"AI" generated code is everywhere. It's impossible to avoid. We can see it. The only thing using the tools would accomplish is expose us to a bunch of cognitive hazards that come with chatbot bullshit.

(It's the delta that worries me in the code I see—the change in quality, that is. Engineers with a quarter century of experience churning out code by the bucketful that's quite a bit worse than what they used to write. So either they aren't reviewing their own code like they used to or their skills have degraded.)

@baldur This is precisely why I revived KScope for Qt6 after years of it languishing after the Qt4 port bit-rotted. The tool exists to aid human understanding. It is not relevant to any so-called "agentic development paradigm", nor is it intended to be. I continue to be more productive by rejecting LLM driven tools.

It is also 3-clause BSD licensed except for the forward port of KFileTreeView, which is LGPL2. There is no requirement for me to release source code which might be subject to IP theft via "training". A prototype Linux AppImage exists.

@baldur Or likely both. But we can fix all that by letting another LLM review the code 🤡

@baldur I'm pretty sure it's both.

@baldur I still remember (in the early days of chatbots) my brother insisting he reviews 100% of the output himself. Nowadays he doesn't even bother pretending anymore.

@chapter1 @baldur I took a pair programming class a couple of weeks ago, and one of the exercises was for two humans to pair with copilot, which I hadn't used before. It was striking how quickly it felt that actually reviewing the code was an unnecessary bottleneck and I was holding things up by spending a couple of minutes reviewing what had been generated in a couple of seconds. I felt like I couldn't justify making the other person wait unless there was an obvious mistake.

@trantion Strong opinion: if review feels like the bottleneck, review is the only part still doing engineering. Speed up everything else, never that. I built AyCode around it: agents run in parallel, you read every diff.

@trantion @chapter1 @baldur

The most critical failures in any system - particularly in code, where there is more than one way to achieve an objective - are the NON-OBVIOUS mistakes!!

The obvious mistakes are normally trivial!

@baldur The lack of self-review is the thing which really really upsets me about this torrent of slop - if the operators of the tool were at least as proud of the *output* as they are of their *use* of the tool then we might not be quite so inundated with crud.

@kinnison @baldur A generation of kaleidoscope-owners convinced that it makes them the most gifted stained-glass artists known to the craft.

@kinnison @baldur also suffering from this. And I do not know solution. E. g. customer needs feature by tomorrow. I can do it by tomorrow with agents accepting “eagle eye review” and customer happy. Or I can deep dive, understand things, use agents, have clean code and no customer.

As a passionate engineer I HATE it. Only ~10% of tasks I can do maintaining decent output quality.

Sadly, it’s not even about engineering. It’s unavoidable outcome of capitalism+LLMs.

@baldur reviewing code is hard - it is likely that no one is capable of reviewing slop and revising it up to "written by a good engineer" quality.

@baldur The most perplexing ones are those who confidently state that the models now produce code better than humans. Either they are all bots, or there are a lot of devs who have no idea what good code even is. The last year has been the worst in my career for encountered code quality.

@danielleigh @baldur My bet is on most of the humans they've worked with wanting to keep their professional integrity and keeping their principles.

@danielleigh @baldur I'm guessing you haven't worked at a place where the majority of code is written by people who just graduated or are still in college, the bulk of whom get fired/promoted somewhere else and replaced after three-six months. But this is how software is written at scale these days. I'm not defending the practice, it just is what it is.

@baldur
i've been strucked by what was saying a Microsoft engineer in one article on vibe-coding, that they were so busy prompting they didn't review anymore, lack the time to do it..
(noted that detail but don't have the source and don't remember it)

@bituur_esztreym I think I came across that same article recently but don't have at hand either.

@baldur are they even given the time to review..... velocity of code seems to be the new token maxing

@bosconet That's basically what I've been hearing.

@baldur @bosconet
Heard that "The AI stuff revealed to us that for the triangle of Good-Fast-Cheap, CEOs everywhere would choose Fast twice"

@baldur I had someone who I had respected quite a bit tell me that AI-generated code no longer needed abstractions that makes the code readable and I found that to be dumb in the face.

@baldur the way I see it you can let an LLM generate code or review it but not both. As a human I'm better at writing it and wrong review suggestions are way easier to ignore.

@baldur Yup, it is everywhere. So is Climate Change. AI is Climate Change denial.

@baldur If the change sets for a supposed trivial refactoring with LLM "assistance" are so complex that I must switch to a new-generation semantic diff tool, rather than a standard unified diff viewer, code flow has suffered, not improved.

@bms Right!

@baldur "If you haven't tried it, how can you know" is one of those takes that make me reconsider the sanity of the humanity.

Deducing something is bad before trying is one of the main purposes of the brain.

@petrikas Yeah. Learning from other people’s experience has historically been key to survival.

@baldur @petrikas And there's so, so many people eager to depreciate it! Even in the form of my elderly relatives replying to insights from me doing just that with a "you're inexperienced, you don't understand it".

Like, I get it, we all need to prop up our egos somehow, but can we do it in a less self-sabotaging way, please?

@baldur @petrikas some people do love to reinvent things from first principles while making all of the novice errors tho; a week in the lab saves an hour in the library, etc.

@petrikas @baldur They usually understand it when it's about drugs or gambling. I don't need to try heroin to know it's bad. I also don't need to lose all my money to slot machines to know it's bad. But somehow, they completely miss the connection with LLMs. 🙊️

@petrikas "Let's not eat at this restaurant, there's tons of negative reviews about getting food poisoning."

"BAH! You can't trust reviews! You need to try it for yourself and form your own opinion!"

Okay have fun spending an entire day shitting yourself. 🤡

@petrikas @baldur

I found myself agreeing, but then I had to stop me, because I think the difference is in the consequences and the cost.

Because in the case of LLMs, it's at worst just a rather cheap waste of time, which is bad, but (in moderate doses) fine, if that means one can end up verdict that is not just based on expectation but actual data collected in an open trial.

Of course, the people referenced here do not mean in that way. They mean it in a dismissive way. But just because they're weak, doesn't mean that we should also counter in weak dismissive ways.

Especially given that an extensive list of all the bullshit one has personally seen the clanker do is an excellent tool to hit tech bros over the head with.

@Hypfer @baldur
"it's at worst just a rather cheap waste of time" - incorrect and rather short-sighted.

@petrikas @baldur

Look man, I'm just saying that to win over people that aren't already aligned, one needs these stacks of empirical data. And especially empirical data that matters to them.

I can also just not reply if that is more wanted, but it showed up in my timeline and I thought there would be some actual interest in doing something about the state of the world being bad.

@petrikas @baldur

I mean look at these guys:

__

jamesponddotco 4 minutes ago | root | parent | next [–]

I gave Claude several pieces of writing from myself, from notes to code comments to iMessage messages with my wife[1], and asked it to write skills that match my voice. I also incorporated the humanizer[2] skill into it.

Worked like a charm. This is the prompt I used[3] to write a version for the company I own with my wife, if it helps anyone. The personal version, which lead to the `jamenizer` skill, is very similar, but has some personal details in it.

___

I doubt one can reach them by talking about.. idk the environment or whatever.
If they are even real people at all and not just ghoulish caricatures.

Anyway, I'll leave you alone now

@Hypfer @petrikas @baldur I predict that won’t end with it having just been a cheap waste of time. It sounds more like the prelude to a messy divorce.

@Hypfer
"in the case of LLMs, it's at worst just a rather cheap waste of time"

I have a feeling that the people who had a data center constructed near them and are having their lives ruined by air, water, and noise pollution would strongly disagree with your assessment of what the worst consequence of using LLMs is.

@petrikas i credit the fact i haven't been normalizing it in my brain with direct exposure for why im still able to spot the glaring, obvious, intern-level errors it makes when asked to review colleagues' PRs (which is only sometimes, as i become more and more grumpy) that they oftentimes haven't even read themselves

yes, i can quite clearly see what it actually does

@baldur

@petrikas @baldur Look, you wouldn't want homicide legislation to be written by people who haven't even committed two or three murders themselves, you know?

@petrikas @baldur

Been a trope in the AI booster circle for a while.

If LLM coding assistants aren't working for you, you have to be doing it wrong, you must be using an outdated model or you aren't prompting it correctly.

Its all your fault!

And just you wait in 1, 3 or 6 months from now these things will be soooo good it'll be transcendental

From Thomas Ptacek's article "My AI Skeptic Friends Are All Nuts", Published June 2025

"First, we need to get on the same page. If you were trying and failing to use an LLM for code 6 months ago †, you’re not doing what most serious LLM-assisted coders are doing."

@petrikas @baldur I've heard this dressed in the scientific method:
"You know, as scientists we need to verify stuff"

But I'm not seeing anyone volunteering to peer review the physiological effects of cyanide through personal verification.

It's all distraction, no amount of evidence is ever going to be enough. Even if you go there and use it, if you don't like, you're just doing it wrong, using the wrong model, not using enough etc etc etc etc

@baldur I have been involved in reading and putting in practice standards specifications for 15 years. I don't need to use the magic lamp to know how people screw up when putting in words what needs to be done.

@baldur ah but you're just observing the drug addict from the outside. It's a totally different experience from the inside!

@baldur

> You do know that we can see the output without using it ourself, right? Some of us are still reviewing pull requests and reading through code repositories, y’know?

Thank you for putting into clear words what I was not able to explain to my colleagues.

@pizzaclick 👍

@baldur Those pull requests must have been done by someone who Prompted It Wrong™

or something like that

idk I'm not a slop defender

@baldur Exactly. Reading the PR is using the tool, just not the dopamine loop. The 'everything changed five minutes ago' line mostly means 'please don't judge the demo by last month's output.'

@aycodeai @baldur also, I don't particularly want to have to relearn my tools every 5 minutes. I am very happy to continue to be productive with tools I learnt 20 years ago.

@baldur Observing drunk people while sober makes me want to stay sober, not get drunk 🤷‍♂️

@baldur My 2 cents: I think this is mostly behavior that seeks exposure (think code influencer), meant to either trigger you or use the "fear of missing out". It is sad that not many people do this for the "love of the game".

@baldur What's honestly even worse is juniors using AI making such statements towards experienced devs and countering a:

"I have used it, but I get better results using it as a helper for mundane tasks"

with a simple

"Well, then you just don't know how to use it - my code is better letting it do the job!"

Congratulations; You just highlighted that you didn't learn anything and why people try to replace you with it.

@baldur but charlie brown, I'm putting better and better footballs out every day!"

@baldur

Yeah, but Baldur, you don't _know_, man. Like... You just... You can't. And if you don't agree it's not that you _understand_ and are simply opposed to it... You just don't _get_ it, man. Like, you know?

🙄

The number of people which I no longer feel like I have to take seriously grows exponentially _every_ day.

<3

@401matthall Yeah, same. In a way that's doing us a favour. Less work to keep up with things that way. 😁

@baldur

So, much this, sir. Sooooo much. <3

@baldur Now that we’re a bit into this hype cycle it’s also fun to see that the same people who said something along these lines a year ago are now saying the same thing but just with an updated timeline.

”You don’t get it, everything changed in March 2026, before that I would have agreed everything from LLM’s were shit.”

…said by the same person who praised the progress in November 2025 ”since February 2025” or something 🤷🏻‍♂️

@HaTetsu

@baldur They've been using that mantra for the past year. It's a kind of Stockholm Syndrome.

@baldur dude reading is so last year.

@quinn Sooo passé

@baldur "everything has been different now" for the last few years.

It's almost like it's not actually different.

@baldur Tried it on some embedded code this weekend. While it did work the thing puked out about 10x the lines a good human coder would have. Nothing elegant and no consideration for the specific application. I would say it took as long to just review the code than if I had created it from repo examples myself.

@baldur
On the bigger scale, the claude-reversed engineered AutoCAD and such crafty tools do not make me forego all security, attestation and certification guardrails around large scale construction projects.

If there's technical debt in there (sure there is, it's in human engineered software too, you know) someone will find it. Either it's you, early on, or the bridge comes down in chunks and injures many people who relied on it.

I get the excitement and enablement. But am I going blindly forward? No. :)

@baldur This happened at work. AI Hyped Coworker: "They set the agenda for the workshop!"
Others: "... yes, isn't that normal?"
AI Hyped Coworker: "The workshop is 2 weeks from now!"
Others "...oki?"
AI Hyped Coworker: "This agenda is going to be completely outdated two weeks from now!!!"

I guess, based on the 5 minute rule, by the time a workshop is over, it will already be outdated, and yet if we don't attend, we will be "left behind".

Can't they just leave me behind.

Please...

@baldur 2026: BRO THIS TECHNOLOGY IS SUBJECTIVELY INCREDIBLE BRO

@baldur Aye. There have been meaningful changes, but the way I put it to people is: the November 2025/Opus 4.5 shift was from “the stuff they spit out doesn’t usually work” to “the stuff they spit out usually works… but it still isn’t *good*”.

It’s quite plausible to me that that change has *worsened* the quality of code coming across the line, because “works” is enough for some folks to say “ship it”. When it just didn’t work, you *had* to actually touch it, which may have helped… somewhat.

@baldur I would say (and in the past *did* say) that this was just a new version of the classic automation paradox, but the *classic* automation paradox involves higher performance than the baseline under normal circumstances. There are axes on which that is sometimes true here—adversarial review, e.g., surprisingly often does better than manual code review given existing code review practices (That’s its own topic of broken engineering practices, of course!) — but it’s not remotely all of them.

@baldur So it’s often ending up being a *worse* version of the automation paradox: the direct outcomes being anywhere from somewhat to significantly worse and the second-order effect still being deskilling. We tolerate the cost of the automation paradox in aircraft autopilot and similar systems because it significantly improves average outcomes. It doesn’t seem to be doing that here.

@chriskrycho It's interesting that the human factors paper that originated the automation paradox (https://en.wikipedia.org/wiki/Ironies_of_Automation) as well as much of the research that followed didn't focus on the idea of higher performance than baseline. The paradox kicks in as soon as the output is good enough to switch the human participant out of "participation" mode and into "monitoring" which is also more exhausting as it requires heightened vigilance.

@chriskrycho @baldur

there have been meaningful changes

are you familiar with the concept of a Shepard tone? Each note in a Shepard tone has meaningful changes, but the overall motion is nil. that's how LLM "improvements" work.

@baldur

I see it every time my coworker sends me an email.

Occasionally there is a gem of useful information buried somewhere among the reams of slop, but it takes so much effort to read it all that I'd probably have been better off not reading it in the first place.

@baldur

And, well, https://mastodon.sdf.org/@argv_minus_one/117215411079697884 bots are shockingly negligent. Ask them a question about programming, and instead of giving you one straight answer, they give you a dozen wild guesses and expect *you* to figure out which of them—if any—is correct.

I used to be better off with pre-2024 search engines and Stack Overflow, but now both of those things are full of bot-generated spam!

These bots aren't helpful for programmers. These bots are an unmitigated disaster for programmers.

@baldur

And before anybody says “lol get a real model”, the guy I was talking to in that thread used a real model, and it epically failed too. https://mastodon.sdf.org/@argv_minus_one/117215673653412253

Bots aren't just expensive and proprietary and full of ethical issues. They also don't work as advertised. They're great at generating generic marketing copy and whatnot, but as soon as the bulls**t stops and the actual engineering starts, they fall apart.

@argv_minus_one
@baldur
In my support role, I get support requests "assisted by", and sometimes even fully "agentic" and they are an absolute nightmare to work on.

Issues that aren't real, applied "fixes" that break things, generated fake "sources" for the prior fix or comments.

The back and forth is like communicating while both parties are suffering a watershed stroke.

@Brett_E_Carlock

It's not just you, and it's not just software. I've heard of legal filings citing nonexistent case law. At this rate, we're going to be living in a society in which everything is made up, and the very concept of truth no longer exists.

@baldur

@argv_minus_one
@baldur
I gather that is quite the point and why it is pushed so hard.

@baldur "if you haven't [seen the output of] these tools over the past five minutes, you don't know what [fresh hell] they're capable of, because everything's different now!"

every time i see someone say "i've reviewed some ai-generated PRs, so i know what these tools are capable of," i get a little perplexed.

you do know that seeing bad code doesn't mean you've seen the best these tools can produce, right? some of us are actually building software with them, y'know?

@baldur I mean, they're not wrong. You ask LLMs to do the same task and you get wildly different results just 5 mins apart 😂

@baldur The slop boosters always leave out a crucial part of the phrase because we absolutely know what sloptools are capable of (slop), what they mean by "you don't know what they're capable of" is "you don't know [how smart they make me feel as I accomplish nothing while offloading the real work to other people who have to review my slop]!"

@baldur They are also so occupied with trying to sell courses and learning material to the others, who they think are simply misinformed about how great this new technology is.

And it's always business school graduates.

Believe me, I've tried this stuff and I spent more time refining prompts into the night than I ever would have spent checking that ticket off myself.

And mind you — it feels amazing and you get instant feedback but the results are overcooked for a lack of a better word.

@baldur Did you see the Reply Guy lecturing Eugen Rochko (!) for having opinions about vibe coding, despite not having used chatbots himself? Basically everyone made the same point: you don't have to have "experienced" using the tech to assess the quality of its outputs.

Indeed, the elevation of subjectivity in this context looks very much like a form of "AI" psychosis. Of course Eugen is more than qualified to tell if the code is "spaghetti" (as he put it). 🤷🏻‍♂️

@ApostateEnglishman No I missed that, but it doesn't surprise me that he got that response.

@ApostateEnglishman If it would be a psychosis, you could treat them with medicine. 😉
I would compare it to a cult.

@baldur

@baldur “Capable of” vs what they are actually producing.

@baldur I know nearly nothing about coding. But these people sound like the guy entering a 3-star restaurant's kitchen. The boss calls him for cutting meat. The guy: why should I use that special knife! If you haven't used that new US weapon that can slice up everything ...

@baldur whaaaaat?! You are doing code reviews?! Just use another AI to do the review!*

* I hope it is clear that this is sarcasm.

@baldur The every-five-minutes argument coupled with the “embrace now or you’ll be left behind” sentiment obliterate each other in an explosion of pure bullshit for me every time I see them in the same airspace.

Like, if the tools are constantly changing, and there’s close to infinite incentive for the providers to make them as easy as possible to onboard onto (and develop a dependency upon), what possible motivation is there for one to _not_ put off adopting them? Explain with math, if you prefer.

@baldur but, but, but maybe you'll feel different when *you're* the one causing the harm instead!!?