> Suspicions aroused, I clicked on the “Document History” button in the top right and saw a clean history of empty document – and then wham – fully-formed plan, as if it had just spilled out of someone’s brain, straight onto the screen, ready to share.
This isn't always a great indicator.
I can't stand Google Docs as an interface to write with, so use VIM and the copy/paste the completed document into it.
I don’t really find it better when someone ads a disclaimer. What am I supposed to do then? There’s still an expected default behavior of reading it, and if I don’t I need to confront them and say “I don’t care what you got an LLM to say, why not give me your view”. It’s inappropriate under any circumstances imo
Why can't the plan be judged on its merits? Rigorous verification of the idea is a good thing that should happen anyways. The main potential problem I see is transmission of privileged information to a third party.
I assume they are working at a business to make money, not a school or a writing competition.
For example suppose that someone likes to work in Markdown using VSCode. To get the kind of Word document that everyone else expects, you just copy and paste into Word. AI isn't involved, but it will look exactly like AI to you.
And there are more complicated hybrids. For example my wife has a workflow where everything that she does, communications, and so on, wind up in Markdown in Obsidian. She adds information about who was at the meeting that includes basuc research into them done by an agent (company directory, title, LinkedIn, and so on - all good to know for someone working in sales). Her AI assistant then extracts out bullet points, cross references, and so on. She uses that to create summaries that she references whenever she goes back to that project. And if someone wants to know what has happened or is currently planned for that project, AI extracts that from the same repository.
There's lots of AI in this workflow. But the content and thought is mostly from her. (With facts from searches that an agent did.) The fact that she's automated a lot of her organizational scutwork to an AI doesn't make the output "AI slop".
> If you know in your heart of hearts that you didn’t put the work in, you’re undermining the social contract between you and your reader.
There's been a lot of social contract undermining lately. Does anyone please know about something that can be done to try and revert back? Social contract of "F you. I got mine" isn't very appealing to me, but that seems to be the current approach.
If I discover you fed me AI output, directly from AI, it really makes me wonder what you are doing here. What did you add to this equation when I could have done it myself?
At least a "Generated by AI, reviewed and edited by xyz" tag would be some indicator of effort and accountability.
It may not be wrong to use AI to generate things whole cloth, but it definitely sidesteps something important and calls into question the "prompter's" contributions to the whole thing.
When it comes to LLMs, the only thing I hate more than the "I don't know, the AI wrote it" people is the "I wrote this" crowd. No you didn't, you asked someone else to write it. If you couldn't claim copyright for it in an IP court, you did not write it. Period.
>Regardless of their intent I realised something subtle had happened. Any time saved by (their) AI prompting gets consumed by verification overhead, the work just gets passed along to someone else – in this case me.
This is _exactly_ how I feel. Any time saved by precooking a "plan" (typically halfbaked ideas) with AI isn't really time saved, it is a transfer of work from the planner to whoever is going to implement the plan.
It was funny. On a more serious note, if one works in a sphere where expanding with AI makes "good enough" documents, then I have bad news for him - the sphere has too much redundancy in the first place (the same place that was used for training). So no new information is created in millions of documents made by humans, and this was noticed by the training pattern recognition. You cannot do the same with historical texts; unless we live in a simulation with predictable random generators, the events are random, and there are no rules like "If the king's name starts with a G, he will likely die in the first week of October."
> So it’s definitely AI. I felt betrayed and a little foolish. But why?
Because the prompter is basically gaslighting reviewers into doing work for them. They put their marks of authorship on the AI slop when they've barely looked at it at all which convinces the reviewer to look. When the comments come back, they pump the feedback into the LLM, more slop falls out and around we go again. The prompter isn't really doing work at all—the reviewers are.
The content of the document matters too. I don't really care if someone was AI-assisted writing a project plan. As long as it's sane and clear I'm not gonna lose sleep over that. However for my performance review I definitely want my manager to put in the effort and actually tell me nuanced thoughts on my performance. I don't want AI output for that part.
I know I'm an outlier on HN, but I really don't care if AI was used to write something I'm reading. I just care whether or not the ideas are good and clear. And if we're talking about work output 99% of what people were putting out before AI wasn't particularly good. And in my genuine experience AI's output is better than things people I worked with would spend hours and days on.
I feel like more time is wasted trying to catch your coworkers using AI vs just engaging with the plan. If it's a bad plan say that and make sure your coworker is held accountable for presenting a bad plan. But it shouldn't matter if he gave 5 bullets to Chat gpt that expanded it to a full page with a detailed plan.
> But if you ship it and people use it, you’ve created an implicit promise: that you can maintain, debug, and extend what you’ve built. If AI assembled it and you can’t answer basic questions about how it works, you’ve misled users about what they can depend on.
Agree with the premise but this part is off. When I find a project online, I assume it will be abandoned within a year unless I see evidence of a substantive team and/or prior long-term time investments.
I will admit to being an LLM workslopper. I don't ever send anything written by an LLM (because anyone who's seen enough LLM writing will recognize it's an LLM) without rewriting it by hand first - with exceptions for parts of READMEs - but for any other task it's pretty much 100% LLM.
I look at the output and ask it to re-re-verify its results, but at the end of the day the LLM is doing the work and I am handing that off to others.
I'm sometimes asked to produce meaningless 30-page documents that nobody ever reads. I mean literally nobody, since I can see the history of who has accessed it. Me and a proof-reader, and occasionally someone will open it up to check that it exists. But nobody reads them, let alone reads them closely. Not the distant funder who added it as a line-item requirement to their grant (their job is adding line items to grants, not reading documents), nor the actual people involved in the project, who don't have time to read a meaningless document, and don't need to. It's of use to no one, it's just something that must be done because we live in a stupid world.
I've started having AI write those documents. Each one used to take me a full week to produce, now it's maybe one day, including editing. I don't feel bad about it. I'm ecstatic about it, actually; this shouldn't be part of my job, so reducing its footprint in my life is a blessing. Someday, someone will realize that such documents do not need to exist in the first place, but that's not the world we live in right now, and I can't change it. I'm just glad AI exists for this kind of pointless yeoman's work.
I think it quickly needs to become good manners to indicate when text was written by AI rather than a person. I read that text differently and I shouldn't have to spend my time guessing.
It will probably be unpopular here, where people appear to have drawn the lines and formed unyielding positions, but...
The whole llm paranoia is devolving into hysteria. Lots of finger pointing without proof, lots of shoddy evidence put forward and nuance missing points.
My stance is this: I don't really care whether someone used an llm or wrote it themselves. My observation is that in both cases people were mostly wrong and required strict reviews and verification, with the exception of those who did Great Work.
There are still people who do Great Work, and even when they use llms the output is exceptional.
So my job hasn't changed much, I'm just reading more emojis.
If you find yourself becoming irrationally upset by something that you're encountering that's largely outside of your control, consider going to therapy and not forming a borderline obsession with purity on something that has always been a bit slippery (creative originality ).
There is nothing worse than this feeling, like fantastic, now i have to go read through this slop with incredible care and minutia. I may as well not read the slop and go redo all the work/thought myself, it will be easier that way.
Just a hot take, but if you ask someone to complete a rote task that AI can do, you should not be surprised when they use AI to do it.
The author does not mention whether the generated project plan actually looked good or plausible. If it is, where is the harm? Just that the manager had their feelings hurt?
2. The author could simply have done the research, created the plan, and then gave an LLM the bullet list points of research and told it to "make this into a presentable plan". The author does the heavy work and actually does the creative work, and outsources the manual formatting to the LLM. My Wife speaks English as a second language, she much prefers telling an LLM what she is trying to say and to generate a business friendly email from this than writing it herself and letting in grammatical mistakes.
3. If I were to write a paper in my favorite text editor and then put it through pandoc to generate a word doc it would do the same thing.
I think that we should have revision control for intermediate stages - for code, documents, even paintings. So we can at least have some idea of provenance, how it's made.
Nowhere in here does it indicate that the generated plan was wrong or broken. I dont care if you use ai to write. I care if you write well. If the author trusted the other person, then it shouldn't matter. If the author didn't trust the other person, then they'd have to validate their output anyway. Granted the tech allows people I dont trust to generate a lot more bs, a lot faster. But i just reject and move on with my life in that case. I am no ai booster but a lot people are expressing distaste for tools when they should be expressing distaste for fools.
38 comments
[ 4.7 ms ] story [ 51.6 ms ] threadThis isn't always a great indicator.
I can't stand Google Docs as an interface to write with, so use VIM and the copy/paste the completed document into it.
This
When I receive a PR, of course it’s natural an AI is involved.
The mortal sin is the rubber stamp.
If they haven’t read their own PR, I only have so many warnings in me. And yes, it is highly visible.
I assume they are working at a business to make money, not a school or a writing competition.
For example suppose that someone likes to work in Markdown using VSCode. To get the kind of Word document that everyone else expects, you just copy and paste into Word. AI isn't involved, but it will look exactly like AI to you.
And there are more complicated hybrids. For example my wife has a workflow where everything that she does, communications, and so on, wind up in Markdown in Obsidian. She adds information about who was at the meeting that includes basuc research into them done by an agent (company directory, title, LinkedIn, and so on - all good to know for someone working in sales). Her AI assistant then extracts out bullet points, cross references, and so on. She uses that to create summaries that she references whenever she goes back to that project. And if someone wants to know what has happened or is currently planned for that project, AI extracts that from the same repository.
There's lots of AI in this workflow. But the content and thought is mostly from her. (With facts from searches that an agent did.) The fact that she's automated a lot of her organizational scutwork to an AI doesn't make the output "AI slop".
There's been a lot of social contract undermining lately. Does anyone please know about something that can be done to try and revert back? Social contract of "F you. I got mine" isn't very appealing to me, but that seems to be the current approach.
At least a "Generated by AI, reviewed and edited by xyz" tag would be some indicator of effort and accountability.
It may not be wrong to use AI to generate things whole cloth, but it definitely sidesteps something important and calls into question the "prompter's" contributions to the whole thing.
This is _exactly_ how I feel. Any time saved by precooking a "plan" (typically halfbaked ideas) with AI isn't really time saved, it is a transfer of work from the planner to whoever is going to implement the plan.
Later, at someone else's desk:
"Chat, summarize these 10 pages into 3 points."
Because the prompter is basically gaslighting reviewers into doing work for them. They put their marks of authorship on the AI slop when they've barely looked at it at all which convinces the reviewer to look. When the comments come back, they pump the feedback into the LLM, more slop falls out and around we go again. The prompter isn't really doing work at all—the reviewers are.
Each can be seen as using a tool to add false legitimacy. But ultimately they are just tools.
I feel like more time is wasted trying to catch your coworkers using AI vs just engaging with the plan. If it's a bad plan say that and make sure your coworker is held accountable for presenting a bad plan. But it shouldn't matter if he gave 5 bullets to Chat gpt that expanded it to a full page with a detailed plan.
Agree with the premise but this part is off. When I find a project online, I assume it will be abandoned within a year unless I see evidence of a substantive team and/or prior long-term time investments.
I look at the output and ask it to re-re-verify its results, but at the end of the day the LLM is doing the work and I am handing that off to others.
I've started having AI write those documents. Each one used to take me a full week to produce, now it's maybe one day, including editing. I don't feel bad about it. I'm ecstatic about it, actually; this shouldn't be part of my job, so reducing its footprint in my life is a blessing. Someday, someone will realize that such documents do not need to exist in the first place, but that's not the world we live in right now, and I can't change it. I'm just glad AI exists for this kind of pointless yeoman's work.
The whole llm paranoia is devolving into hysteria. Lots of finger pointing without proof, lots of shoddy evidence put forward and nuance missing points.
My stance is this: I don't really care whether someone used an llm or wrote it themselves. My observation is that in both cases people were mostly wrong and required strict reviews and verification, with the exception of those who did Great Work.
There are still people who do Great Work, and even when they use llms the output is exceptional.
So my job hasn't changed much, I'm just reading more emojis.
If you find yourself becoming irrationally upset by something that you're encountering that's largely outside of your control, consider going to therapy and not forming a borderline obsession with purity on something that has always been a bit slippery (creative originality ).
The author does not mention whether the generated project plan actually looked good or plausible. If it is, where is the harm? Just that the manager had their feelings hurt?
1. If the output is solid, does it matter?
2. The author could simply have done the research, created the plan, and then gave an LLM the bullet list points of research and told it to "make this into a presentable plan". The author does the heavy work and actually does the creative work, and outsources the manual formatting to the LLM. My Wife speaks English as a second language, she much prefers telling an LLM what she is trying to say and to generate a business friendly email from this than writing it herself and letting in grammatical mistakes.
3. If I were to write a paper in my favorite text editor and then put it through pandoc to generate a word doc it would do the same thing.
Until AI is used to fake that, too.