Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Don't paste the AI, please

dontpastetheai.com

Post reply on HN
Print view View original

581–590 of 632 posts

Re: Don't paste the AI, please

#581

Post by NateEag »

mapontosevenths wrote: post #578
NateEag wrote: post #351

Earlier quoted context omitted.

I am literate and hyperlexic - I read ~500 words a minute, and have read thousands of books for fun. I hate genAI deeply, for many reasons, but one of them is the way it bastardizes and ruins the written word. It writes incoherent, worthless prose, taking hundreds of words where two sentences would do. The signal-to-noise ratio in genAI writing is just wretched. Yes, I can read the thousand-word ball of muck you past…

> It writes incoherent, worthless prose, taking hundreds of words where two sentences would do. Either you haven't used GenAI in a few years, you've only used bad ones, you're really bad at using them, or all of the above. > I hate genAI deeply, for many reasons Probably explains the above issues. In my experience people who don't want things to be good: a) Always find them to be bad b) Are always bad at using them

> Either you haven't used GenAI in a few years, you've only used bad ones, you're really bad at using them, or all of the above.

I use Claude every day, by corporate mandate, generally with Opus 4.8 or 5, but sometimes Sonnet 5 (per corporate demand to both use LLMs heavily and simultaneously keep the bills within a budget).

I hate doing it, but I get results my coworkers, PMs, and end users are happy with.

> Probably explains the above issues. In my experience people who don't want things to be good...

I admit regularly that LLMs do amazing things. They're astonishingly good at breaking security, for instance.

That doesn't mean I like them.

And no, they really are not good at writing.

When you feel the impulse to be defensive, that's often a sign that asking questions will be more worthwhile than leaping into the defensive posture.

Re: Don't paste the AI, please

#582

Post by mapontosevenths »

NateEag wrote: post #581
mapontosevenths wrote: post #578

Earlier quoted context omitted.

> It writes incoherent, worthless prose, taking hundreds of words where two sentences would do. Either you haven't used GenAI in a few years, you've only used bad ones, you're really bad at using them, or all of the above. > I hate genAI deeply, for many reasons Probably explains the above issues. In my experience people who don't want things to be good: a) Always find them to be bad b) Are always bad at using them

> Either you haven't used GenAI in a few years, you've only used bad ones, you're really bad at using them, or all of the above. I use Claude every day, by corporate mandate, generally with Opus 4.8 or 5, but sometimes Sonnet 5 (per corporate demand to both use LLMs heavily and simultaneously keep the bills within a budget). I hate doing it, but I get results my coworkers, PMs, and end users are happy with. > Probabl…

Claude might be the problem here. It really is bad at writing and ticks all the boxes you mentioned. It's much better at code.

I don't mean to be defensive, just to point out that most of the statements you made are provably and obviously untrue... And given that you admit bias that's the most likely explanation.

I hadn't considered that you have experience, but only with one that is legitimately bad at this particular thing because someone forced it onto you. Fair enough.

We don't all have to like the same things.

Re: Don't paste the AI, please

#583

Post by disillusioned »

mmikeff wrote: post #90
disillusioned wrote: post #17

Heh. Just got done writing (by hand!) a Principles of AI Use document for my (ironically) AI enablement firm, the first of which is: Write as yourself. You’re being paid for your expertise and insights. Communicate them directly to us. Copying and pasting Claude responses into Slack or an email directly shifts the burden of comprehension and understanding to everyone else, and worse, risks skipping that step for your…

Any chance you would publish that guide and share here?

Here are the rest. I wrote these to specifically be principles that can be easily internalized and pointed to for accountability:

2. Own the Output. Everyone is encouraged to use these tools to accelerate the pace of their work and to take the toil out of their daily output. But that doesn’t mean you simply defer to the model and skip your own careful review step. Part of the paradox of these models is that they’re wordy and it can feel exhausting to read through everything they’re writing and confirm its validity. But that’s an enormous part of the job: you’re the human-in-the-loop to make sure what it’s saying is correct, and that you’ve reviewed it before you’ve shared it with the rest of the class. Read all of the words before they reach the rest of your team or a client, and make sure you stand behind having your name attached to them. In situations where a full review isn’t possible for expediency, disclose it directly! “Hey, I haven’t had a chance to review, this is pure Claude, but here’s what it says.” This isn’t a get out of jail free card, but it can be used occasionally when the situation dictates. Most of the time, you need to review and adjust the output before other people see it.

3. Be the expert. Have an opinion. This is a big one. Do not offload your role as expert to the model. The model is an assistant. You need to internalize the work being asked of you, and apply your education, instincts, and specific expertise to form your own opinion. You can use models to work to adversarially challenge things, but part of being an expert is knowing when to discard irrelevant feedback. Use this skill daily. Don’t let it atrophy.

4. Don’t Slop-Bomb. This is an amalgamation of all of the above, but again, don’t just spew large volumes of model slop into comms channels. Bring your own perspective, internalize the problem in front of you. And be judicious about what you share, and how wordy these things are. Put another way: edit for clarity and concision. It’s too easy to generate busy-work-seeming reports and outputs, and that burden shift to everyone else is multiplicative. Use fewer words, distill things into insights (backed with longer form deep dives), but make sure you’re not just making the rest of the team read a bunch of useless pablum. I don’t want to have to make my model read your model’s output to glean what I’m supposed to out of it. Do that for your coworkers and your clients. Think in terms of information hierarchy.

Re: Don't paste the AI, please

#584

Post by disillusioned »

gh5000 wrote: post #359
disillusioned wrote: post #17

Heh. Just got done writing (by hand!) a Principles of AI Use document for my (ironically) AI enablement firm, the first of which is: Write as yourself. You’re being paid for your expertise and insights. Communicate them directly to us. Copying and pasting Claude responses into Slack or an email directly shifts the burden of comprehension and understanding to everyone else, and worse, risks skipping that step for your…

I’m curious if you’d share there rest of your principles. This is so well articulated.

Thanks for the compliment! I just posted the other 3 in another reply!

Re: Don't paste the AI, please

#585

Post by DenisM »

prepend wrote: post #573
DenisM wrote: post #560

Earlier quoted context omitted.

You missed my point. As a possible reader of your email I’m facing dozens of emails every day, my default reaction is to skip past and “read it later”. Maybe you did put a lot thought into it, how would I ever know? If you show up in person I will know you are invested into it.

Perhaps I missed your point. I’m kind of lucky to be in the position where if you don’t read my email you will regret it because something bad happens. Or at best you miss out on something cool that would help you. I expect everyone I send my email to to read it and pay attention to it, or I don’t send it to them. Also, I’d love to have only dozens of email. I expect me and most of my coworkers go through 50-200 emai…

Fair point. I’m in a position of “creating horizontal influence” and I’m lucky enough to be able to focus on less than a dozen outcomes per quarter, so that colors my perception.

Re: Don't paste the AI, please

#586

Post by anotherevan »

blensor wrote: post #244
mscbuck wrote: post #48

Earlier quoted context omitted.

Just want to say, "Copying and pasting Claude responses into Slack or an email directly shifts the burden of comprehension and understanding to everyone else" is something that I don't think I really internalized until you wrote it in that way. Which I think is evidence of what you are talking about.

Tangentially I also hate if someone answers with a voice message. I can understand that there are situations where you just can't type out the answer and those exceptions are of course fine, but if you post a voice message or god forbid a video message in a group chat you are forcing everyone to take the time and sit through it.

You just gave me strong memories of using voice messages for a couple of months some years ago. In my defence, I had dropped a circular saw blade through the tip of my finger which slowed down my typing for a while. (I did manage to get surprisingly fast as a nine-finger touch typist, sans the left index finger.)

Fortunately my colleagues were very understanding, and I reverted back to text once I could type reasonably well and with minimal pain again.

Re: Don't paste the AI, please

#587

Post by anotherevan »

bonoboTP wrote: post #246
knorker wrote: post #179

Earlier quoted context omitted.

This reminds me how I've had to tell people to try to type better and nt us wird abbrvs, even if it means they can type it faster. Because if you want 100 people to read it, then every second you waste for the reader is worth you fixing up your message for over a minute. It's very likely that proper punctuation and capital letters help the reader, then it doesn't matter than not using them is faster to type. I'm sure…

That's a good use for AI. Type loosely and ask it to fix up the spelling, capitalization and punctuation.

Except that it will get it wrong infrequently but often enough that you should really proof-read the result properly. All of which you end up taking more time than if you wrote proper if the first case.

Re: Don't paste the AI, please

#588

Post by Forgeties79 »

simonra wrote: post #509
bluGill wrote: post #406

Earlier quoted context omitted.

In many cases someone else already knows the answer and your interruption of them is less time lost for the company than you looking up the answer.

Unless the organisation is on the brink of collapse there is valuable learning and understanding in honestly attempting the research. If nothing else one gets a bit more knowledge on how to research the topic, although some knowledge of the state of the field is hard to avoid if you're so new to it that it's resource intensive to research. For the company it would probably pay off better (over time) to let the people…

I’m sorry dude but if it’s going to take me two hours of sorting through documentation to find a feature in the software I’m using and I know someone one Slack message away could point me to it in 30 seconds, I’m going to ask them. You’re basically making a case against any sort of technical collaboration outside of one where everyone comes to the table with all of their knowledge set and prepared to look up everything themselves. That is very inefficient and simply not how people operate, not even you do that

Re: Don't paste the AI, please

#589

Post by rpdillon »

Forgeties79 wrote: post #588
simonra wrote: post #509

Earlier quoted context omitted.

Unless the organisation is on the brink of collapse there is valuable learning and understanding in honestly attempting the research. If nothing else one gets a bit more knowledge on how to research the topic, although some knowledge of the state of the field is hard to avoid if you're so new to it that it's resource intensive to research. For the company it would probably pay off better (over time) to let the people…

I’m sorry dude but if it’s going to take me two hours of sorting through documentation to find a feature in the software I’m using and I know someone one Slack message away could point me to it in 30 seconds, I’m going to ask them. You’re basically making a case against any sort of technical collaboration outside of one where everyone comes to the table with all of their knowledge set and prepared to look up everythi…

No, it's more like a basic case where you put five minutes into it and if you can't get there in that time frame, then you ask someone. But I am asking for the five minutes of work.

Re: Don't paste the AI, please

#590

Post by chrisweekly »

nativeit wrote: post #490
jimmaswell wrote: post #469

Earlier quoted context omitted.

I wish this were simply a fireable offense and we were expected to read. These kind of people who act like reading 3 paragraphs is being asked to swallow glass make me blindingly mad. They should be unemployed.

I’m usually very tolerant of others’ varying capacities, but 100% this. High-school level reading and comprehension shouldn’t be a hard ask for professionals.

Sending a wall of text in an email is, IME, ~never^1 the right thing to do. If more than a fraction of colleagues fall into that bad habit, soon all anyone will be doing is email. Let alone if you try to respond to one of the too many points in an overlong message, or loop someone else into the conversation(s) which are now hopelessly fragmented. It's akin to a self-inflicted DoS attack.

1. Exceptions do exist (eg if the message is intended solely for and addressed to a single recipient, and includes a disclaimer / request / reminder about general email hygiene), but they're vanishingly rare.

Post reply on HN