AI;DR (AI; Didn't Read)
rickmanelius.com
rickmanelius.com
Features are coming out and metrics are improving, but we’re basically in a post readability code base, with the occasional performative comment about a variable name.
I don’t really know how to address this situation or if it needs addressed. I certainly don’t read the long-winded AI comments or the AI documentation, but perhaps it’s useful for the AI on its next pass.
Write REVIEW.md.
I have CC check itself pretty well.
I also put into agent/claude/review instructions to write using simple English skill and humanizer skill. Then not to write redundant comments.
It’s not perfect but definitely catches lots of slop.
How do you enforce that?
In this fucked up job market, it is easier said than done.
I feel like it's all just perception and how companies can sell their stories to investors or potential acquirers, and everything else can be punted and dealt with later when we get acquired or when share prices are a zillion dollars etc.
It's a race to the bottom, for sure.
If you really want to save your career, learning how to have real conversations is a vital skill.
Asking because when I was at capital one, certain comments or questions about topics like this, would actually get you noticed negatively by your manager and being disruptive to what leadership wants hurts your career.
In fact being the one to ask and point out questions like yours ultimately got me PIP'ed and removed from the company. So, like, yeah, being vulnerable and the first one to tell the truth is risky. That's why the bystander effect can happen and in politics or risky situations silence is a common response, and why we don't have more courageous people doing the right thing - because the risk is higher on the person speaking up and the rewards aren't disproportionately in their favor (but are evenly distributed though, so its not favorable for you as I learned from a game theory perspective to voice up / defend certain stances).
I started my first post-Capital One job a few months ago, and I'm still not used to the fact that this company expects everybody working here to still be working here in two years. Most employers are not going through everybody with a fine tooth comb every 6 months hunting for any hint of an excuse to PIP them.
It was really a bizarre place to work, because work-life balance was amazing, the benefits and vacation were great, but twice a year there was this insane hunger games.
They really chilled out the performance management process from 2020 or so through 2023, so people who worked there then didn't always notice.
I don't recommend going to Plano, TX and joining the Auto Navigator product division.
Also, if you go back you'll note that I didn't suggest that people "tell the truth". I suggested people ask a neutral question about an obvious phenomenon. If it's unsafe to give an honest answer, most people won't, so it'll be a short discussion. But if it's a topic with some room, it can be the start of a good conversation and possibly a useful change.
I can see how that technique works great at a non-toxic company. Thank you.
There's a reason 3rd places exists. There's a social construct that off-work discussion (even when on-work) stays off work.
If your coworker is professional and cares about keeping channels open, they too will respect the unspoken rules that govern 3rd place neutral locations.
Blame it on the alcohol later on if it actually pisses someone off.
Not anymore these days.
Besides, GP was not talking about corporate, but about the inability to even suggest an improved usage of LLMs among their developer peers, as if every line of code generated by AI is now gospel and questioning it is treated as heresy. This level of defeatism is beyond cynical and approaching childish.
"My peers are all bad", "Management has no idea what's going on", "My boss is out of touch", "My staff keep screwing up".
But if you're consistently seeing the problems while everyone else becomes defensive, you (that think that way) need to learn how to communicate effectively in a way that doesn't result in being 'the baddie' and solve the interpersonal dynamic along with the technical problems. "Crucial Conversations" is a good place to start.
All the execs know that they can refresh the screen faster and see the features for marketing and sales.
When has engineering ever had a voice anyway? The bean counters have been looking to cut us out since the inception of our industry.
That's the behavior being discussed. In my experience, overly large PRs are a detriment to productivity, so if anything my suggestion (asking one's coworkers to stop that) might lead to an improvement in velocity. If I have to read an essay before and throughout every PR, I'm not going to be able to review that very quickly.
1) Understands the problem ("what are comments?")
2) Accepts that it is a problem ("how can more comments be a bad thing?")
3) Cares enough to solve the problem ("is this issue really a priority to solve right now? just accept the PR and we'll go back and fix if needed some other time cough never cough")
4) Believes you (this can take many forms, but the most common is, subconsciously, "this other engineer says it's not a problem, so I'll just assume it's not since that's easier")
All of these logistical, political and social factors are "the ocean"
To be fair, I would not be surprised.
And, really, it works. You can copy and paste between tickets and Claude, and then do manual testing. Then you tell Claude to self-review for clarity and minimalism, and stop worrying. Sure, today it's not as good as a human, but for almost all the code out there, it gets the job done. There's no skill needed any more, and if the boss doesn't care about quality, I don't see why I should.
There may not be a career left in software development with these monstrosities, but if there is, it's in using decades of experience to get the abominations to produce something other than unreadable dreck.
If the code itself truly doesn't matter any more, programmers will no longer be paid.
But the managers won't do it, because you still need manual testing to make sure that the AI models got things right. Managers will want to hire someone to do the manual testing.
If we remove that requirement (as most around me seem to be doing), then yeah, not much skill required, but the system is going to steadily trend towards being a buggy, unmaintainable pile of sloppypasta.
Maybe the models get good enough they can handle maintaining that, but anecdotal reports so far seem to indicate we aren't there yet.
I guess we'll see if AI 2027 pans out.
I don’t have time to go through and flag everything or even read the thousands of lines of code changes that are happening.
Corporate Capitalism's main innovation was virtualizing feudalism. Now anyone with a bit of cash flow can be a feudal lord with peasants to do his will. No need to maintain land or fight over it with other feudal lords. There's still fighting, naturally, but now the game isn't zero sum.
The serfs don't get any real say in this model. At best you can bend the lord's ear if you're sufficiently trusted an advisor. But part of the reason you're trusted is to be trusted not to treat the lord like he's too much of an idiot.
A huge problem is that capitalism rewards cash flows and accumulated capital so powerfully that lots of other things just don't matter, potentially for a very long time. Look at GitHub. This leads to immensely extended fuck around phases and "find out" looks like IBM (fossilization and bizarre holdings corporation / dead company parking lot) just as often as it looks like Enron.
I am a programmer but also in management, and if I knew my team thought like that I would be horrified. We absolutely care about what the team members want , like, hate etc. I for once would love someone to bring up stuff like this to me. As long as they are as open minded about things as they would like me to be, there is no issue. But it’s a serious issue to pretend everything is fine while thinking like you say.
Do the people above you in management care as much? It's the top levels of management that decide hiring, firing and budgets.
Reframe your context, pretend you're in a communist society, and you could be making the exact same point about speaking up against the grain being bad for you. It feels as though people are brainwashed by anti-capitalism sentiment at the moment, unable to see outside of that context.
Mega corporations pervert free markets and democracy, and we have a huge number of mega corporations At this point.
I'm pointing out that, due to how strongly capitalism rewards cash flow and previously accumulated capital, that feedback/correcting mechanism is often _extremely slow_.
That feels relevant to me.
It doesn't feel sustainable now. It won't feel sustainable in the future when companies are trying to operate on an entire fossil layer of LLM-generated legacy code.
Do I expect managers to get smarter about LLMs and their limits or do I "retire" and find an entire new industry to work in?
It's really depressing and there has been a lot of grieving since April that it feels like my entire 20+ year career path disappeared overnight, replaced by something unsustainable and ultimately terrifying.
They totally gotta be doing that at OpenAI. Meeting invitees: You, co-workers, GPT 5.6.
Yes, that's the entire point. And it is extremely useful. Why wouldn't I want this?
Give feedback that about the docs and block merging till the issue is resolved.
Actually, rejecting them is precisely what will make them easy.
"Sorry, the comments are so bad I'm stopping here. Please fix them and then I'll resume the code review."
You're giving everybody (including yourself) more work by:
1. Reviewing the code (even if you skip the documentation).
2. Letting too many abstruse comments in which everyone in the team will have to read.
3. Allowing the behavior to continue.
Become the bottleneck so the team can talk about it. If they decide this shouldn't be a blocker, just declare you won't review the comments going forward.
Not to mention that the response to each review will come with its own set of new comments and new documentation. The ability of people to write things has exploded tenfold. You can’t out review the slop.
Every PR can be rewritten and re-architected on a whim.
I'm currently not working in tech, but I will again sometime within the next year or so, and I've been reflecting on my career in light of the recent AI madness. I think the biggest mistakes I've made over the last decade+ in tech as a software engineer have been caring. I've worked at small, medium, and large companies. Famous big names you've definitely heard of, less famous ones you probably haven't. In every case, in retrospect, as an individual contributor (non founder, non board member, etc) being personally invested in some outcome--quality, efficiency, cost, ux, customer value--is a mistake.
If you aren't actually in a position to change something, caring about it is futile. [edit] And I don't mean in the sense of "feeling empowered" or some such, but actually having the clout to steer the ship and make it happen. Very few people in a tech organization actually have that power, by design.
The industry wants to do away with code review, meaningful testing, computer security, and reliable services. Why swim upstream? You can't actually do anything about it so why try?
When I go back to work I'm going to do my 9-5 40hr/wk, smile, nod, punch the clock, and make the bosses smile or whatever. But I'll be damned if I'll ever give a shit again.
Because habits beget habits, and in form imprint into your identity. I'm not the biggest fan of GPA fixation, but it does have the side effect of imprinting work habits (I won't say "ethics", because the extremes I saw peers go to to get A's is anything but). It doesn't necessarily imply mastery of your class, but your ability to adjust to someone's standards and meet them. Which is something that will follow you into the workforce, and roughly correlate with people who will meet the standards of their company.
If I don't give a damn about code quality in a place I spend a third of my life, why do I expect to give a damn when I want to work on my own project in the future? Even if I try those habits will wear into something I can call my own. This might be fine if your overall goals have nothing to do with your work; that you are fine just going through the motions and getting a paycheck to empower your non-tech hobbies or focus on supporting your family. But that's not the path I've chosen.
But in this case, we're talking about reviewing a deluge of AI slop. Other employees have been incentivized to generate all that slop. It's not your job as an IC reviewing said slop to push back and try to make it better. It's your job to approve the PR and get out of the way. If you do something else it'll come back at you in a bad way.
And probably if you're not generating a comparable volume of code, your output will be labeled deficient.
These things happened at companies that value their reputations as "good" places to work. Not Amazon. Not Google. I mistakenly thought fixing these problems would not only be possible, but also that it would be appreciated.
Maybe I'm tarring the industry with too broad a brush, but in my experience this is how it is: either you go along with the malfeasance or you're targeted. Being targeted fucking sucks. I'll never paint a target on my back again.
Then remove the whole process and let it go through automatically. I've worked in high trust teams where we would do that - there was no gate.
I left. I will always recommend finding a new job when you have dishonest management. (Dishonest by not admitting they are lowering the quality.)
Fortunately, none of my jobs since that job had that problem. They either allow good quality reviews, or they explicitly accept the technical debt that comes out of it and don't blame me for poor codebases.
(But then again, I make sure to probe these things during the interview process...)
I absolutely plan to do so throughly when I begin interviewing later this year. But I've been fooled before, so it'll take extraordinary evidence to get my hopes up, and I don't think I'll ever truly believe it again. There's always the possibility it isn't real or I'm misunderstanding something.
If you ever trust an employer you put yourself at a profound disadvantage. Employers are not so disadvantaged by placing (judiciously earned) trust in employees. That's because employers are companies, and companies are not people.
As I said here and elsewhere: If management is concerned code review is a bottleneck, let management be aware that the process has to change. Either the human is not as thorough, or he uses some combination of his brain + LLM to conduct the review.
It's a basic fact: They have a machine that produces a lot of code, but don't have a machine to review that code. The bottleneck is obvious. I'd love to build planes for $1 too, but physics applies.
Do you not have team meetings with the manager? Examine some of the PRs your coworkers reviewed, find the flaws, bring it up to the whole team, and get buy-in from the manager that he's OK with it. If he is, stop worrying and let the PRs through.
Also, start writing those CYA emails.
This is also how I do code review of AI work on my projects. If the work is offensive to the point I can't complete the review, I simply reject the code and tell the AI why. Then it goes off and fixes it. This repeats until the issues are either gone or are small enough that I can just fix them myself and move on.
There is no need to be upset. Just iterate until it's right. If it's cheap to write, it's equally cheap to rewrite.
I've denied for poor branch names and commits from AI. I've denied for too verbose of comments from AI. I've denied for parts of the code base being touched that are not relevant to the case they are working on (login isn't broken, your case is to add a check box in the settings pane, remove the changes made to login).
Pre-AI I wasn't fine with PRs with multiple features and touching irrelevant areas of the code base. Why would I be fine with it because my co-workers got new toys? You want AI to refactor a part of the codebase? Fine. Separate branch, new PR, and in the description, present an argument for it. Don't shoe horn it into something else. Also, I'm not obligated to hit approved on shit.
I may also be a bit privileged because I can be a pain in the ass to whole team. I may be slower, but I've got the numbers, my code is creating way less bugs then my "fast more efficient" co-workers.
Turns out not every team has the idea of "progress" being yet another iteration of "more code submitted = better code". Some people actually desire or need to make quality products.
It might be touching one line of actual code in a file, and take advantage of it to remove 20+ lines of actual useful comments.
Everybody is talking about the opposite, so I'm wondering if this is rare.
I ask because when I started informing it I made personal edits, it stopped doing this kind of thing so often and let me work _with_ it more.
My saved prompt now says never to assume a file has not been edited since the last time it was read between prompts.
Sometimes I have luck interrogating Claude on why it did something. It'll either point to a skill or agent file with the culprit, or it'll respond with some vapid nonsense and apologize.
It's been working well for me for months now. If only I could convince the team to do it too
Still I always find it funny when I give it code to review that it itself wrote and it immediately says how good the comments are, it's like that obama giving medal to obama meme.
I asked Sol to reduce the length of some documentation we had by making it more concise. It came back after 20 minutes of work, did a line count and was aghast that the line count had somehow increased...
Even then it is conservative. For the love of God, compact the comments.
Comments become a huge maintenance burden, especially in the age of AI. They just grow and grow, and then mislead the AI later on.
That is a dramatic shortcoming, but I fix it with permalinks to other files. https://replicated.live/blog/link
I also love how it'll build local plans with phases, tasks, or decisions, then reference those numbers in those same useless comments.
Now people think I'm just parroting what Claude said. It sucks. I want my catch phrases back, I guess this is how em dash users felt
When someone uses the presence of em-dash as the only signal for AI-generated then they are simply lazy.
They're only a LinkedIn-ism when used to create an unduly dramatic juxtaposition for an otherwise mundane idea. But now they set off people's AI radar when used for any reason at all.
You ask it to do something, then tell it to do something in a different way, then it assumes it needs to do the refactor in a backward compatible way, or creates migrations for it etc.
1) use the edit in-place functionality; or
2) ask it to rewrite to remove something
It'll write the prompt as if the agent (codex) knew about the conversation and add "don't do X" etc. At first that bothered but I realized it doesn't really change the output so I stopped caring.
Still, no experience is unique I guess.
It contains (Among other things) stuff like "this is a single user personal project, I'm the only user, this will never be open to the public internet" etc.
It kinda-sorta tones down the proclivity to worry about backwards compatibility and slight edge cases where if someone has edited some template and the new code doesn't support it.
I wish they trained the bots to be a notch more relaxed and less hysterical. Less is more.
But maybe that's just a consequence of the RL training being essentially AI torture to make them do what we want.
When I'm writing technical documentation, it keeps the explanations in. Same when writing non-technical documentation. When I was having it attempt to generate a Pathfinder 1e class for a Sword Dancer, it was leaving in notes about why it removes things I told it to remove/rework.
And it isn't just Claude. I've seen the same with GPT models, with Grok, with Deepseek. Each AI isn't quite the same with how it approaches this, but in every case they seem to have a strong bias to retaining information, even bad information that we want gone, so it is like they have a, dare I say, subconscious bias to retain the information. Putting a note in a comment or explaining why to not do something or something was undone is a good way to retain information while still achieving the goal (well, if you ignore the part about the human intention for the information to be gone).
This then weakens the AI in the future, as I find AI struggles with the more incorrect information. Sure, a comment saying "not X because Y" is less 'context damage' than a comment saying "X" (assuming X is wrong), but it is still a slight shift to X being present in context in some way. One off, AI's seem to perfectly handle this without issue. But after hundreds or thousands of cases build up? The attention mechanism seems unable to keep up and incorrect information flows it. This effectively creates a sort of vibe coding maximum size unless there is a human janitor cleaning up the bad information on the context stays nice and clean.
But this is all simply a feeling I get as I use AI to do different things and isn't at all backed up by any formal study.
1) Document what's there, not the diff. Documentation of how code was removed or changed to fix a bug or add a feature is not useful and difficult to maintain; documentation should explain how code works now.
2) Documentation should live close to the source as possible. Prefer line based comments and standardized function documentation. Top-level sweeping architectural essays are not maintainable for every change.
The last will depend on your codebase. It CAN be very useful to have a human-readable spec documented for the entire program and have it updated when anything changes. But the key is again, you're CHANGING it every time. If you add a whole new disconnected documentation file it should set off alarm bells; nothing in one system is truly disconnected.
We recently added a similar thing to our style guide, It’s astonishing to me that we have to spell this out, that something as obvious as this needs to be explained to LLM’s at all. They’re supposed to be exceeding human intelligence, at least at things like programming, but can’t understand basic things like what code comments are.
This perception is a good part of why this market is irrational. LLM's aren't "intelligent". They do not reason, they are a very fancy kitbash of whatever it trains on.
Ad yeah, I'm not surprised that a lot of documentation on every bit of readable code online is awful. "Document the diff" sounds like an anti-pattern learned from people with an incentive to get as many PR's submmitted as possible, not make the most friendly documentation for people maintaining a project.
Hehe. Yeah, that tendency of LLMs to document "the story" of the code instead of its current purpose (or non-obvious implementation details) is a pet peeve of mine too. I've added a slew of guidelines to try to sway Claude to not do this, but it still does it often.
At the same time, it feels like something to be expected to have this "failure mode". The model has its context to work on, and what is on its context if not the conversation you've been having (and its internal monologue) and the files it has read? It makes sense that it references the story on its text generations, because that behavior is usually a good thing for an LLM to do. Otherwise, what would it generate? If it generated things that had nothing to do with the conversation in its context, in many cases those things would be seen as "hallucinations", and they'd tend to be RLHF'ed out. So the models that we end up having are the ones that have been reinforced to be most "contextually relevant" and less "hallucinatory".
I might be completely wrong on that of course. It's just my intuitive reasoning of why this seems to be such a prevalent behavior.
It's doubly insidious because it trips up the human brain too: When I'm reading a PR saying "fix lock ordering to avoid deadlocks on user deletion", and there's a comment somewhere in the diff saying "// use the fixed lock ordering here", my brain tends to completely forget the fact that the comment doesn't make any sense in its surrounding context. Because it makes perfect sense in the context of being the human reviewing the diff. But it's a slight bit of mental effort to remind yourself "what is this comment going to look like to someone reading the code after this is merged?"
I wouldn't be surprised whatsoever if the RLHF supervisors forget to apply that extra bit of mental effort and say "yup, this comment looks great", forgetting to check the surrounding code to see if it makes sense on its own.
When vibe coding the content of user prompts is ground truth and the only way any human thought affect the code base. So if the vibe coder says "do X not Y", recording int comments that "we shouldn't do Y" is important. It ensures that the agent doesn't accidentally decide to do Y tomorrow, which would frustrate the vibe coder who'd feel that the agent doesn't "remember" what it was told yesterday.
So for people who look at the code the comments are obvious and completely superfluous, but for the vibe coder it's a way to ensure that their tiny (relative to the size of the code base) input is not forgotten.
A recent example I saw was an agent leaving a comment "// return an error here instead of panicking, as a panic will abort the process". Because likely the original human in the loop caught the AI putting a panic in there and told them not to, and then the comment to not do panics was placed in there. But to a future agent, it'll see that and think "ok, this comment must be here because we usually do use panics instead of returning errors, this place must be an exception", and now its context window is primed to think of using panics first.
It's pretty well-documented at this point that spending a lot of tokens explaining what not to do can actually increase the likelihood of an LLM doing that thing, especially when its context window is getting full.
It's like if you go to the grocery store and see a sign saying "Vegan tomatoes". It sounds fine until you think "wait, aren't all tomatoes vegan?" and now you start doubting yourself and start imagining what a non-vegan tomato would be.
Another semi-related issue is the tendency for LLM's to make up their own dumb little short-hand words for things that it has been talking about over and over in the context window. "The lock that prevents a user from being deleted while another thread is updating it" becomes a "user-fence", and now there's comments saying "// this function returns a user-fence", and I have absolutely no idea what that's supposed to even mean.
I wonder if I can get a MacArthur grant for this epiphany...
I wouldn’t say “post reading code” but it’s definitely trending in that direction.
I’d rather the agents put jumbo verbose descriptions in the PR description than in code comments TBH.
My worry is that after several passes this compounds and starts introducing errors or biases, a bit like in the "telephone game" children play.
I don’t think we need to write code anymore unless we want to - but we DO need to be humans, and respect the time of other people.
Then another coworker just pasted the Claude response with no signature and no comment and no context as if they had written it themself. That made me livid. I used to like that guy. No respect anymore.
I think enforcing black box testing is the best way to get useful tests out of both humans and robots. They must not know the internals, or it will lead them to do bad things.
In this day and age, https://hypothesis.works/articles/claude-code-plugin/ might be useful.
I do wonder if the idea of TDD has influenced AIs to be too prone to testing even when a human would never consider it. I've seen some really silly tests, especially when it starts writing tests for things that are setup to only allow testing. Normally a comment later and it agrees it was pointless, but unless you have something in context to force it, it simply defaults to "test all the things".
still get walls of text sometimes
I don’t trust it with code anyway - every feature needs comprehensive test, and then a live deploy on a real working test system before it is approved - I mostly measure success with - after deployment is it doing what it’s supposed to be doing. It’s like “helping another team managing their work stream” experience rather than coding yourself.
Funny enough models seem to have personalities and the dis on each other - when I had an opus orchestrator dispatching fable workers, they would comment on how “unreliable” it was and it had “evidence to prove it” and fable thinks opus is too rigid and needs more hand holding… it really starts to feel like managing team egos and verifying work. And I code scan mostly to just spot check if it’s not doing anything super stupid. But my goal is to make sure anything shipped is easy to change and fix, and every mistake has a test behind it so it doesn’t happen again.
I ship more problems, but they get discovered and fixed quicker. Before they reach prod of course. And from time to time you do reorganisation and refactoring passes where I brainstorm how things could have been better with the help of evidence- chat sessions, tests, bugs etc.
It feels less like rigorous engineering and more like gentle gardening. Or like “project management” not “coding”.
Honestly given my age now I’m fine with that. Have enough “hard” projects under my belt (ORMs, sql parsers, etc) that I don’t feel I need to prove anything to anybody, but I don’t think that’s even relevant- the velocity change is … I guess around 5-10x for me - with provable metrics, so I try not to lent the good old days but figure out how I can now live in this brave new world and be happy with my work.
If you don’t trust it with code, surely you need the code to be readable so you can understand what it is writing?
I'm not saying that's good or bad because I don't know, but I think that's the idea of dumping all that junk into PR descriptions.
However, annoyingly, we still need to review those descriptions very closely, because the robots are trained to put a lot of weight into things they read in the documentation. And they tend tospresent loose speculation as fact. They often end up documenting some assumption that isn't true, then end up writing code as if it were.
They're just helping you understand the whole picture!
> Features are coming out and metrics are improving, but we’re basically in a post readability code base, with the occasional performative comment about a variable name.
You futilely grasp for control and it eludes you. The Way is to ride the tides of life, move with the forces that shape you. Your code base is in the hands of the Machines now.
I've created myself a pre-commit harness hook to explicitly discard superfluous or too lengthy comments. Within code-review I also added comment review as blockers
Maybe I'll sneak in a change to CLAUDE.md telling it to stop
But half the time these paragraph-long comments don't even make sense, or refer to a previous iteration of the branch that has since been deleted and nobody including the author ever looked at
When technical language is necessary instruct the agents to use ASD-STE100 Simplified Technical English.
- PR descriptions should be human-written, you are a human communicating to another human. it should be limited at 4 bullet points, and maybe a "details" section in the (rare) needed cases
- in general, if you want human attention, produce human effort
After reading the long verbose promising article, in the end, the guy (with the help of the agent) only managed to implement access to the PCIe config space so that lspci on the remote machine works and shows the remote PCIe device, but that's all. It never addressed the issues above nor even mentioned them. The code was AI generated. The article was AI-written. The article never made a reference to DMA, interrupts, MSIX-X, IOMMU, IOTLB, virtual memory, etc, but it made big claims on next-gen datacenter disaggregated architecture, boosting GPU utilization, reducing large scale inference costs, etc.
Anyway, you get my point: big long beautiful words, but zero nuance.
Sometimes that's a yes, even if AI generated it and I can tell despite someone making half an effort to be a human for a minute during the PR process. But sometimes it's not.
It gets really dicey when you've kept an architecture a certain shape because there's a difficult feature you intend to slot into the negative space. There are two or three ways to implement that feature, and the one they chose just stomps all through your potential. But that's why we have code reviews. You don't have to read my mind if I just tell you. But it can be frustrating if someone asks you to do it a different way, and then has comments on how you accomplished that, and then comments on the fixes for your other comments.
My point being, docs were already an unsolved problem in code. Of course LLMs dumping reams of docs is not good, but are we really sure it's worse than before? Undocumented code is bad, and mis-documented code is the worst.
If I'm reading something on the internet, I'm either reading it to learn, or I'm reading it to be persuaded. If I wanted the LLM to teach me (thank you, no), I would ask an LLM. I'm reading your website/newsletter/email because I want to hear from you.. If you can't be bothered to put your time into writing it and teaching me what you think, why should I be bothered to read it?
I do feel like there is some middle ground where AI helps, but there's this one guy who keeps emailing everyone with walls of dumbass text and does not integrate or understand the feedback he gets when we are telling him what he needs to do. He just runs it though the LLM and replies and then forgets half of it in the next email engagement which functions as /clear on his end apparently. Like... dude I have ChatGPT and Claude, too. Actually read your goddamn emails.
it probably doesn't help that the US education system is so deeply broken with how rooted it is in segregation-era practices [0][1][2]. the average gets dragged down quite a bit when some districts receive so much money that they're flying their Spell Bowl teams to nationals and putting them up in hotels while another one a few miles away can't afford to pay for textbooks. average reading levels suffer and consequently so too does the effort that people put into writing well and reading critically
[0] https://www.shankerinstitute.org/segfunding
[1] https://www.jchs.harvard.edu/research-areas/working-papers/s...
[2] https://edlawcenter.org/research/the-color-of-opportunity/
Depends on his countenance and how good a mood I'm feeling in!
In any case I might well ask how much time went into at least reviewing the output. Or I say (rather than a flat AI;WR) "I'll scan that and have an AI summerise it for me later".
I would crumple up the flyer, throw it on the ground, and then ask for another one. Then I would crumple up that flyer, throw it on the ground, and ask for another one. I would repeat this until he refuses to give me another flyer.
If not, I would politely keep my mouth shut, then pocket the flier to trash later, and on the way home I would complain about it to my wife.
Right, exactly, that's what you'd do whether he spend 10 hours on it or 10 minutes.
Suppose you check something with ChatGPT, get nine hundred words, edit that down to three or even one and send that. You won't get AI;DR for "I checked, we're good."
Maybe. The issue here is we know most people are simply posting 900 words wholesale, maybe editing down 20-30 words to try and sweep some dust under the rug (meanwhile the trash is still fully featured on top of the carpet).
For your style: I'm not a writer but I question if it's even more efficient to edit 900 AI words down to 100. I can see some people finding it easier to change other's words than generate your own ideas, but I don't think it's a majority. I certainly know it's more annoying to edit 900 lines of AI code into a succinct 100 line function/class.
I'm sure that's the intention, but the alarming trend, at least here on HN, is crying wolf at the first em-dash.
Please note the key item is without checking. I'm not denying that many AI-generated documents are likely to be garbage. But refusing to read them just because of it will result in a lot of throwing the baby out with the bathwater.
In my opinion, this is not the same as people who use AI comment bots.
Of course, other cheaters had other reasons, but the few I was friends with had similar motivations. (Actually, a few just created tools because they enjoyed creating tools, not so much actually using them.)
I'd like them to go further and delete links to AI-generated content, but I haven't been able to persuade management to do that yet.
(Sharing next sentence for the newbies, since you know-) All our jobs to flag on site & even email the worst offenders to hn@ycombinator.com . That said I rarely flag or even downvote anything. People either really behave, or perhaps the rulebreakers break rules that don’t grind my gears. Politely admonishing good faith rulebreakers is another tactic in the arsenal.
AI-generated comment posters with comment histories of exclusively slop get email alerts from me. Don’t know whether they’re planning to sell or spam but they clearly need help finding the door :)
Very low effort the ensnare someone and waste their time for a bit, even though it is pretty quickly obvious what they are doing.
The above still implies some effort going into good prompting, however. I completely agree about the uselessness of "build me a castle; no bugs plz" style generated articles.
It is almost always low information density. Sometimes I can get it to remove most of the extraordinarily cliche phrases it injects into everything, along with the tortured, gauche prose (especially Claude). But what remains is an expansion algorithm: it takes a simple idea and expands it like polyurethane foam to fill however much space it can take, the volume filled mainly by grammatical syntax and only traced with any actual semantic meaning.
https://claude.ai/share/ac747c6d-2a62-40b7-b33e-06a94cf34344
Because it wasn't pre-AI. It was normal for people to post links as arguments, counterarguments, etc. It annoyed me, and I didn't bother clicking most of them, and would occasionally tell them "If you can't bother articulating your thoughts, I can't be bothered with that link".
But I was always in the minority. If that behavior is acceptable to the masses, so is AI generated content.
(And, BTW, the "link as argument" bugs me a lot more than AI responses. If I know the person, I can assume the person has done some due diligence in reviewing the AI response before sending it to me.)
> If I wanted the LLM to teach me (thank you, no), I would ask an LLM. I'm reading your website/newsletter/email because I want to hear from you.
To play the devil's advocate: it's far from a given that had you asked the LLM it would have given you a comparable response to the one you got. You're precluding the possibility that the other party instructed the LLM to write what he intended to say.
No, linking blog post and posting AI generated content is not the same. One is ok, other is not.
Meet "the masses".
Sorry, but no. Sending me a link as an alternative to having a conversation is totally not OK.
And it is not the same as "I had AI generate this".
Please reread my comments. That's not the scenario I was discussing.
This is, of course, before people started sending Instagram Reel or YouTube Short links. Or, maybe worse, links to 3-hour-long YouTube podcasts.
As with everything, it depends on who is sending it. For some people, yes. For the rest (majority), it's a case of
He: Makes some claim
Me: I question some aspect of it.
... Potential back and forth
He: Oh, just look at this video/article/book that explains and answers all your concerns.
Thing is, in the majority of the cases, it didn't address my concern. I'd write up something in detail, and get back either silence or yet another link. Took way too long for me to realize I shouldn't bother clicking the link to begin with.
Too many people put up a lazy comment when it is clear they haven't thought about $it, googled $it, or asked AI about $it. Where $it == {whatever they are unsure about or arguing about}.
The lazy commenter desires a human to waste their time writing a response. Often others judge that the commenter doesn't deserve the time or effort, so instead of a thoughtful response others often just reply with a relevant link - giving the commenter a chance to learn or reply.
Unfortunately offhand responses like LMGTFY become a common behaviour in forums, even when wildly inappropriate (as you note).
When it happens to you, it isn't personal, you are just a victim because everyone has been trained to a pattern of behaviour by thoughtless commenters.
It is also likely your comments get pattern matched as not worthwhile reasoning clearly to. You can do something to change your writing to avoid appearing waffly/bullshitty/unworthy (or whatever the reason responders are not replying to the meat of your point). Here's where I've done that pattern match in the past although usually spent time writing reply too: https://hn.algolia.com/?query=by%3Arobocat%20duckduckgo&sort...
See also: https://meta.stackexchange.com/questions/19665/the-help-vamp...
Sorry in advance if I've misunderstood your point or if I'm being Captain Obvious.
People replying with LLM responses is not really functionally different from lazily responding with a link. If the OP is astounded that LLM responses are tolerated, where was he all those years when people were lazily responding with links?
Perhaps it is a wider problem of "flipping the bozo bit", that someone perceives they are being treated as worthless (or worse) and then they respond similarly?
A considered AI response can be good if quoted correctly. Unfortunately I've seen a lot of AI-hate, where even fantastic AI answers (carefully vetted/curated) are shot down.
After a few years of this I also just stopped clicking thinks and refused to engage in conversations that seemed to be going that direction.
Not so. Posting links is not the same. If you were to copy+paste someone's blog post into the forum and post is as your own, without attribution, it would be seen as plagiarism. Everyone understands this; you're trying to pass another's work as your own.
On X there's actually a good modern equivalent to "posting links" which is @grok, and letting Grok add to the thread. There is occasional teasing but it's nothing like posting slop under your own name.
ChatGPT actually allows something like this; you can share a link to the conversation.
Everyone is free to do this, instead of using ChatGPT as a ghostwriter.
"ChatGPT says ... "?
> Everyone is free to do this, instead of using ChatGPT as a ghostwriter.
I guess the bulk of folks use ChatGPT as the preferred chatbot, but people like me have a completely different interface that doesn't use ChatGPT. I can't send someone links to my conversation.
(And as an aside, over 80% of ChatGPT conversation links I click on don't work - not sure if they clean them after some time, or I need to be logged in, etc).
And even if it did work, I really don't want to read the whole conversation. Just send me the relevant part that ChatGPT said!
Does it look lazy? Well, if you respond _as its sockpuppet_ you look like a hustler, too.
Arguably things like Cliff's Notes and "for dummies" books were sating this impulse before that, I wouldn't be surprised if the line can be traced back further.
Since dropping LLMs altogether I've made the connection to the impulse for a superficial jump to the conclusion just by noticing how I use web search, or feverishly scan some text to get to the point.
Theoretically I studied critical thinking and have learned decades ago how to recognize when this is OK and when it's a superficial placation of some addiction to closing the book on a subject or micro-subject... but boy whatever good instincts I might have developed back then have significantly eroded over time. Trying to reclaim that again is pretty humbling.
That is not currently possible. At best, you could train on millions of personal examples of style, eg historical emails, reports, etc, and you'd still never get what the person would have written.
I'd be astonished if an LLM didn't even miss important nuance, data, and focus on the wrong point.
It think the parent's point was, hearing a summary of what someone was thinking, from a second party isn't interesting as data from the primary source.
When you come to listen to Bob the Great, having a minion walk put and paraphrase Bob's words isnt the same. Human, LLM, or not.
You wouldn't get it verbatim, but your mistrust in AI reflecting what you intended to say is unwarranted.
I use LLMs to clean up poorly OCR'd hand written notes. It's the exception where I need to "fix" its output. It figures out what the notes intended, and writes it up fairly well.
> It think the parent's point was, hearing a summary of what someone was thinking, from a second party isn't interesting as data from the primary source.
For a lot (most?) of AI slop, I agree. But pre-LLMs, this was universally a bad idea. The first draft always sucked, and was easy to misunderstand. A good LLM, with a good system prompt, will usually produce something better than the input it is fed. Most people are poor at expressing their thoughts.
> When you come to listen to Bob the Great, having a minion walk put and paraphrase Bob's words isnt the same. Human, LLM, or not.
Most people aren't Bob The Great, but Bob The Incoherent.
Yes there are deterministic tools for this, but grammar is actually kind of complicated and the tools either miss mistakes, or suggest obviously worse edits. Looking at you, Microsoft Word.
All the writing is my own and I read all of it. Naturally there’s no LLM-isms because I wrote 99.9% of the original text. But, it was AI assisted. So, I don’t think it’s cut and dry.
Personally, in this new, different context you've stated, I don't think many have an issue with that. If you write 6 pages of text, and then ask an LLM to look for errors, that's not remotely the same, and it is more like a spell checker, or a second set of eyes.
There has even been an advert extoling the behaviour (I forget which phone OS brand it was for, which is a shame as I might like to try avoid buying anything from them): "if a friend asks a question and your phone can answer directly, shouldn't it?", to which my gut reaction is "absolutely fucking not, it shouldn't". My more considered reaction is pretty much exactly the same.
For a start I'd consider it rude and if I found out it was happening reduce my fucking my communication with that person, or call instead if messaging (unless it seems they have a fake them taking calls, no doubt that'll be a thing sooner rather than later). And quite frankly I will reply when I'm good and ready, thankyouverymuch, instant automated reactions plays far too much into the "always available" thing that I don't care for even without AI.
I look forward to future news stories where people with such features turned on suffer PII exfoliation or other hacks when someone finds a prompt injection hole, or just uses the feature as designed because the user has been too open with what they have given it access to and who it should respond to (perhaps by inaction, because the defaults will likely be wide open for "convenience" until there is uproar).
That it is plausible to have come from any of the major ones just means that they all should be avoided. Which is of course not a very practical insight.
> unless it seems they have a fake them taking calls, no doubt that'll be a thing sooner rather than later
That would be the point where our friendship is over (or rather I acknowledge the fact that it was already over) and I delete & block the contact.
If you're not familiar with the concept... https://letmegooglethat.com/?q=lmgtfy
None of this is new. People who act like they were born yesterday at every instance in this parade of slop induced phenomena are more grating to me than the slop itself.
Just take your response. Reads as incredibly lazy to me. I've read pretty much the exact same sentence so many times by now. At a certain point, what distinguishes a mob of humans repeating each other from an LLM? Original, high quality commentary is exceedingly rare. Surely you must already know this. Of all the words written on the internet, most must be garbage. It was not LLMs that tipped this balance. In my experience, the average LLM actually speaks as a superior interlocutor to the average person online. It comes with the benefit that if you get a lame response, you probably deserve it for asking such a lame question.
Most humans are not worth reading. Sorry, but this should be obvious to anyone who has tried to research or investigate anything with significant rigor. People don't appreciate enough that many times when something comes up where the LLM is doing something "dumb" it is obviously because they are trained on a corpus authored by dumb humans. My favorite example of this is when people complained in a geopolitical simulation the LLM happily deployed a "tactical nuke" as if the phrase "tactical nuke" is not itself aptly described as a wholly human authored hallucination.
People are using LLMs at work or w/e to get back time from their corporate bosses that don't care about them. Not everyone has the luxury of being able to act like an enlightenment era university student. Most words published have to be akin to "we'd like to reach out to your about your car's extended warranty" anyway. Many blogs are more or less cover for the same, or at the very least indirectly subsidized by such commercial interests. If that is not offensive, frankly it's not clear why it would be more offensive to have an AI author it. It's mostly all pablum anyway.
No one who actually cares about a topic will use a LLM. The rest of them might as well, because it makes almost no difference. If you had a clearer model of this incentive structure you doubtlessly on the whole benefit from as a modern human typing on hackernews, instead of this knee-jerk reactive one ("universally offensive and reviling") you would learn to just move on and get over it. If you want more people to have the luxury of your enlightened discernment, you should probably celebrate LLMs automating away the majority of so-called "knowledge work" so that the economy, which has largely already solved for basic human needs, can achieve a state of hyper optimization, algorithmic foresight and UBI where none of us have to work and we can spend all our time in symposium with each other, given that is what you seem to desire for yourself and other humans. We will not get there if at every turn LLMs are demonized on these incessantly banal terms.
While I'm hoping the AI bubble pops sooner than later, as I still hold out hope that knowledge work will continue to require knowledgeable humans at the helm, I wanted to add on that I think you make a good point.
A lot of users online can't read[1], and between a lack of pure reading comprehension, poor media literacy habits draining one's brains and ability to focus, and finally the increase in AI-enabled content, I'm not surprised at how redditor-cultured netizens love to throw up their hands and say "AI;DR!", as if that makes them somehow morally superior.
These users won't read dense texts or try to elucidate meaning out of human-written works, they ask for AI summaries or manual "TL;DR" entries as it is, now the AI is use is being used as an excuse to enable mental laziness and lack of attempts to read information.
So while I get lost in wikipedia walks because I'm looking for written entertainment for 'learning', they are getting mind-melted by algorithms, therefore, an uncharitable part of me wants to let the meat-proxying continue.
And then the other part of me that gets annoyed by having to sift through AI-slop-infested text to see what if any meat there may be in there decides that perhaps that's not as helpful of an attitude as I should take after all, because if everybody decides to meat-proxy the world away, I'll be unhappier than letting lazy functionally-illiterate-users complain aloud, even as the latter is boring reading as well.
[1] in the https://coding2learn.org/blog/2013/07/29/kids-cant-use-compu... sense
I view LLMs optimistically as the catchup phase. After all, they aren't really "reinventing" anything, rather they are a tool to synthesize the vast human corpus and catch up to the scale. Yes, at the moment they are used for all sorts of things, most of which are trivial. But when the bubble pops it will not mean knowledge work has a renaissance, rather it will be the nail in the coffin. The bubble pop will mean it will be cheap, compute/LLMs become a utility. So let the AI fill out the spreadsheet. Allow UBI to flourish, implement a severe wealth tax. End pointless wars. Slim government, which is a lot of bloat and obfuscation around what should be automated, open and transparent disbursements. End private taxes on the economy (credit cards duopoly, etc) and essentially, everyone gets to act like an enlightenment era university student. On this view, the hope would be that the present times are the wars of religion, the storm before the calm. If we are able to make it through in a hyper-networked, nuclear age.
LLMs are not going away, no matter how many Americans may want to go back to the proverbial land. Reactive movements have always existed, and always failed. I think you would simply be better served by advocating for the best case scenario, rather than aligning yourself with the AI skeptics who as you acknowledge, are also rather widely boring and intellectually lazy - that which you seek to remedy. The question of assuring UBI is not obvious. Imo there is far more rigor and interesting questions in the camp trying to create the optimistic future, than in those who resort to unexamined, parroted skepticism.
1. This work was largely optional to begin with. Few people actually need to post on LinkedIn, write blogs, etc. It's arguably a mountain of content waste (which is why the term "slop" is so fitting). Not only does it waste the producer's time to produce it, but it also wastes the reader's time to try to consume it.
2. It's not a substitute for actual work. If it becomes clear that you're using LLM to do work you are paid to do, eventually the person who pays you will notice, and replace you with the LLM. If your employer values you because of the work you do, it's in your best interest to do the job yourself.
2. Yes, people get replaced at work all the time. And actually most employees who figure out who to maximize their own ROI (and thus their companies) get rewarded, not punished. If your slop code is "good enough" and you are moving the metrics the C suite cares about, your colleague taking far longer to create far less is obviously going to be punished.
If some sea change happens that replaces a whole class of workers, that is beyond the scope of the individual worker to think about. The modern economy can be viewed as one long process of automation, where most workers have been subsidizing their own eventual replacement, given the privileged position of capital over labor. This is historically obvious. It does not mean it benefits the individual to go back into some state of nature, or not reap what benefits they may gain from the prevailing system while it is still the one at hand.
Perhaps it's worth looking at this issue from the demand side instead. Who is the customer or buyer of the AI-generated work; and if there is one, how much do they value it? I contend that there are few actual buyers who are actually asking for the AI-generated output, and thus the value is low. These buyers are usually asking for some other outcome that flows downstream from the work, and AI slop is either being misperceived as the actual ask, or is a step in the middle somewhere.
Next, I wasn't talking about slop code. I was talking about slop narrative work--the work preceding the coding phase that gets teams and goals aligned, and reports results to management. In a world in which code is increasingly machine-written, the focus turns to the planning and decision stages, and that's where communication skills are most important. If you're outsourcing that work to an LLM, you're providing little added value, and that makes you replaceable. Not by a machine, but someone better than you at writing.
As for your last paragraph, I'm having trouble understanding whether you're in favor of AI-written content (not code, let's put that aside). That's the question I'm more interested in discussing than how economies evolve. I know how economic progress works ;-)
I similarly don't think your second point makes much sense. I think people going on about writing or domain experience as a skill are missing the obvious writing on the wall. If you are not an exceptional writer you will be replaced. Most people will be replaced then. Or at the very least, their diminished market power is inevitable. The end state of these models is in my view clearly aligned to less human in the loop and more automated self prompting. Most people's prompts are not that unique. If someone has a superior prompt that eventually gets integrated into the model. Eventually the model learns to ask itself the superior question, even if the spark is a profoundly lazy human invocation. The progress along code generation is then joined hand in hand by progress to full automation. Why is generating a prompt harder to solve than generating code? Why will both not advance in the next decade? Suggesting code will be increasingly machine written while saying humans will still need to be around to write instructions just strikes me as denialism of the scale of the change. If you think that in 10 years most of the back and forth is not going to be LLMs talking entirely among themselves I think you are in for a rude awakening.
As I said earlier, if a human cares a human will author. As for the rest of it, it makes almost no difference. Bad, lazy writing has always existed. Why would I take solace that a human has produced slop over a machine? The question is not "do I like it" but "do I care". The answer is no. Most things in the economy I don't care for or like. But certain things are inevitable as humans conglomerate. A well organized mind learns to get over it. Quite frankly if LLMs push LinkedIn over the edge and lead to the destruction of that platform I would be pleased. It was always a center of corporate slop and backpatting, something that in theory was optional but was more like a mandatory career chore, a step in the middle to actual work. It just is kind of amusing that your example of why AI writing is bad is because it is ruining LinkedIn, as if that place wasn't always trash. It just makes my point. In domains were humans care about human authorship, it's just not an issue in the same way. I've never heard someone care about LinkedIns viability as an ecosystem from someone who didn't work there. If LLMs are making the site unusable they are doing God's work.
That is precisely the observation we're not seeing.
What is the basis for this claim? To me, my formulation reads as logically self evident, to the point of near tautology. Your claim is what? People who genuinely care about things are delegating those things to LLMs? Then in what sense do they actually "care" about them?
I also think, respectfully, that you're overindexing on LinkedIn. I was using it as an example to illustrate a point. My primary concern is with teams I work with day to day, and to a much lesser extent, AI-written blog posts being featured here on HN.
Thanks! I appreciated it.
Edit: I see what you mean. In my formulation the "human" in question was one and the same. Perhaps this clarification makes my point more intelligible. Thanks again!
You are likely not the target audience anymore and they don't care if you're going to read it.
- I'm open to seeing information from an AI (I'm fine with an LLM "teaching me") so long as it is high quality. That does require human inspection.
- I've seen a related bit about AI-written emails where the commenter would rather see the prompt. I agree. As a software engineer, I like reproducibility. If an AI outputs an interesting result, show me the "source" (i.e. the model, effort, prompt, etc.). I wouldn't say no to an added one-liner summarizing a long response.
- In the end this will be a difficult attitude to prosecute as it boils down to a Chinese-room-style preference. You want to hear from some being on the other end of a labyrinth of transmission lines and logic gates. As language models continue to develop, what goes on inside the room will be difficult to verify.
- The day may come where, rather than people passing AI off as their own thought, they will pass off their own thought as AI. "If I wanted a human to teach me (thank you, no), I would ask a human." The coming generations may value AI's opinion much more highly, whether that's right or not. And people may want to usurp that credibility for themselves.
this a thousand times this
LLMs are information transformers. if you're trying to take some rough idea and blow it out into something that "feels" substantive, you're just combining your high-signal idea / prompt with a bunch of meaningless bloat and noise from the model weights.
I think we're already there. We started having database issues after a rushed deploy (the PR diff had +100k lines). Everyone was pasting random Fable output as gospel and grasping at straws running random "create index concurrently" statements directly in production while I was slowly reading library documentation, testing hypotheses, gathering logs.
One colleague was completely in denial that his vibe coded "optimizations" that went into the update could have anything to do with the problem. I eventually found that the "optimizations", naively trying to optimize for page "first contentful paint" were prefetching pages whenever users hovered over anything and creating a barrage of requests.
And people seem really proud of what they "generate" too, not at all worried that they might be offloading the consequences to others. It's a stupid, tiring time to be working in software.
In the book-publishing context, ghost writers have to be credited, even if the attribution is in small print.
If your favourite creator is engaged in generating this style of content, then I think it says more about your tastes than their morals.
Don't smash the like button, don't share and unsubscribe.
Like I don’t need to be placated by someone else’s AI agent when I call out something in a PR, especially when they have such an aversion to conciseness and resolution
Back in the 00's you would have been allowed to mock, abuse and bully people who did stuff like that in the workplace -- and the shame might have put an end to it. Maybe the loophole is you tell their Ai agent to f' off not them.
LLMs enables them to post (or, rather, paste) something, in an attempt to trick others into believing they do have something to say.
They think they sound smart now, because LLMs are "very much the “stupid person’s idea of an intelligent person”"[1]
So now we are flooded with stupid people's attempts to show us how smart they are.
I remember talking to a business type who was so incredibly proud that GPT could turn his incredibly vapid thoughts into bland pointlessness on LinkedIn... not realizing that everyone in his feed was doing the same thing. We never did business with him for ... reasons, obviously :)
Various youtubers I subscribe to are using apparently AI-written scripts now. They all wrote and spoke just fine before, each with their own particular style. Now they just sound like Claude. I know it's still their thoughts, and I'm still interested in what they have to say. Maybe they're using AI to save time, or maybe they've read too much AI text and now that's how they write. But it kinda takes the sparkle and personality out of everything.
In this era, I live by two principles
1. Do my best to avoid AI slop. If you agree, this is straightforward. I don't always succeed and I'm sure I've "fallen for it" in quite a few places, but I disengage if I sense it.
2. Let sleeping dogs lie. Because I'm not some almighty judge of AI, I'm not going around accusing others of doing so. There is zero benefit for me to call out random internet pieces on if something is AI. And ultimately, I'm in discussions to engage with humans, not to monitor bots nor even to argue about the existence of bots (I feel I've said all I can on the topic, I have moved on internally in terms of discussion).
>It was apparently good enough for people to praise it specifically. Why is that a problem?
Writing is subjective, but I can certainly see some dangerous anti-patterns develop if you take bad advice as a beginner, especially advice that ultimately came from hallucinations in an LLM.
I was watching a Youtube video about "developing taste" and how culture has been impacted in the modern age. TYou risk becoming "tasteless" consuming AI slop, because AI by its nature is trying to average out all the inputs and cut of all the edges of any given culture. That's not necessarily a bad thing if you just wanted to get a brief overview of a culture, but is poison if it comes time to try and become part of that culture.
as a huge summation of an interesting video: There's 3 aspects of a culture: embodied,objectified, and institutionalized. it's never been easier to pretend to be a "poser" in this era of AI and be rewarded for it, because AI makes objectified easier than ever. But it doesn't mean it truly embodies you into the culture. And that's dangerous for multiple reasons (in my eyes).
Well that is not really what happened at all. Most people seem to believe Hank Green was honest when he said the line "thanks for the pushback" was in fact made up at the spot and not by an LLM, and the scandal would have died there.
Except in follow-up to that, Hank Green made several very open and personal posts about his LLM usage getting completely out of hand to the extent that his wife and brother called for an intervention about it, that it's been like an addiction to him, and that he had been letting it guide him way too much and how this was a breach of trust to his audience.
There were also one or two other incidents for people who are paying attention that this did another rather large dent in trusting him as a "smart person who says smart things", any more.
they probably weren't.
Really, does it matter? Lots of YouTube comments are zero-effort attempts to juice a video/channel's engagement metrics (aka "commenting for the algorithm"), and whether they're accomplices/sockpuppets, or just unimaginative parasocial parasites, is basically unknowable from the outside.
Then you say "it's because I don't like human eyeballs in my food", and someone and replies "but the mass market clearly doesn't care. You shouldn't be shocked by this."
uhm this is so heavily dependent on the kind, type and model of AI used that I'm calling you out on making this "statistic" up on the spot.
but I'm not entirely sure what your point is
if you don't tell people which food has been hyperprocessed and laced with corn syrup or whatever you can imagine, they will also prefer that one
also 9 times out of 10, people prefer the needle with heroin in it
Now? My friend, we’ve been flooded with stupid people’s attempts to show us how smart they are ever since internet forums came to be. Heck, just look at 90% of the comments on any given HN post. It’s mind-numbingly stupidity by the handful disguised as PhD talk—big words, fancy terms, heavy phrasing, all to say a bunch of nothing or just pretend they have a part in said conversation.
But I don't buy your argument: if a person used AI to produce a text it doesn't mean they haven't steered the I in the direction they wanted the conversation to go and (hopefully) review the output.
That's an added value
Yes, it would be nice if it was also pleasant to read
"Why are you upset with this restaurant meal? Okay it wasn't actually cooked here per se, but the chef certainly reheated it and carefully monitored the microwave oven while doing so!"
This crap just needs to end. "Steering" words at another human being is incredibly disrespectful.
The chef is however responsible for the outcome so you're absolutely right to complain if the meal sucks and most of this AI written crap definitely sucks to read.
That said, it's not about the "principle" of having your meal cooked by the chef / the text written word by word by the human.
It's about the outcome. People should learn to tame their AI tools/harnesses in such a way to produce quality output (code and prose) and it's absolutely despicable that so many people are so eager to lower our standards in quality just because everybody else seems to be doing it.
No. It very much isn't about outcome. It's about interpersonal relationships.
> People should learn to tame their AI tools/harnesses in such a way to produce quality output (code and prose)
The quality isn't the problem. The quality is that something that needs to be "tamed" is writing words that the "tamer" expects me to read.
> the fact that i'm at risk of seeing a 14 year old's opinion at any point during my day is a human rights violation
This always strongly resonated with me; if something was written by a 14-year-old, I don't want to read it, end of story. There is no way it can provide value to me. I shudder to think that I may have unknowingly argued with 14-year-olds on the Internet in the past.
But if I had a child who was 14, I wouldn't object to interacting with them, obviously.
It's the same with LLMs: the only time I want to be reading LLM output is when it's my model that I prompted myself.
It astonishes me that in 2026 AD, people still like to pretend that our calendar is not based on Christ's supposed birth. Christians have no problem calling days after pagan gods (Wednesday = Woden's day, Thursday = Thor's day, etc), but for some reason we have to pretend like the calendar isn't based on the Christian god.
I like how this could be offensive to both Christians (Jesus isn't god) and anti-Christians.
Recently on a call with a junior coworker where they asked me to look at some changes they had made, they were completely unable to answer any questions that I asked on their own. The way the conversation went was: I ask a question, they type my question into a ChatGPT prompt, we read the output together and discuss whether it makes sense. This is a person I've had great discussions with in the past, and who has frequently had really good insights, but seems to have completely tapped out.
To be fair, the outputs from ChatGPT tend to be quite good. It's the inputs people give it where problems most often arise - and people seem far too eager to charge ahead with a vibe coded solution built on invalid or incomplete inputs/assumptions. And it's shocking to see how quickly and willingly people have handed over control to LLMs.
I'm so tired of AI Slopped architecture documents, code, messages in slack, emails etc. I'm close to flat out rejecting any architecture doc I get that's written by an LLM, because I'm not sure the engineers even understand what they're proposing. The code is often a nightmare under the hood, even if it works. I've got smart coworkers who now just copy paste everything via an LLM giving themselves something approaching a lobotomy in the process.
I may just get out of the tech industry altogether. But I'm also tempted to try my hand at an indie dev project that I have in mind. With AI, it seems like it would be easier than ever to get a new project off the ground. And I don't really mind AI-assisted development personally. It's mainly all the organizational and interpersonal disruption that is giving me a headache at the moment.
So I've disabled that and now I'm explicit if I want to solve a problem because then it means my brain's actually engaged.
It seems to be like a shortcut to early onset dementia because the brain's never being actively involved in problem solving.
I do think that it is very difficult to not be a meat proxy when you utilize an LLM for writing tasks. These tools are computer programs that generate text, so what else would you do with it than have it write for you? This is being a meat proxy.
Trying to avoid being a meat proxy can quickly eat the supposed productivity gains of using an LLM in the first place.
Like say I need to write an article about some subject and it would take me the entire day to complete this task without an LLM. I can get some sort of article out of an LLM by simply asking, I don't need to think about the subject myself at all. The LLM productivity promise is also fulfilled, because I did an entire days work in a minute or so.
Now if I also need to read and understand the work, and be able to defend it or explain it to others as if I produced it myself, then this will take a lot of time. If on top of this I need to edit the text to sound like me rather than an LLM, then I probably still need the whole day for this task and I have lost all the productivity increase that I sought from using the LLM.
It's the thinking that takes time when you write, not typing. But you can't do the thinking and avoid doing it at the same time.
The same idea applies to coding too. Doing code review and requiring the author and reviewer to read and understand the code basically eliminates all the productivity gains that using the LLM was supposed to provide. Throw out the reviewing, reading, and understanding and we can have each developer producing a thousand kloc per day, over 1000x productivity. If we don't throw those out, we are comparatively in the exact same spot as not using LLMs at all.
Of course, given that posting one is not mutually exclusive with posting the other, I would still think that posting both is the easiest way to satisfy both people who prefer to read the prompt and those who prefer the read the output.
Sharing information that is already in the corpus is simply not that valuable in a world where everyone has access to their own LLMs.
If I think the task requires a human to think about it. And the person is the human paid to think about it. Whether I get a wall of AI text, prompt, or an entire AI conversation... there is no human insight in it.
If that person took the time to think about the problem, had a chat with the AI, and understood the output, surely they can take the final step and tell me the short result/summary themselves, right?
---
However, if I had to choose lesser evil, I would of course prefer to have a link to the AI conversation over having a wall of text directly in the conversation.
Clearly you either didn't read it, misunderstood it, disbelieved it, or just DGAF :)
Which is fine for you, but doesn't change that many people can get more insight out of your original text than they can out of the generated adaptation, and can do so with more interest and less fatigue.
And indeed, you don't have to care about that. But one would think that in publishing or sharing your writing at all, you do in fact have some investment in whether people receive those insights that you have. Otherwise, you could just keep the AI writings to yourself in the first place.
I don't want the prompt - I may not have access to an LLM to evaluate it. The entire argument here is misguided, and deeply angry about AI.
It's equivalent to answering a question with "google it". Which is often times just plain rude.
But I do see a consistent trend here regarding hostility to AI. It seems to be a pattern, and a strange one.
I've never had someone angry at me for posting a Google search, for example.
I have.
For example, I was looking for a bathmat. My wife pasted me a list. Just a week ago.
I don't think my wife is bizarre. It's unfortunate that you do.
SO BIZZARE!
You called my wife bizarre. Then you said "TECHNICALLY, I didn't call your WIFE bizarre I called her ACTIONS bizarre."
I would ask you to stop talking to me, but it seems unlikely to work.
So. You have a nice day, now.
I do not think that forwarding a listing from Google is bizarre.
These seem like reasonable positions to me.
I am very sorry if my voicing these beliefs makes you uncomfortable.
I knew you wouldn't like me pointing it out, but the comment wasn't really for you. It's for the other reasonable people who are having to read or endure your bad behavior.
But I disagree with your “toxic”. I think HN has a fear of conflict: the “civility” argument from politics/Clinton circa 2016. And look where we are today.
Sadly, conflict is sometimes necessary to resolve problems.
For example, right now, the anti-AI thing is pure madness. I understand it is based in fear, and I have sympathy for that. But much of human ugliness is based in fear, and sometimes conflict is necessary to get a community past its fears. Humans are stubborn.
So, again, I am sorry for your discomfort.
> bub
And now I’m sorry for your anger.
Really? I wonder where else you haunt because this is the most pro-AI place I'm around, both in fervor and number.
The topmost comments were ugly attacks.
Except I'm lying, i don't want the prompt either, i want your human writing because even bad human writing is less insufferable than LLM slop
Just because those of you who disagree are more vocal, downvote and flag-happy doesn't make you right.
He would make up an excuse that is entirely untrue and then flag them.
Oh wait, I'm thinking of my last encounter with him.
But I am glad you've decided to engage. Now will you answer my question?
You have an account on HN dating back to 2013 with over 6000 comment karma, you clearly didn't have trouble communicating like a normal person for a decade on this website before the advent of LLMs. Your spelling and grammar in posts predating LLMs are also far better than what you've produced in this prompt, so I'm tempted to think that you're exaggerating your inability to communicate normally to make a point.
But even so, the original wall of text communicates the point much more strongly and sincerely than the LLM trash.
It is the articulation of ideas that I struggle with and so I usually edit my comments multiple times. And to get rid of some of the more stupid stuff I say.
To be fair, that word salad would read more like this comment if I gave it some time, but it would take many hours to produce a simple post with the same readability as the LLM.
Your style reveals the way you write sentences, the language and vocabulary you choose to employ, and even your emotion.
Here's an obvious example:
> I dont care ion people dont read my stuff. There are people that do and they find it useful. My ideas solidified have already started to bring dividends for me - and not from others reading beut from me being able to ficus my thoughts.
which converts into
>I don't particularly mind if people skip my stuff. Some people do read it and find it useful. And honestly, the biggest dividends haven't come from readers at all — they've come from the act of solidifying my own thoughts. That alone has been worth it.
spelling aside, just look at the first sentence. You show a different kind of derision for your audience here compared to the LLM trying to sugar coat it and soften the blow. That's style, and a small part of why AI feels like it "slops" up any given style.
"Brings dividends" isn't grammatically correct but reveals how your mind sees the idea. you see it as a result to manifest from your ideas. The Ai interpretation of "The biggest dividends..." grinds out the edges back to a more passive action happening more as a side goal.
There's that much subtle changes in meaning from a 50 word passage being converted. Which AI should be "good" at. Now imagine how this breaks down slowly over the 600 word post you made if I broke down every little bit of it. That's what people are doing internally and quickly concluding "this all sounds the same". Which is about the worst thing you can be in an attention economy; boring and unoriginal.
----
P.S. Yes. I don't want to come off as rude. But your prompt also does reveal how you approach writing as a whole. It's clear you a) value the presentation of the writing as much, if not more, than the contents of the writing and b) admire the writing "sounding smart and proper" but do not wish to impart effort to achieve those means yourself.
I do believe there can be interesting prompts to read that is a back and forth correcting and managing the AI to give a more tailored writing as an end result. A conversation that reveals how you go past the basics spelling and grammar and tell the AI how to inject your style and taste into the generated results. That would be true "AI-assisted writing".
This was clearly not that. This was a single prompt: "make me an essay that sounds good and is formatted well". One and done. Not a conversation, but a demand trusting the standard of the average as output.
I won't reject the idea of AI-driven writing one day making bespoke works. But I highly doubt it will be a one-and-done prompt that achieves this.
I have made longer notes using exactly the process you prescribe, but that wasn't necessary for me in this case as it wasn't the most complex of ideas - I just regenerated until it gave me a fair approximation of what I wanted.
And to be fair - even if I had given it follow-up guidance it would still have the LLM "smell". It would receive EXACTLY the same treatment from the anti-llm crowd. Thats pretty much one of the core points of the post (judging the medium ahead of the content).
AI writing is like the voice anonymizer they'd use in TV interviews when they wanted to hide the identity of the person speaking. The output is more adherant to convention but likewise more impersonal and exhausting, and when it's used by too many people it's so indistint that it becomes noise.
It might look good to you in isolation, because you're excited to see that it says what you meant to say without the features you feel self-conscious about, but to others it just reads like yet another sample of the same old AI slop. The message gets entirely lost behind that.
If you must use AI in this way, I'd tell it to take a _much_ lighter touch. It cleans things up without washing away the personality of your writing. ChatGPT example:
> adjust the following for spelling and grammar, making MINIMAL alterations to the original wording and content. do NOT rewrite, just clean up what is already there: [post]
https://chatgpt.com/share/6a8382b9-3620-83eb-b17e-4397aee5ad...
Thank you :)
There’s also a difference between “please edit my writing” and “please generate a blog post from these bullet points.”
The LLM response has none of that character, it doesn't give me that little sliver of insight into the writer. It strips the idea from the author, making it generic and anonymous. The only thing I can gleam from text edited with an LLM like this (besides the literal meaning of the text), is that the author didn't care enough to express themselves.
Sure, it's more polished, but doing the writing manually is half the point. You have an abstract idea, you refine into something that can be communicated, and in doing so, explore that idea and improve your understanding of it. I've re-written this paragraph a dozen times trying to do exactly that.
It's getting those abstract ideas back out from the word-vomit I find very difficult. It is almost completely impenetrable to me to read that text - even though I wrote it!
Someone in another comment (https://news.ycombinator.com/item?id=49338214) has shown a prompt to just break it up and fix the sentence structure rather than rewriting, which should hopefully help me to at least re-read what I have written.
...This part immediately made me link it to the hipster stereotype:
"oh, it's sooooo much better on vinyl because you can hear the cracks & pops"
Neither pieces in the chat log work for me: The prompt was a wall of ramblespeak. The output was a presentation that should've been an email.
Even the attempt of using Claude to be a fancy formatter failed when you look closely at the language employed in the final prompt. It's trying to convey the same ideas with better foratting, but still softens up the derision the author clearly has here. Which is the interesting part in my eyes.
You either do not know what derisive means or have worse reading comprehension than I do! /derision
Seriously though - it's literally just talking about the use of an LLM as an accessibility aid and argues that the LLM-assisted writing can still contain human thought. What specifically did you consider contemptuous or mocking?
FYI: https://claude.ai/share/601e313e-905c-4e4a-bb6a-24a1315013e5
> slightly ill intent
There was no ill intent. The content itself described the reason for the process. Even simple comments I write here end up with multiple edits over the hour-long edit window.
If you read my entire comment history you will find I rarely make more than a single observation per comment. Anything more complex becomes too difficult to read and revise.
Cognitive impairment is extremely difficult to understand until you have experienced it. I am particularly lucid this afternoon. That doesn't mean the difficulty is not real.
Meanwhile, the LLM-generated essay is hard to read because it's essentially the high-school essay that's padding its paper to meet a minimum page count. I'm surprised that you find it easier to read than most human text, because I've invariably found that my first reaction to LLM-generated text is "where's the point?"--everything is just buried in so much fluff.
So yeah, I find the prompt much better than the LLM output.
The practice of writing can serve the purpose of conveying an idea or information to the readers, but it can also be an invaluable tool for the writer to process, digest, and internalize a message.
The act of wordsmithing, and shaping the idea from its raw rambling form into a coherent narrative, is intellectual metabolism.
A good writer will both distill the ideas, present them in such a way that will be pleasurable to their readers, and also bring the reader along the way so they can learn as well, and feel like an active participant.
The LLM tend to produce prose that defeats the purpose for the writer, takes the joy out for the reader, and ultimately does nothing for anyone but take up space.
It's not more words that make good stories, and it's not more code that makes good programs. It's what use is this to anyone, and only humans understand what other humans want and only humans can care.
Seems like people in this thread agree that it's better to express oneself in a less skilled fashion but actually write something that represents your idea, than to present some polished turd and pass it off as your own writing.
Same as a birthday card on some stock paper written in crayon will always be more valuable than a store bought card with generic sentiment.
And then fed it your input.
And the output was a very easily readable, enjoyable plain-ascii-text, with no Ai-ick, except for an em-dash or two.
Tried pasting it here, but that removed paragraph formatting
> Iff you dismiss AI writings just because it sounds like it’s written by aI you will end up missing out on some truly valuable content.
Yes that might be the case. But with Ai now being in so heavy use, there is so much content generated that most of us dont have time to just read output of someone elses basic prompts, if we can't determine how much original thought went into it.
So, with your Ai generated post, I don't know how much of was is actually based on your thoughts/opinons, and how much of it is just a conglomerate wordsallad from a consensus machine based on random internet thoughts/opinions. Thus I ignore a majority of everything that looks Ai, just to protect my cognitive circuits.
Unfortunatly this means that I sometimes even miss out on genuine human content that is just close enough to Ai-style to trigger my "Ai filter" . I've even dismissed content narated by voice actors, such as Benjamin Pemberton, just because his style is so close to Ai voice (although his style obviously predates the Ai voice style)
And it seems this is such a common occurence that, some videos now put a "Human narator, not Ai" text on screen at the begining of the videos.
I wish we werent here, but here we are.
I don't use one because my aim is to intentionally produce the obviously LLM-sounding output as that is what I find easier to read (although as others have said, I can preserve more of my "voice", if I tweak the prompt somewhat).
Do you read over what the AI spits out? Do you read over what you write?
It takes time to do either, and both are the things that I think are actually useful.
I could not finish the output made by Claude. I tried, but having to translate back from "mouthy words" to "plain English" is tiresome.
I am not trying to say that I don't like well written words, a good book with carefully chosen words is a good way to spend an afternoon. But for LLMs, it feels very forced and not natural because it doesn't insert those words where it would make sense for a good author.
So the polite version I use now is: "Hey, just to get a bit more context, what was the original problem you were trying to solve?".
That gets them to distill their own problem a bit further.
A lot of people replying seem to be misunderstanding, though. The only point of human-to-human communication is to transmit thoughts from one brain to another. If you simply send me stuff from AI then I don't need you. I want to hear your thoughts. If you don't have any, then I don't need or want to hear from you. Your thoughts are entirely contained within your prompts. Just send me those.
Sounds like this isn't everyone, but rather, it's people who have nothing to say, and -- even after they've "sourced ideas" -- they don't know how to reason about it, nor how to communicate it.
This isn't everyone. This is people engaged in generating noise, for the sake of noise.
~actual prompt i sent in to claude last week. doing what you propose would be basically impossible / nonsensical, and further undo the entire exercise, sending all the money spent on the tokens down the toilet.
put differently, if you use agents in any actually useful manner, the principle you describe erases whatever value they did end up providing. though since the whole underlying premise is that ai never provides any value, maybe that's the intended result and is as expected. in which case, i'd argue such a desire is more than a bit demagogue.
It turned out his prompt was equally uninspired... 1 or 2 sentences. That was the thought he put into it, which then had us spending hours trying to figure out the AI reply. He could have saved the whole team a full day by just sending the prompt... or nothing at all.
"Here is the prompt and model I used, here is the portion of the output that stuck out to me as relevant and useful, and here, in my own words, is why I think that is so."
i.e. Highlight the good part of the output for me and, preferably, explain why you think it's good. Otherwise it's the same thing as posting slop, only with extra steps. You thought it was worth posting and I thought that made it worth looking at, but now I'm sorting through a bunch of monkey garbage looking for what made you think that. Being a different person, I may not find that hidden gem and I may become quite frustrated with you.
________________
It's a tool like any other. It assists the intelligent, but makes the lazy - lazi-er.
I guess the same thing goes for articles written by humans. I don't really read the articles that are "Shower thoughts that I think are profound" or "The way I do things is the way everyone should do things"
There is a fairly common wisdom that the way you get good at writing is by writing a lot, I'm prepared to cut people a lot of leeway at people putting all of the stuff out there before they are good at writing. It's just the path you have to take. That's OK, because I don't have to read it.
And I'm fine with people making crappy AI generated tutorials while they figure out how to make better AI generated tutorials. I am not fond of people who are creating bulk content by volume without any regard to quality or improvement. Some may be doing this to make money, and maybe that works, I suspect for the most part it doesn't. I don't appreciate it though, whether it be AI or human produced bulk content.
I'd spare a thought for those who feel that they are producing high quality content when they are not. Everyone has a poor perspective on something sometimes. Usually the harshest punishment they will face is the feelings their future selves experience when they look upon their work with a more balanced perspective.
I have coworkers who say they will stop reading something if they believe it is AI.
For them, it's not about quality, it's laziness disguised as righteousness.
Alice being AI or a real person doesn't change those factors much.
> Look, I get it. It’s Q3 2026, and we should expect that everyone is utilizing AI at SOME point in their process (sourcing ideas, creating outlines, refining prose, etc.).
I stopped reading there.
it's more like, "refining prose so that nobody sane will bother reading it". like wow it's so useful to claudefy your thoughts, wow look at how amazing your blog post looks now with how much more slop you produced from your prompt, amazing.
I really wonder like, do those authors not think from the perspective of the reader?
However there's only 10% of AI-generated writing that is worth reading. If some articles convince me that there's somebody behind it contributing valuable understanding, I'm willing to overlook the telltale AI mannerisms. It's understandable that some domain experts aren't very proficient in English.
That said, statistically speaking 90% AI-generated pieces are a waste of readers' time. It's a good strategy to simply skip them.
I know this looks like a bot comment. It's not. It's a much older style of writing known as Chinglish. That's why as a non-native English speaker, I can totally understand why someone would want to use AI to help with their writing.
How do people not realize how up their own @$$ they sound when they say things like this
It's frustrating dealing with them since they all use AI in communications. Everything is written so flowery and with so much extra words. Any feedback I give is met with overwhelming praise about "how right I am". Its exhausting.
I understand many of them are in developing countries and aren't confident in their English skills. But I always tell them I'd rather they write in broken English to communicate their own ideas. I want to hear what they have to say, directly from their brain. I don't have issues with parsing bad english. Usually its straight and to the point.
They always end up defaulting back to AI messaging. Even when I tell them to be genuine. I'd love to give one of these firms a chance, but they fail this basic step. I think they think everyone in the developed world likes AI or something.
I work in FAANG at a non-US office, and this is what I tell my colleagues. I always get pushback: "They're too shy to speak English".
They'll often write messages in their native language, then ask AI to translate.
This is in a developed country, by the way. Engineers working for a US-based company refusing to speak English while getting paid handsomely.
The idea was to just avoid sending email altogether and just simulate the interaction entirely using Claude.
hard disagree on this
Hell no. Sourcing ideas? From AI? Are you kidding? Essentially guaranteeing they won't be original? Refining prose? Making it much more likely it'll be generic as hell. Call me old fashioned, but while I can certainly understand asking LLMs concrete questions - how do I fix this, where can I unsubscribe from that - using AI to help communicate seems bass-ackwards. LLMs have no idea what you're talking about - they're free associating probabilistic sentences! What on earth value could that have in communicating. They self contradict, hallucinate, speak most frequently in a weird smug corporatese. How is this an attractive use case of the technology?
> This can be navigated, of course, but it is truly perilous: our writing is an important vessel for building trust — and that trust can be quickly eroded if we are not speaking with our own voice.
From https://rfd.shared.oxide.computer/rfd/0576#_llms_as_writers
Be careful not to [give up your own voice][1] to the [machine][2].
[1]: https://bookshop.org/p/books/you-are-not-a-gadget-jaron-lani...
Also, I may have already started looking into it, but now this other person just wasted money.
Not to toot my own horn, but...[0]
On a more serious note, we're planning a trip with family, and my mother-in-law asked if I'd read "the planning doc" yet.
It was a shared ChatGPT thread. Thankfully I read enough LLM output to skim it easily and avoid offending her (:
No, no, no, no - the whole point of customer support is that I am having a problem that requires attention from a representative of the business I am dealing with. Businesses really don't like doing this; they call it a cost center to have to talk to their own customers, so they staff their call centers with increasingly useless script-readers designed specifically to shield the people with knowledge of the matter you're facing against customers like you. This is a great solution for dealing with technically illiterate people who don't do basic troubleshooting and terrible for literally anyone else who actually tries to be a good customer.
I, of course, had no choice but to apologize. I hope I was right, but clearly there is some chance I was wrong.
Before I think he was using AI and other tools to translate his messages since english isn't his first language. His own words, then AI translated it. Sounded kinda clunky, but I could sense the human behind the words and I gave grace since I can only imagine how hard it is to properly communicate your ideas when English isn't your first language.
But now it's gotten to a point where I can tell he's not using it for just translation. I'm gonna need to chat with him.
Redacted example below
```Hey @PERSON_WHO_ASKED_QUESTION Both good, and the retrieval one isn't written down anywhere. In order.
One corpus or four. My lean is one. Same chunks and embeddings tables, source_type column to tell them apart.
Values I'd propose, flat rather than nested: paper, dataset_description, dataset_readme, dataset_contributors, dataset_records, dataset_files. Description and readme split because their units already differ, one row versus one row per paragraph. A discriminator that can't separate those isn't doing much. The existing 440 rows would need backfilling to paper.
Chunk id in the same spirit: dataset doi, source type, ord. So EXAMPLE_DOI.
Worth checking before any of this matters: does chunks.doi carry a foreign key to papers.doi in 0001_init.sql? If it does, a metadata chunk with a dataset doi can't go in that table at all, and separate storage stops being a choice. Ten second read, I haven't done it. Shout if you get there first.
Retrieval is the one that's bigger than it looks. Search once across everything and nothing guarantees a paper chunk and a metadata chunk both land in the top k. The facts we want relate the two, and the generator can only write those if it sees both sides in the same window. So if one type systematically wins the ranking, that class of fact doesn't get worse. It becomes impossible.
Which way it goes I don't know. Two mechanisms pull opposite ways. Metadata chunks are short, tens of tokens against roughly 450 for a paper chunk, so they may just lose. But we embed context header plus text, and on a short chunk the header is most of the vector. The header is the dataset name, which is also most of the query. That points the other way.
Cheaper to measure than argue. Load one dataset's metadata, run a normal dataset level query, record the rank of the first chunk of each source type. Runnable as soon as any one of our four subtasks lands.
After that it's one pool, per source with quotas, or one pool with a floor per source type. I'd rather not pick before there's a measurement. ```
I already saw someone getting fired for only producing AI text as part of their entire output, after being unable to explain what they “wrote” in multiple situations.
And recently my company enacted a mandate that text made for humans should not be AI generated.
Sooner or later every company will start realizing that this is just people too lazy to actually work and coasting on a paycheck, at the expense of every other worker.
Much how like essay's have a word limit - and people think you should max it out.
We will go in the opposite direction - which I'll be glad about!
There were too many incidents of “this doesn’t look right” in the same week, so when their boss did a thorough check a couple days after, it was all fake, everything. Eventually they admitted and got canned.
But AI was trained on good writing. Good writing looks like AI. I've had to reduce my use of em dashes for example because they are an "AI tell". I used to like using the term "load bearing".
And the list gets longer every day. Eventually every mark of a good writer will be an AI tell.
I don't know how to resolve this. But I've been accused multiple times of copy/pasting AI when I did no such thing, and it's getting equally frustrating.
I even studied writing in college. I’ve written published books and articles. I can write well. But thanks for the tip!
AI writing isn't actually all that bad. But it has a very specific style, and there are MANY tells. I use Claude every single day, and it's immediately obvious to me when I'm reading something Claude wrote. "It's not X, it's Y" "That's the reframe" "The gap is" I could go on.
I too like using em dashes, and have been since I learned the keyboard shortcut over a decade ago. I've never been accused of posting ai-written text because of it.
AI is indeed trained on good writing, but it's also trained on bad writing. And it's post trained on specific things that are writing that aren't good or bad for reasons unrelated to style. What you're left with is a one-size-fits-all writing style. Doing math? Doing medicine? Doing anime waifu role play? It all starts from the same generic milquetoast style until you tell it not to. And to me (and apparently many other people) that's super easy to pick out.
The typical highly voted HN AI-written post has an annoying clickbait title, then jumps into a meandering chain of “so there I was” scene-setting, assertions of dubious reliability & relevance, and supposed supporting examples. All punctuated by short, unhelpfully dramatic sentences that would interrupt the high-level narrative, if only there were one.
I’m really curious what “good AI writing” you had in mind. My personal intuition is that for almost any topic, the HN LLM-generated post du jour has a far superior cousin sitting in the archive pre-2021.
Beyond that, most AI content is WAY WAY WAY WAY WAY longer than it needs to be to get its point across and is oversaturated with unneeded "mic drop" phrases
* Cover Letters
* Most test reports
* Lots of docs required by the government
* I'm sure there are more you can think of
So now we can have stuff written that no-one ever writes and no-one ever reads? Hmm...
I think sometimes people have a big long thread with Claude where they feel enthusiastic about the back and forth being very productive -- and then they genuinely want to share the summary of the conversation with their colleagues so that they can feel it too. It's not just typing a short prompt and copy/pasting the response.
But, Claude, especially Opus is so unnecessarily flowery, and smug -- and it's so identifiably characteristic.
It somehow rubs salt in the wound: "I didn't take the time to write this out in my own words, so now you get to listen to this petulant asshole mansplain it to you".
If it would get to the point, and with some humility, it would be a much different proposition.
I don't know if the contributors are writing any of the code they're submitting or just using AI to hop on every 'help wanted' tagged ticket in every product database but if it looks mostly like a smart human wrote it, which usually takes two rounds of reviews, then it can probably go in.
But if you make me work for it more than you did I'm just going to close the PR and do it myself.
On the other hand, I suspect that we are pretty quickly going to get to the point within a few years where we realize that the best AIs are not just being overly verbose. But rather their communication is just more compressed and thorough.
So even though we still see a lot of jaggedness and weird failures in some places，in other ways I think this is actually an aspect of intelligence where AI has surpassed humans. Just in terms of being able to deal with a certain density of information.
At the end of the day I don't care how much of what you're presenting is your own words vs AI generated, as long as I can see evidence that you have reviewed and understood what you've passed to me. My resentment and ire is raised when the onus is pushed on me to be the first real human reviewer, having to do all the weighing of merit, practicality and in some cases plausibility that the quote un-quote author should have done before ever hitting send. It's passing the buck and an abdication of responsibility. It is lazy and makes more work for those you foist it upon when your role should be making the effort in your communications to make it easier for the recipient. Not wasting their time having to spend orders of magnitude more effort refuting bullshit you shouldn't have put out there in the first place.
It is something that has always pissed me off, however AI has made it fair easier for lazy people to think they sound smart and externalise their costs into my face.
[1] https://drewdevault.com/blog/Stop-externalizing-your-costs-o...
I like to think it helped, but likely what helped was that nobody seems to use Deep Research anymore, haha.
It would make sense, for example, if these people who are uncomfortable with output from others' use of "AI" are people who do not themselves use "AI"
The last thing in the world that I want is to read an entire article written by AI, nor am i happy when i ask a human a question and they don't answer but instead just copy paste to Claude and back as if that's comparable
After you've exhausted your favorite audiobooks, I mean
Like, I love The Wind in the Willows, but I wish the part where Toad and the gang go on the road in the cart lasted for longer. I asked ChatGPT to invent some fill-in chapters matching the same style and speak it out loud, and it kinda did OK actually.
Yeah...I'm probably dropping whatever services do this. If the company can't invest the capital to hire humans that help humans, then what does that say about their products?
Don't Be a Meat Proxy: https://gruhn.me/blog/2026-08-03/
No Meat Proxy: https://nomeatproxy.com/
How to Passive-Aggressively Shame People Who Use LLMs Selfishly: https://joshmoody.org/blog/selfish-ai/
I value my time. With a LLM, it takes very little effort to create a vast mountain of crap that will take a lot of time and effort to read through. True, it is also possible to use LLMs as a tool while creating things that are not worthless. But the fact that someone has not taken the time to make their thing not sound obviously AI-generated is a strong signal that it will not be worth my time.
Are you sure about that?
Sure, before, there was always the possibility that the article wouldn't deliver. Sturgeon's Law. But now it's like 10× more, and the amount of slop out there means its even harder to find that (human-written) gem amongst all the chaff. This last month especially, I've started to wonder "what even is the point of HN?" as more and more of the frontpage ends up dominated by slop. The thing that made HN worth it in the beginning is that it had a high gem:chaff ratio — at least compared to other corners of the Internet at the time I joined.
i have what ill call a "condition" where i fear assuming someone did something stupid is insulting. i try not to patronize people because i also assume that would be insulting. this leads to issues where sometimes people really want to be lead to water rather than make an implied connection themselves
this is a roundabout way of saying, is there a genuinely polite way to tell colleagues posting ai copypasta makes them seem ignorant? i dont think ai:dr is workplace appropriate. id like to tell people gently and avoid patronizing if possible
AI-writing is definitely very annoying to read, but I don't know if it's ideal to start dismissing interesting content before we even get to its substance just because we don't like its form.
A quick mental filter is absolutely needed though. But we should consider tuning that filter for signals other than AI writing. I've built one up where a quick skim through the post tells me if there is any substance there or just some shallow "it's not A, it's B" punchline. It took some work to tune that filter, but has let me find some very interesting articles that I otherwise might not have read.
TLDR: Content != Art
TL;DR works because I can see the length at a glance
And you cannot tell all AI. Only mainstream crap like ChatGPT and Gemini which is full of cached tropes and boilerplate formats so it all sounds the same.
My custom invoicing software automates notes based on years of my previous notes and tasks I did - looks nothing like AI.
Use a real library and local model, mess with settings, there is a lot more to this than the mainstream consumer apps.
https://hn.algolia.com/?q=ai%3Bdr
This piece is still slop, human or ai created
fwiw, I'm partial to ns;nt (https://blog.kinglycrow.com/no-skill-no-taste/), but it is different in nuance
Claude, post that wall of output to Slack yourself. Make no mistakes.
(Newsletter popup,. Didn't read)
No. Very much no.
Its awfully similiar to bitching about reading mails on a computer and attacking the sender.
The virtue signaling is beyond me. I dont care about AI;DR or TL;DR.
Give me good content. And I will read it.
> I’ve been thinking about it ever since. Why? Because there is growing grumbling among everyone about AI writing. And it’s not just others; it’s me!
Classic AI "its not just, its..."
Really hate it when I send a tricky spec out for review and people respond a couple minutes later with a Claude summary. Way to add value bro.
The problem is that those kinds of responses are rare in the tech world. But when you have someone competent at the helm, it's much better than reading human slop.
> AI;DR (AI; didn’t read) is the solution for AI slop.
Maybe I'm the odd one out here and didn't plumb the worse-depths of social media, but I feel I need to defend good old TLDR as a slightly different, and less-hostile animal.
It can be a moral condemnation of the poster, where they're disrespecting everyone else's time by posting something big and vague... But it usually isn't. At least half the time TLDR is someone providing a summary because they think it'd be useful (even of it's sometimes a hostile interpretation.)
I'd say that sounds about the same for ai;dr. I could see myself posting a prompt like that in front of the generated report.
- Why does it matter who or what wrote a thing? And how would you know if the /content/ is worthwhile, unless you read it...
- The ability to "detect AI" is imperfect at best. 90% AI written? 5%? How would you know, unless you read it....
- If you didn't read it, then why brag about it with an "AI;DR"?
- If you are wrong, and you announce it to the world, how would it feel to someone who spent a lot of time and energy writing?
Bragging about it is weird, but I will use it like I use 'RTFM'.
And detectability - My AI-DAR (radar), is fairly fined tuned but not perfect, so if I make a mistake, oh well, the point is I'm protecting my time.
> If someone tells their unattended LLM agent to create a blog, write whatever content will maximize banner ad revenue, and then share each post on Hacker News, would you read it? What if the content is actually interesting in your opinion?
Yes, I think if one of that content had a headline that interested me I would definitely click on it, because that's already happened!
The purpose of writing is to communicate. The person communicating is an important part of the message.
Consider the sentence "I'm waiting for you at home." It really matters if it's said by your loving wife or an evil clown. You can extrapolate from this.
It's a reference to TL;DR. If you can understand that, you can understand AI;DR.
Two reasons:
1. I can't effectively engage in follow-up conversation with the "author." I can't do it with the human (principal) because they didn't write it and can't explain it--and, in fact, they may not even personally agree with all if it!
2. It shifts the cognitive burden from writer to reader. What makes writing hard, and valuable, is the effort put into translating one's raw, unfiltered thoughts into writing that is easy to comprehend, and the writer's own unique personality that is expressed in it. As ambient prose, AI-generated writing is often not only more difficult to read than human-drafted prose, but it also has this "sameness" quality that makes every "author" have the same personality.
Because what the clanker produces is almost always bad output compared to what a capable human would produce. Furthermore, even an incapable human is better than a clanker, because at least humans have intrinsic value that they bring to the table. Finally, it's just plain disrespectful to refuse to take some time to answer a question and hand it off to the clanker.
Why would you think it would?
You have some solid ideas there but they are very poorly organized. I think you would benefit from a class in writing. I had a month-long technical writing workshop back a couple of decades ago that my company sponsored and it helped me tremendously with starting with wall of consciousness, progressing to writing plan, outline and then editing. It had a very positive impact on my career in professional services where clear communication via documents is an essential part of value prop.
That is not the part I take issue with at all! In fact many (if not all) of my comments to those people are thanking them.
I think you need to go and double check exactly what I have replied! Feel free to link to anywhere you think I have been irate in response to someone stating their preference! I think you will find my only grievance is where people have called the INPUT thoughtless or zero effort.
I’m finding it fascinating how people are commenting on what they think they have read rather than what they have actually read - both in relation to the post and in my replies.
I do appreciate your advice on following a writing course. I did exactly that to great effect in the early 2000s. Unfortunately it is of little help these days
AI writing is rapidly improving. Human writing, on average not so much.
https://www.promarket.org/2026/05/04/consumers-prefer-ai-mus...
>If you’re not bothered enough to review and edit it...
>...then I’m not going to bother reading it.
Just playing Devil's Advocate here, but we've had copywriters and editors for generations now. Not reading something because AI did the post-writing work feels kinda weak. Skip stuff started from AI, fine, but this is just because it's AI, not because the author didn't do the task. And I do get the point is that if an AI did /most/ of the work, ok to that, too, but not post-write processing alone.
did you formerly hire copywriters for this task?
So in the end, I don't care if a human wrote it or an AI, I don't want to read it either way.
What's wrong with copyrighters? They write like corporations, not like humans (see the Cluetrain Manifesto for more on this). They sound like nobody - literally inhuman. I don't want to read that. I want to hear a human voice. If I can't hear that in your writing, I don't need to read it.
As an anecdote, someone posted a blogpost on Linkedin on using agents to implement a driver to access PCIe devices over TCP/IP. I was intrigued because that's not an easy task for several reasons, like handling PCIe interrupts and DMA. For exmaple, how does the remote machine map the device's PCIe BARs? And when it issues I/O to the devices registers, how are these reads and writes transferred to the remote device. In the end, this is just some virtual memory. In a local machine, this is either directly mapped to the PCIe physical addresses or some IOMMU virtual address space which is then translated by the hardware upon CPU/device/VM access.
After reading the long verbose promising article, in the end, the guy (with the help of the agent) only managed to implement access to the PCIe config space so that lspci on the remote machine works and shows the remote PCIe device, but that's all. It never addressed the issues above nor even mentioned them. The code was AI generated. The article was AI-written. The article never made a reference to DMA, interrupts, MSIX-X, IOMMU, IOTLB, virtual memory, etc, but it made big claims on next-gen datacenter disaggregated architecture, boosting GPU utilization, reducing large scale inference costs, etc.
Anyway, you get my point: big long beautiful words, but zero nuance.
The implication is that expertise is being devalued in real time, and in a sense it really is, but I think that most people really do need to face the reality of their emotions and where it comes from. For myself personally, I have no qualms about reading or even engaging with AI content, as long as the content itself is of quality and truthful.
I described a problem to a new hire. Symptoms, how to reproduce, etc. and asked him to go investigate it, pin down the problem and solve it. It was a non-trivial and hard to reproduce bug.
Instead of doing any of that, he sent back very quickly a wall of text. Generic advice. Nothing relevant. Along with this he sent a PR that would have broken prod.
It was useless, and basically amounted to throwing the assignment back in my face. Since I, like literally everyone else, already have Claude Code, having an expensive person use it for me (slowly and badly) is worthless. In fact, it's negative value, and a huge waste of time. ChatGPT would have given me the same wall of text for pennies.
In the end I got rid of him.
I'm sure he thought I was "intimidated", and had "ego" issues. But he's actually just redundant.
Is that an ego issue? Doesn't seem like it to me.
The bar for what counts as expertise is shifting. If someone with relatively little background can use an LLM to get to a reasonably informed answer much faster than before, then experience and credentials alone carry a bit less signalling power than they used to. That obviously doesn't mean the inexperienced person suddenly has the judgement, context or intuition of an actual expert, but I do think some of the hostility toward AI comes from discomfort with that change.
AI is still wrong, goes in weird circular loops, and doesn't solve the fundamental human problems around judgement, organization, learning curves, accountability, etc. But it does seem to be raising the baseline expectation for what a competent knowledge worker should produce, and I feel that shift is genuinely uncomfortable for a lot of people who based their identities around being rigorous abstract problem solvers.
I agree that it shifts the notion of what is expertise, but for the worst. It's very hard to tell who is actually an expert, and who is mimicking the expert.
The consequences vary from annoyance to downright catastrophe. e.g.: a confidently wrong medical advice
On the other side, sometimes the expertise is not a requirement for the user: "If it works and does what it says on the box, that's all I care".
I like the fact that you use the word "some", awesome to see some nuance. It's a tricky situation.
---
[1] I am being an idiot on purpose, I think the word skepticism was completely derailed from its origin and meaning.
A minimum amount of skepticism is very healthy, especially for software engineering. Most software engineers, and science-based professions for that matter, are skeptics at heart. And it's a good thing.
Skepticism is the base of science; science thrived when we admitted ignorance and became skeptic towards our own knowledge.
Plus I’ve been seeing where AI productivity is highest and it’s been when leveraged by an expert.