"ChatGPT said this" Is Lazy
terriblesoftware.org
terriblesoftware.org
ChatGPT is free and available to everyone, and so are a dozen other LLMs. If the person making the comment wanted to know what ChatGPT had to say, they could just ask it themselves. I guess people feel like they’re being helpful, but I just don’t get it.
Though with that said, I’m happy when they at least say it’s from an LLM. At least then I know I can ignore It. Worse is replying as if it’s their own answer, but really it’s just copy pasted from an LLM. Those are more insidious.
The only workaround is to just text as-is and call it out when it's wrong/bad, AI-generated or otherwise, as we've done before 2023.
I think typically, the reason people are disclosing their usage of LLMs is that they want offload responsibility. To me it's important to see them taking responsibility for their words. You wouldn't blame Google for bad search results, would you? You can only blame the entity that you can actually influence.
Because if they'd actually read the output, then cross-checked it and developed some confidence in the opinion, they wouldn't put what they perceive as the most important part up front ("I used ChatGPT") - they'd put the conclusion.
Or, if you preface a comment with "I am not an expert, but...", it is not often about seeking to avoid all blame, but to simply give the reader reasonable context.
Of course, you are right, it is also sometimes just lazy dodging.
I find a good workaround is to just say "some very quick research of my own leads me to ...", and then summarize what ChatGPT said. Especially if you are using e.g. an LLM with search enabled, this is borderline almost literally true, but makes it clear you aren't just stating something completely on your own.
Of course, you should still actually verify the outputs. If you do, there is not much wrong with not mentioning using the LLM, since you've don't the most important thing anyway (not be lazy in your response). If you don't verify, you had better say that.
My experience is that the vast majority of people do 0 research (AI assisted or not) before asking questions online. Questions that could have usually been answered in a few seconds if they had tried.
If someone preface a question by saying they've done their research but would like validation, then yes it's in incredibly poor taste.
When you put it that way I guess it kind of is.
> If someone preface a question by saying they've done their research but would like validation, then yes it's in incredibly poor taste.
100% agree with you there
The "let me Google that for you" was more trying to get people to look up trivial things on their own, rather than query some forum repeatedly.
they're more clueless than condescending
That's not like pasting in a screenshot or a copy/paste of an AI answer, it's being intentionally dismissive. You weren't actually doing the "work" for them, you were calling them lazy.
The way I usually see the AI paste being used is from people trying to refute something somebody said, but about a subject that they don't know anything about.
Which was just as irritating.
No. With Google you get many answers. With AI you get one. Also we know that AI is unreliable with some possibility, it’s highly probable that you can get a better source on Google than that. This is especially bad when the question is something niche. So, it’s definitely a worse version of lmgtfy.
udm=14 my beloved
> What can be asserted without evidence can also be dismissed without evidence.
Becomes
> That which can be asserted without thought can be dismissed without thought.
Since no current AI thinks but humans do I’m just going to dismiss anything an AI says out of hand because you are pushing the cost of parsing what it said onto me and off you and nah, ain’t accepting that.
Hell, Feynman said as much in 1985. https://www.youtube.com/watch?v=ipRvjS7q1DI
One of these courses of action is helpful, while the other one very much isn't. Copy-pasting 5 paragraphs of ChatGPT is equally not helpful. If however find one small blurb in that ChatGPT output that's germane to whatever point is being made, and copy and paste that, along with a source and/or your reasoning as to why that is correct, then you're probably back to being helpful again. This however is not what most instances of 'chatGPT said' are. Most of those are 'chatGPT said [insert word salad here]' or '[chatGPT said [thing obvious from even the most cursory knowledge about the subject at hand]'.
Whether the hatred is justified or not, the openness of said hostility just breeds users who'll hide their usage and deliver the same content, just repackaged enough in attempt to throw off the BS smell. And now you, the user, might be two paragraphs into something that may be truthful enough to fool you unless you're also an expert on the topic. I'd rather know up front what their source was and weigh the data with that in mind.
Otherwise we're now all going to have to assume everything we read is AI generated and suspect everyone else is going to look at our own content under such a microscope. Maybe that's a positive though.
I had a consultant I’m working with have an employee do that to me. I immediately insisted that every hour they’ve billed on that person’s name be refunded.
They removed the analyst from our project and credited the hours along with a bonus.
I start responding again when they can be bothered to formulate their own thoughts.
If a stupid LLM doesn't understand what something might be, chances are your user won't either.
This is not a problem I've run into fortunately, but I'm appalled that this exists.
If anyone gives me an opinion from an AI, they disrespect me and themselves to a point they are dead to me in an engineering capacity. Once someone outsources their brain they are unlikely to keep learning or evolving from that point, and are unlikely to have a future in this industry as they are so easily replaceable.
If this pisses you off, ask yourself why.
I would rather people go find the actual whitepaper or source in the footnotes and give me that, and/or give me their own opinion on it.
Not write "Wikipedia says..." and paste the entire article verbatim.
Why would it piss me off that you’re so closed minded about an incredible technology?
Like sure that is cool that is possible, but if I do not do the work myself I will not get stronger.
Our brains are the same way.
I also do not use a GPS because there are literally studies with MRI scans proving it makes an entire section of our brain go dark compared to London taxi drivers required by law to navigate with their brains.
I also navigate life without a smartphone at all, and it has given me what feels like focus super powers compared to those around me, when in reality probably most people had that level of focus before smartphones were a thing.
All said AI is super interesting when doing specialized work at scale no human has time for, like identifying cancer by training on massive datasets.
All tools have uses and abuses.
How many IQ points do you gain per year of subjecting yourself to this?
People are using LLMs to generate code without doing this.
If anyone gives me an opinion from a book, they disrespect me and themselves to a point they are dead to me in an engineering capacity. Once someone outsources their brain they are unlikely to keep learning or evolving from that point, and are unlikely to have a future in this industry as they are so easily replaceable.
If this pisses you off, ask yourself why.
(You can replace AI with any resource and it sounds just as silly :P)
It's so strange that pro-AI people don't see this obvious fact and keep trying to compare AI with things that are actually correct.
I find models vastly more useful than most technical books in my own work because I know how to feed in the right context and then ask them the right questions about it.
There isn't a book on earth that could answer the question "which remaining parts of my codebase still use the .permission_allowed() method and what edge-cases do they have that would prevent them from being upgraded to the new .allowed() mechanism"?
No one really cares how you found all those .permission_allowed() calls to replace - was it grep, or intense staring, or AI model. All that matters is you stand behind it, and act as an author. Original post said it very well:
> ChatGPT isn’t on the team. It won’t be in the post-mortem when things break. It won’t get paged at 2 AM. It doesn’t understand the specific constraints, tech debt, or your business context. It doesn’t have skin in the game. You do.
You're so close to realising why the book counter argument doesn't make any sense!
Those people exist and they’re wrong.
More frequently, however, I find I’m judging the model less than its user. If I get an email that smells of AI, I ignore it. That’s partly because I have the luxury to do so. It’s largely because engaging has commonly proven fruitless.
You see a similar effect on HN. Plenty of people use AI to think through problems. But the comments that quote it directly are almost always trash.
Because the output is almost always trash, and it takes re-doing the work and then claiming that it came from LLMs for it not to be.
These tools are being sold to me as similar to what a Jr engineer offers me, but that's not at all true because I would fire a Jr that came to me with such bullshit and needing such significant hand-holding so often as what I see coming out of an LLM.
—-
It’s interesting that we have to respect human “stupid” opinions but anything from AI is discarded immediately.
I’d advocate for respecting any opinion. And consider good or at least good willed opinion.
This does not apply to AI of course. In most cases, if a person did an AI PR/comment once, they will keep doing AI PRs/comments, so your explanation will be forgotten next time they clear context. Might as well not waste your time and dismiss it right away.
The fact that you're presenting this as a comically absurd comparison tells me that you know well that it's an absurd comparison.
It’s not the source that matters. It’s not the source that he’s complaining about. It’s the nature of the interaction with the source.
I’m not against watching video, but I won’t watch TikTok videos, because they are done in a way that is dangerously addictive. The nature of engagement with TikTok is the issue, not “I can’t learn from electrical devices.”
Each of us must beware of the side effects of using tools. Each kind of tool has its hazards.
Let’s try it with other stuff:
“Looking at solutions on stack overflow outsources your brain”
“Searching arxiv for literature on a subject outsources your brain”
“Reading a tutorial on something outsources your brain”
There’s nothing that makes ChatGPT et al appreciably different from the above, other than the tendency to hallucinate.
ChatGPT is a better search engine than search engines for me, since it gives links to cite what it’s talking about and I can check those, but it pays attention to precisely what I asked about and generally doesn’t include unrelated crap.
The only complaint I have is the hallucinations, but it just means I have to check its sources, which is exactly the case already for something as mundane as Wikipedia.
Ho hum. Maybe take some time to reevaluate your conclusions here.
But it's a very different problem from "outsourcing your brain". In fact, the tendency to hallucinate is the very reason you still need to use your brain when using an LLM.
As I said, it's not any different from a web search or a wikipedia read: It's one input, and you still have to think for yourself.
I have recently started to use codex on the command line. Before I put the prompt in, I get an idea in my head of what should happen.
Then I give it the instructions, sometimes clarifying my own thoughts while doing it. These are high level instructions, not "change this file". Then it bumps away for minutes at a time, after which I diff the results and consider if it matches up to what I would expect. At that point lower level instructions if appropriate.
Consider whether it was a better solution or not, then ask questions around the edges that I thought were wrong.
It turns my work from typing code in to pretty much code design and review. These are the hard tasks.
Unfortunately that logic does not apply to models.
If your interaction with the junior dev is not much different than interacting with an LLM, something is off.
Training a junior dev will make you a better dev. Teaching is learning. And a junior dev will ask questions that challenge your assumptions.
It's the opposite of "outsourcing."
So, working with CLAUDE doesn't count. Gotcha.
> If this pisses you off, ask yourself why.
It doesn't piss me off, but your comment is disingenuous at best.
At my previous company they called it 'sparring with <name of the software>'. You don't 'work' with Claude.
You use the software, you instruct it what to do. And it gives you an output that you can then (hopefully) utilize. It's not human.
You don't have to outsource your thinking to find value in AI tools you just have to find the right tasks for them. The same as you would with any developer jr to you.
I'm not going to use AI to engineer some new complex feature of my system but you can bet I'm going to use it to help with refactoring or test writing or a second opinion on possible problems with a module.
> unlikely to have a future in this industry as they are so easily replaceable.
The reality is that you will be unlikely to compete with people who use these tools effectively. Same as the productivity difference between a developer with a good LSP and one without or a good IDE or a good search engine.
When I was a kid I had a text editor and a book and it worked. But now that better tools are around I'm certainly going to make use of them.
We work with junior engineers because we are investing in them. We will get a return on that investment. We also work with other humans because they are accountable for their actions. AI does not learn and grow anything like the satisfying way that our fellow humans do, and it cannot be held responsible for its actions.
As the OP said, AI is not on the team.
You have ignored the OP’s point, which is not that AI is a useless tool, but that merely being an AI jockey has no future. Of course we must learn to use tools effectively. No one is arguing with that.
You fanboys drive me nuts.
Yes, when someone builds a straw man you ignore it. There is a huge canyon between never use AI in engineer(op proposal) and only use AI for all your engineering(op complaint).
If so, congratulations for being old or belonging to the 0.01%. Good luck finding a first job where that holds in 2025.
If you looked me or my work up, I think you would likely feel embarrassed by this statement. I have a number of world firsts under my belt that AI would have been unable to meaningfully help with.
It is also unlikely I would have every developed the skill to do any of that aside from doing everything the hard way.
Do you do all your coding in ed or are you already using technology to offload brain power and memory requirements in your coding?
Also I use VIM. Any FOSS tools with predictable deterministic behavior I can fully control are fine.
So your ok with using tools to offload thinking and memory as long as they are FOSS?
It took some iteration and hands on testing to get that right across multiple operating systems. Also to pass shellcheck, etc.
Even if an LLM -could- do that sort of thing as well as my team and I can, we would lose a lot of the arcane knowledge required to debug things, and spot sneaky bugs, and do code review, if we did not always do this stuff by hand.
It is kind of like how writing things down helps commit them to memory. Typing to a lesser extent does the same.
Regardless those scripts are like <1% of the repo and took a few hours to write by hand. The rest of the repo requires extensive knowledge of linux internals, compiler internals, full source bootstrapping, brand new features in Docker and the OCI specs, etc.
Absolutely 0 chance an LLM could have helped with bootstrapping a primitive c toolchain from 180 bytes of x86 machine code like this: https://codeberg.org/stagex/stagex/src/branch/main/packages/...
That took a lot of reasoning from humans to get right, in spite of the actual code being just a bunch of shell commands.
There are just no significant shortcuts for that stuff, and again if there were, taking them is likely to rob me of building enough cache in my brain to solve the edge cases.
Also yes, I only use FOSS tools with deterministic behavior I can modify, improve, and rely on to be there year after year, and thus any time spent mastering them is never wasted.
I decided to see if I could get an old Perl and C codebase running via WebAasembly in the browser having Claude brute-force figuring out how to compile the various components to WASM. Details here: https://simonwillison.net/2025/Oct/22/sloccount-in-webassemb...
Here are notes it wrote for me on the compilation process it figured out: https://github.com/simonw/tools/blob/473e89edfebc27781b43443...
I'm not saying it could have created your exact example (I doubt that it could) but you may be under-estimating how promising it's getting for problems of that shape.
The best form of (AI) plagiarism is to simply wrap the original tool in your own facade and pretend like you built anything of value.
Is this intended to be a bad joke?
Honestly, I've seen the occasional bad faith argument from people with a passionate dislike of AI tooling but this one is pretty extreme even by those standards.
I hope you don't ever use open source libraries in your own work.
<https://github.com/simonw/tools/blob/473e89edfebc27781b43443...>
At least run a pretty-printer on the code so that it can be reviewed by anything but a robot.
Part of developing a good software system is about exercising taste in vendored-in libraries and, especially, the structure around them.
P.S. I've gone to look at other chunks of Javascript and see that I was unlucky to grab this steaming pile first.
Originally I tried to get it working loading code directly but as far as I can tell there's no stable CDN build of that, so I had to vendor it instead.
Even then, I would never use it, because it would be some proprietary model I have to pay some corpo for either with my privacy, my money, or both... and they could take it away at any time. I do not believe in or use centralized corpotech. Centralized power is always abused eventually. Also is that regurgitated code under an incompatible license? Who knows.
Also, again, I would rob myself of the experience and neural pathway growth and rote memory that come from doing things myself. I need to lift my own weights to build physical strength just as I need to solve my own puzzles to build patience and memory for obscure details that make me better at auditing the code of others and spotting security bugs other humans and machines miss.
I know when I can get away with LTO, and when I cannot, without causing issues with determinism, and how to track down over linking and under linking. Experience like that you only get by experimenting and compiling shit hundreds of times, and that is why stagex is the first Linux distro to ever hit 100% determinism.
Circling back, no, I am not worried about being unemployable because I do not use LLMs.
And hey, if I am totally wrong and LLMs can create perfectly secure projects better than I can in the future, and spot security bugs better than I can, and I am unemployable, then I will go be an artist or something, because there are always people out there that appreciate hard work done by humans by hand, because that is how I am wired.
Have you been following the developments in open source / open weight models you can run on your own hardware?
They're getting pretty good now, especially the ones coming out of China. The GLM, Qwen and DeepSeek models out of China are all excellent. Mistral's open weight models (from France) are good too, as are the OpenAI gpt-oss models.
No privacy or money cost involved in running those.
I get your concern about learning more if you do everything yourself. All I can say there is that the rate and depth of technical topics I'm learning has been expanded by my LLM usage because I'm able to take on a much wider range of technical projects, all of which teach me new things.
You're not alone in this - there are many experienced developers who are choosing not to engage with this new family of technology. I've been thinking of it similar to veganism - there are plenty of rational reasons to embrace a vegan lifestyle and I respect people who do it but I've made different choices myself.
I will admit that I find deploying and hacking on the tech itself super interesting. Hell I founded a machine learning company and got a paper published with the AAAI for my cheap bulk training data acquisition techniques back in 2012 before most cared about this stuff.
I even think there are a ton of great and exciting use cases for this tech. Like identifying cancer in large photographic datasets, etc. I have a lot of hope about medical applications in particular.
All that said, I just don't think LLMs are remotely competitive or useful at the type of threat modeling, security engineering, and auditing work I do on average. They are the wrong tool for my job, which require a level of actual reasoning that LLMs are nowhere near capable of right now, or are likely to be any time soon. Maybe they could help with a script here and there which might save me a few hours a month, but for 95%+ of it, they would just waste my time regurgitating the same industry standard bad advice and approaches that I am trying to change while making me duller at writing code by hand when I need to.
As contrast though, I would not fire someone for using LLMs for learning or inspiration as long as they consistently prove they fully understand and can explain every line of every PR they submit, can pair program or usefully contribute to engineering discussions without LLMs, and maintain a competitive level of quality with the rest of the team. Not everyone has to make the same tool choices I do, as long as they can hold their own in a team with me and are not dumb enough to regurgitate AI slop they don't understand.
It is amusing you use vegans as an example. I am not a vegan, but I often describe myself as something of a digital vegan that is very very selective about what tools I use and what I expect from them such as why I also don't use a smartphone or GPS.
They believe that the entirety of human ingenuity should be theirs at no cost, and then they have the audacity to SELL their ill-gotten collation of that knowledge back to you? All the while persuading world governments that their technology is the new operating system of the 21st century.
Give me a dystopian break, honestly.
Maybe someone vibe codes their way through some webapp stuff and they review and clean it up, I will not stop them, but if someone turns in obvious AI slop they clearly do not understand they are dead to me.
I do security auditing and engineering though, which is in very high demand and something LLMs are incredibly bad at right now.
Soc. At the Egyptian city of Naucratis, there was a famous old god, whose name was Theuth; the bird which is called the Ibis is sacred to him, and he was the inventor of many arts, such as arithmetic and calculation and geometry and astronomy and draughts and dice, but his great discovery was the use of letters. Now in those days the god Thamus was the king of the whole country of Egypt; and he dwelt in that great city of Upper Egypt which the Hellenes call Egyptian Thebes, and the god himself is called by them Ammon. To him came Theuth and showed his inventions, desiring that the other Egyptians might be allowed to have the benefit of them; he enumerated them, and Thamus enquired about their several uses, and praised some of them and censured others, as he approved or disapproved of them. It would take a long time to repeat all that Thamus said to Theuth in praise or blame of the various arts. But when they came to letters, This, said Theuth, will make the Egyptians wiser and give them better memories; it is a specific both for the memory and for the wit. Thamus replied: O most ingenious Theuth, the parent or inventor of an art is not always the best judge of the utility or inutility of his own inventions to the users of them. And in this instance, you who are the father of letters, from a paternal love of your own children have been led to attribute to them a quality which they cannot have; for this discovery of yours will create forgetfulness in the learners' souls, because they will not use their memories; they will trust to the external written characters and not remember of themselves. The specific which you have discovered is an aid not to memory, but to reminiscence, and you give your disciples not truth, but only the semblance of truth; they will be hearers of many things and will have learned nothing; they will appear to be omniscient and will generally know nothing; they will be tiresome company, having the show of wisdom without the reality.
Phaedr. Yes, Socrates, you can easily invent tales of Egypt, or of any other country.But e.g. ChatGPT with search enabled is often an invaluable research tool, and dramatically speeds up finding relevant sources. It basically automates the spidering of references and links, and also handles the basic checks for semantic relevance quite well, and this task requires little real intelligence or thought. Only once you hit a highly specific and niche technical domain will it start to fail you here (since it will match on common-language semantics that do not often align with technical usage).
For a lot of topics, I now have the reverse feeling that a person who has NOT used an LLM to facilitate search—on basic or even intermediate questions—is increasingly more of a concern.
Until then, LLMs are of no interest to me. I am pretty good at using grep on local data caches.
Perhaps a model fine-tuned or specifically built for the niche domains might help non-experts and people learning to better do semantic / common-language searches, and this might accelerate learning. But, it might not. Not all domains have their core knowledge encoded in documents and/or language, and summarization is sometimes only harmful (especially for highly technical writing).
It also seems to me almost definitional though that LLMs aren't going to beat true expertise, because LLMs aren't true intelligence or GAI, they really do just estimate token path densities / curve-fit. And real understanding is clearly more than that.
> Once someone outsources their brain they are unlikely to keep learning or evolving from that point
It doesn't piss me off, it makes me feel sorry for you. Sorry that you're incapable of imagining those with curiosity who use AI to do exactly what you're claiming they don't - learning. The uncurious were uncurious before AI and remain so. They never asked why, they still don't. Similarly, the curious still ask why as much as ever. Rather than Google though, they now have a more reliable source to ask things to, that explains things much better and more quickly.
I hear you grunt! Reliable? Hah! It's a stochastic parrot hallucinating half the time!
Note that I said more reliable than Google, which was the status quo. Google is unreliable. Yes, even when consulting three separate sources. Not that one has the time for that, not in reality.
You've got it the wrong way around. LLMs do the exact opposite - they increase the gap between the curious and the nots. It accelerates the learning rate gap between them. The nots.. they're in for a tough time. LLMs are so convenient, they'll cruise through life copypasting their answers, until they are asked to demonstrate competence in a setting where none are available and everything falls apart.
If you still find this hard to imagine, here's how it goes. In your mind LLM usage by definition goes like this - and for the uncurious, this is indeed how it would go.
User: Question. LLM: Answer. End of conversation.
By the curious, it's used like this.
User: Question. LLM: Answer. User: Why A? Why B? LLM: Answer. User: But C. What's the tradeoff? LLM: Answer. User: Couldn't also X? LLM: Answer. User: I'm not familiar with Y, explain.
Telling when you're asking the wrong questions is also a skill that the curious such as yourself will develop in no time through experience. Are you saying this didn't give a shred of intuition how to do so in the future?
In practice it’s more like:
User: Question. LLM: Answer. User: Why A? LLM: Actually, you’re right, it’s not A, it’s B. User: Why B? LLM: Actually, you’re right, it’s not B, it’s C. User: Why C? LLM: Actually, you’re right, it’s not C, it’s D.
I’ve had this happen pretty much every time I asked an LLM a non-trivial question.
Of course never use chat models, only straight API, as the system prompts for consumer versions make them induce sycophancy as that is what the uncurious like. For this conversation, don't use things like Claude Code or other harnesses either. Just have a conversation through API calls.
You also changed the questions in the scenario in a way that makes them more likely. Keep them open-ended, the curious' most used phrase should probably be "Why or why not? If so, explain why. If not, explain why not".
That said, for me personally, almost all of the work I do is on things that have never been done before in security engineering and supply chain security and the entire body of relevant public research an LLM could have trained on is like 10 people, most of whom I am in frequent touch with and very familiar with their work.
LLMs in general are very very bad at threat modeling or security engineering because there is so little training data on the -right- way to do things. They can often produce code that -works- but most would be unable to spot when it is wildly insecure.
There are many many cases where the overwhelming majority of the way things are done in open source training data are completely wrong, and are what an LLM will respond with on average.
Honestly the only way I could maybe see using an LLM in teaching the type of security engineering and auditing work I do is having LLMs generate code examples, and train humans to spot the security flaws the LLM confidently overlooked because it cannot reason or threat model.
This is a framing issue that's keeping you from doing clearer thinking, faster. LLMs, especially in agentic configurations, are a cognitive tool. Just because many people use a tool unskillfully doesn't mean that tool is useless.
>If anyone gives me an opinion from an AI, they disrespect me and themselves to a point they are dead to me in an engineering capacity.
Hmmm...
> Once someone outsources their brain they are unlikely to keep learning or evolving
Agreed. But there are many ways in which people can "outsource their brains". LLMs are one way. Dogmas are another.
>If this pisses you off, ask yourself why.
It would definitely piss me off to see an engineering leader fail to grasp the obvious due to some poorly reasoned beliefs.
A simple solution would be to mandate that while posting coversations with AI in PR comments is fine, all actions and suggested changes should be human generated.
They human generated actions can’t be a lazy: “Please look at AI suggestion and incorporate as appropriate. ”, or “what do you think about this AI suggestion”.
Acceptable comments could be: - I agree with the AI for xyz reasons, please fix. - I thought about AIs suggestions, and here’s the pros and cons. Based on that I feel we should make xyz changes for abc reasons.
If these best practices are documented, and the reviewer does not follow them, the PR author can simply link to the best practices and kindly ask the reviewer to re-review.
Like, it's fine for you to use AI, just like one would use Google. But you wouldn't paste "here are 10 results I got from Google". So don't paste whatever AI said without doing the work, yourself, of reviewing and making sense of it. Don't push that work onto others.
Stories full of nonsensical, clearly LLM-generated acceptance requirements containing implementation details which are completely unrelated to how the feature actually needs to work in our product. Fine, I didn't need them anyway.
PRs with those useless, uniformly-formatted LLM-generated descriptions which don't do what a PR description should do, with a half-arsed LLM attempt at summary of the code changes and links to the files in the PR description. It would have been nice if you had told me what your PR is for and what your intent as the author is, and maybe to call out things which were relevant to the implementation I might have "why?" questions about. But fine, I guess, being able to read, understand and evaluate the code is part of my job as a reviewer.
---- < the line
PRs littered with obvious LLM comments you didn't care enough to take out, where something minor and harmless, but _completely pointless_ has been added (as in if you'd read and understood what this code does, you'd have removed it), with an LLM comment left in above it AND at the end of the line, where it feels like I'm the first person to have tried to read and understand the code, and I feel like asking open-ended questions like "Why was this line added?" to get you to actually read and think about what's supposed to be your code, rather than a review comment explaining why it's not needed acting as a direct conduit from me to your LLM's "You're absolutely right!" response.
> uniformly-formatted LLM-generated descriptions which don't do what a PR description should do, with a half-arsed LLM attempt at summary of the code changes and links to the files in the PR description. It would have been nice if you had told me what your PR is for and what your intent as the author is, and maybe to call out things which were relevant to the implementation I might have "why?" questions about.
If I want to see what the code changes do, I will read the code. I want your PR description to tell me things like:
- What the tradeoffs, if any, to this implementation are
- If there were potential other approaches you decided not to follow for XYZ reason so that I don't make a comment asking about it
- If there is more work to be done, and if so what it is
- Any impacts this change might have on other systems
- etc.
Sure, if you want to add a handful of sentences summarizing the change at a high level just to get me in context, that's fine, but again if I want to see what changed, I will go look at what changed.
Think of it as a dynamic opinion poll -- the probabilistic take on this thing is such and such.
As a bonus you can prime the respondent's persona.
// After posting, I see another comment at bottom opening with "Counterpoint:"... Different point though.
I know ChatGPT exists. I could have fucking copied-and-pasted my question myself. I'm not asking you to be the interface between me and it. I'm asking you, what you think, what your thoughts and opinions are.
Wait no, if your boss is making you use this tech, then its his fault.
Wait no, if the companies are selling this as the holy grail of firing everyone to save money, its the fault of whomever is buying this trash without testing it
Wait no, if this is trained on the entirety of human information, then we are all wrong for not deleting all of our bad answers written on the internet
No one I know who says this kind of thing would read this article. People love being lazy.
One example: Code reviews are inherently asymmetrical. You may have spent days building up context, experimenting, and refactoring to make a PR. Then the reviewer is expected to have meaningful insight in (generously) an hour? AI code reviews help bring balance; it may notice stuff a human wouldn't, and it's ok for the human reviewer to say "hey, chatgpt says this is an issue but I'm not sure - what do you think?"
We run all our PRs through automated (claude) reviews automatically, and it helps a LOT.
Another example: Lots of times we have several people debugging an issue and nobody has full context. Folks are looking at code, folks are running LLM prompts, folks are searching slack, etc. Sometimes the LLMs come up with good ideas but nobody is sure, because none of us have all the context we need. "Chatgpt says..." is a way of bringing it to everyone's attention.
I think this can be generalized to forum posts. "Chatgpt says" is similar to "Wikipedia says". It's not the end of the conversation, but it helps get everyone on the same page, especially when nobody is an expert.
Certainly citing GPT is better than just assuming it's right and not citing it along with an assertion.
I wrote before about just sending me the prompt[0], but if your prompt is literally my code then I don't need you at all.
----
In many of the Hacker News comments, a core complaint was not just that AI is sometimes used lazily, but that LLM outputs are fundamentally unreliable—that they generate confidently stated nonsense (hallucinations, bullshit in the Frankfurtian philosophical sense: speech unconcerned with truth).
Here’s a more explicitly framed summary of that sentiment:
⸻
Central Critique: AI as a Bullshit Generator
Many commenters argue that: • LLMs don’t “know” things—they generate plausible language based on patterns, not truth. • Therefore, any use of them without rigorous verification is inherently flawed. • Even when they produce correct answers, users can’t trust them without external confirmation, which defeats many of the supposed productivity gains. • Some assert that AI output should be treated not as knowledge but as an unreliable guess-machine.
Examples of the underlying sentiment: • “LLMs produce bullshit that looks authoritative, and people post it without doing the work to separate truth from hallucination.” • “It costs almost nothing to generate plausible nonsense now, and that cheapness is actively polluting technical discourse.” • “‘I asked ChatGPT’ is not a disclaimer; it’s an admission that you didn’t verify anything.”
⸻
Philosophical framing (which commenters alluded to)
A few participants referenced Harry Frankfurt’s definition of bullshit: • The bullshitter’s goal isn’t to lie (which requires knowledge of the truth), but simply to produce something that sounds right. • Many commenters argue LLMs embody this: they’re indifferent to truth, tailored to maximize coherence, authority, and user satisfaction.
This wasn’t a side issue—it was a core rejection of uncritical AI use.
⸻
So to clarify: the strong anti-AI sentiment isn’t just about laziness.
It’s about: • Epistemic corruption: degrading the reliability of discourse. • False confidence: turning uncertainty into authoritative prose. • Pollution of knowledge spaces: burying truth under fluent fabrication.
People did this with code as well. DDG used to show you the first Stackoverflow post that was close to what you searched. However sometimes this was obviously wrong, people have just copied and pasted that wholesale.
the other two are still incomparably better in practice though.
Fortunately
1. The person was transparent about it, even posting a link to the chat session
2. They had to follow on prompt to really engage the sycophancy
3. The forum admins stepped in to speak to this individual even before I was aware of it
I actually did what you suggested, fed everything back into another LLM, but did so with various prompts to test things out. The responses where... interesting, the positive prompt did return something quite good. A (paraphrased) quote from it
"LLMs are a powerful rhetorical tool. Bringing one to a online discussion is like bringing a gun to a knife fight."
That being said, how you prompt will get you wildly different responses from the same (other) inputs. I was able to get it to sycophant my (not actually) hurt feelings.