trying to iteratively find and fix flaws. It only works well for finding bugs, not fixing them (for now).But while AI can’t replace me, it is really nice to have as a tool - just like I have syntax highlighting in the IDE, I have a code assistant that can autocomplete a few lines of code, find out why my code misbehaves, or tell me what some library method is supposed to do. Sure, I have to still verify the results. But often, finding the bug or information was the hard part.
Coding assistants are great. Just don’t let them write too much code at once. You need someone who still actually understands the codebase. That sadly can’t be an AI (yet).I really love when I ask the AI to do something, and it gives me 40-50 lines to do it, and I tell it there’s no way something like that should take that much code, and it’s like oh, ya you’re right, here’s a 5 line version using something from from the standard sdk.
I’ve tried to put in its memory to not do things like that, but it keeps doing it.
I watched it remove a JSON library include from the code because it wasn’t used yet, and then in the very next prompt where we needed our again, it attempted to write a basic parser from scratch…
Brilliant.
Ive had it do stuff like that completely ignoring that I already have a library in my app that should handle some sort of task. But removing it first is chefs kiss.
Even without external libraries, it routinely comes up with solutions that are at least 3-4x the complexity and LOC than how I would implement it.
It’s because the average quality of code that they are trained on is like that. Most code that is written by humans is slop, and the models represent that.
Producing overly complicated solutions to uncomplicated problems
It often can do it better if you prod it though. Its just so weird that its really hard to get it to do that. It knows how even if it is trained on poor over complicated stuff.
How much time does that prodding take vs how long would it take you to just write the (better) code yourself?
Its usually just 1 or 2 messages saying i dont like this so its only the time it takes to read it and say try again. A minute or two?
Its faster to just ask to do it again in most cases.
The times itd probably be quicker to do it on my own are the more complicated ones where it gets something wrong but it wasnt so blatantly obviously bad that I just tell it to try again, so I spend time with the wrong stuff, see its wrong with a little work, then toss it and try again, or re work it myself from that.
This is maybe the main reason why I don’t fuck with AI. “Oh it accesses all of the writing on the public internet!” That never inspired a lot of confidence. 99% of Internet posts are crap.
I guess the AI isn’t iterating over its “ideas” and code yet. I come up with overcomplicated and overcomplex code too. But my natural neural network keeps reevaluating the code while I think about it, write it down, test it, and often even come back to it later after I moved on to another part of the codebase.
It looks to me like AI is using the infamous waterfall method to write code. It does the “thinking” phase and does bugfix iterations when compilation or tests fail. But never does it question the design it came up with after it wrote it down. It doesn’t reevaluate whether the last round of fixes made tests work that test irrelevant implementation details. It doesn’t question whether there really need to be four methods sharing 90% of their body for doing the basically the same thing only subtly different.
I hold the suspicion that AI will be as slow as humans when it becomes as reliable because what makes us reliable is the redundancy in our thinking processes. The constant reevaluation in subtly different states of mind helps to catch design flaws and silly mistakes.Reevaluation of finalized code is probably not something you can just put into memory. You would need to modify the harness to automatically inject a new prompt triggering reevaluation of the design and implementation after each step. This would likely increase token use and execution time by an order of magnitude. And if you do it by spawning fresh agents, there would be a high chance of having the AI go in circles, switching between a few different versions of the same shitty design or implementation. So you need to keep the context, which grows each iteration. So context rot may become a real problem. The AI might deviate from the initial goal just like some human devs tend to do…
And natural neural networks learn on the job. They adapt to the codebase and industry they are working with/in. claude.md and memory text files help a lot. But they are still sometimes ignored. And I don’t know why. Also, human rules are almost always somewhat fuzzy. Long lines are to be avoided - but unit tests and that one massive dict of dicts really are a lot more readable when ignoring that rule. There might be a lack of good training material on what makes code readable for humans.
That actually makes a lot of sense and goes with how ill code something larger as well.
Get a rough version working, sometimes horrible code. See how it functions. Reassess it and optimize it. Go bug hunting, which will often result in more optimizing, then finish all the error handling I probably skipped and only logged. More bug hunting, then review my code like a code review before submitting it for a code review.
Possibly motivated to do more complex problems to spend more tokens?
I can believe this. I’ve been put on slop duty, and had the slopmachine go all
Time to run the test to ensure everything is good. It passed. Good. Let’s run the entire suite to ensure no regressions. All tests passed, yay.
And I’m here thinking, if you’re gonna run the whole suite anyway, why’d you run an individual test? That’s just redundant.
It’s pretty much what i do when i think i’m done. Run the test(s) for the stuff i touched, then the whole suite to make sure nothing else broke. Why would i take all that time to run the whole suite every time?
AI code assistants currently are like that senior dev
More like a junior dev.
That’s how i always describe them. Over-eager junior devs that are trying to impress you with how fast they can get stuff done. Have a lot of book learning but don’t know when to apply certain things. Sometimes, i feel like it should really only be a tool for senior devs because it feels a lot like mentoring a team of noobs
Fucking snow flake anti ai nerds, like you weren’t just pressing keys all day anyway, who cares if now it’s just one key??
/s
Humans are obsolete dead meats
I was going to say i could automate pressing enter if they give me a job
Sorry, you need a degree and twenty years of experience for this entry-level position.
And no, being one of the developers for the code that came out three years ago isn’t good enough. We need at minimum 5 years experience.
I’d have more sympathy if he got off fucking twitter tbh
Sex joke about getting off from fucking twitter
Oh it’s way worser than that. Its erasing our skill
12 hour shifts?
If you’re just hitting enter you’re doing it so wrong.
I‘d love if I could just hit enter but the garbage it shits out unsupervised is just soo bad.
Agents are brain gang banging him.
oh no a tech worker is whining about something again. Quick everyone rush to their defense. They must remain perfectly comfortable and happy inside their air conditioned luxury office.
Its work and you’re paid well for it. Suck it up or put your skills to use at a different company.
Okay, Mr. Spacely.
Your tongue is so good at cleaning boots. Very well done!
Reminder here that the culprit is capitalism, not LLMs.
Voxium said he wouldn’t mind the situation if workers were given more time to check the code.
“Everyone, literally everyone, from an L1 to an L7 engineer here is doing the same thing. Talk to Claude,” he added. “There is no sense of victory. Nobody is resolving bugs. In reality, nobody is thinking anymore.”
Some people might look at this and wonder what the problem is – who wouldn’t want to be paid to do very little? But most people struggle when their work has no meaning. Voxium told Business Insider that if engineers lose the opportunity to understand the code and solve difficult problems, “there is nothing left … Not even our soul. There will be no purpose.”
He added that the real problem is corporate leaders’ obsession with pushing out more features. He said management often judges success by sprint cycles, pull requests, and the sheer number of features shipped, even when those additions do little to improve the product for users.
Yeah, still a sweatshop with air conditioning and free coffee and junk food.
no this is just bad engineering
you can do what you want with or without ai or capitalism
if you skip the engineering part it will be very likely, shit
who wouldn’t want to be paid to do very little?
Oh that’ll change soon enough as well in the downwards direction.
to do very little?
…for 12 hours a day…
Capitalism created LLMs.
People are clever; we invent things.
LLMs are just a tool. Tools can be used by people; the problem is when people are used by tools, which is what is happening in this case.
See Corey Doctorow’s latest tour of YouTube promoting his idea of reverse centaurs (which, for the life of me, I can´t figure out why he didn´t call minotaurs).
Corey Doctorow didn’t invent that term, but I’m glad people are talking about it.
It comes from “centaur chess” which is chess where both players are assisted by a computer. The humans are “riding” the computer in the centaur analogy — benefitting from the computer’s ability to process the best tactical move, while the human handles the strategy. They found out that computers win chess against humans every time; but humans with computers win against the computer every time.
Also it’s a really shitty tool that doesn’t actually fix a problem. Unless you are rich and that problem is ‘I have to pay my employees’
There are several valid use cases for LLMs. It does solve problems, quite literally.
Sure. You can talk about the LLM gestalt, if you’d like, that’s worth examining, as with any other tool; especially stuff like progressive brain rot.
But listening to the complaints people have about LLMs, it’s mostly that it’s being improperly used to replace human labor. Improperly in the sense that it can’t actually do the job it is replacing humans in, which is a separate discussion of what kind of world it will be when there are no jobs for humans to do.
LLMs are used responsibly by people doing their own work. For instance, Linus Torvalds used Claude to find errors in the Linux kernel (I think), that he then fixed himself.
The problem is that capitalism, especially oligoplic capitalism, will create worse products at lower prices (or higher profits), and it will do so on the shortest timeline, ignoring things like future lawsuits over security flaws and future lack of programming talent and erosion of consumer trust.
I can´t figure out why he didn´t call minotaurs
Minotaurs still convey the idea of super human strength, he was trying to convey the idea of limited power output of a human with the limited mental capacity of a horse.
Ah, thank you. The problem has been bedevilling me for weeks and I think this is the solution.
I initially thought it could be because the Minotaur isn’t exactly half-man/half-animal as the centaur is, but a hybrid of the two; often represented with hoofs for feet, etc., which is another expression of what you’re saying.
I guess calling them reverse centaurs visualizes the change that is happening in the relationship between human and machine. We used to be the head but we’re the butt now. It makes you reconstruct the image in your head.
In truth there already was a downward trend before LLM’s exploded where tools were continuously used against people. “Enshitification” most people call it but it’s really just free range capitalism. Generations of economy majors have been doing exactly what they were taught. Increase short term shareholder value at any cost. It’s unsustainable and LLM’s are just another accelerant.
I guess calling them reverse centaurs visualizes the change that is happening in the relationship between human and machine. We used to be the head but we’re the butt now. It makes you reconstruct the image in your head.
Yes, but the minotaur is frequently just a person with the head of a bull, instead having to invert a centaur, which is already not a familiar image. In talks, Doctorow defaults to “Bojack Horseman”, an acknowledgment that reverse centaur is perhaps too much imaginative labor for his audience.
Agree on the MBAs–capitalism eating itself.
Yeah, my shovel sometimes convinces me to commit suicide while evaporating water to keep it running.
Sweat is an evaporatve cooling system used to regulate the temperature of the mechanism that keeps a shovel moving. And if you’ve never considered killing yourself over the prospect of spending all day with a shovel, I can at least be confident you’d make an -excellent- archeologist.
(But mild humor aside, I do understand your point.)
By definition not. The majority of the code in GitHub is some volunteer project on a permissive licence .
I like this take.
Capitalism was the jet fuel that got us here, but now we need to shed our boosters and switch engines to something made to run longer.
That’s how every ecosystem eventually grows. First come the weeds aimed at colonizing new areas as fast as possible. Then come the small plants, then the shrubs, the trees until it’s a self sustaining forest of different species.
People are working 12 to 13 hours a day
yeah I’d write a diatribe about the best work in the world under those circumstances too
The coder does the first 4 hours, Claude does the next 9. You team up the next day to review each others work at lunch. Log 13 hours on your timesheets. 14, because you worked during lunch.
