Live data from Hacker News

AI;DR (AI; Didn't Read)

rickmanelius.com

521–530 of 729 posts

Re: AI;DR (AI; Didn't Read)

#521

Earlier quoted context omitted.

Two very useful directives to give AI when it comes to documentation: 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 s…

> Document what's there, not the diff 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.

Most models are trained to be as "helpful" as possible which may work for a chatbot but not for code.

Re: AI;DR (AI; Didn't Read)

#522
post #516
post #263

Earlier quoted context omitted.

It's not just claude, all AI is unable to produce something concise. On the surface everything looks 'good' whether code or prose, but then if you dig a bit, try and understand the whole text you quickly realise that 80% of it is unecessary and the whole thing could have been re-worded/re-coded into something a fraction of its size and complexity. I asked Sol to reduce the length of some documentation we had by makin…

I have a theory that AI code looks good because you never subsequently come up with your own alternative.

"AI code looks good for you because you are a bad developer"?

Re: AI;DR (AI; Didn't Read)

#523

Earlier quoted context omitted.

It's not the ocean, it's the poster's own team. A simple "AI comments suck" in a sprint retro would be trivially easy and would at least start the conversation.

While I agree with the spirit of what you're saying, it is very similar to whenever anyone complains about things where they work. "Well, why don't you just raise it with management?" You're assuming that management 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…

I usually require to file an issue in bug tracker about problem in #3 and then follow all its updates to prevent silent closing by management. It won't help but annoys management a lot and they afraid to get more backlog issues of such sort from me in the future.

Re: AI;DR (AI; Didn't Read)

#524

Earlier quoted context omitted.

Two very useful directives to give AI when it comes to documentation: 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 s…

> Document what's there, not the diff 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.

> 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.

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.

Re: AI;DR (AI; Didn't Read)

#525
post #263

Earlier quoted context omitted.

My "favorite" Claudism is when I critique its work and ask it to remove some unnecessary part of the design -- and then the diff has more green than red because it added comments about why the code is no longer there -- the code that was never in the mainline and never asked for!

It's not just claude, all AI is unable to produce something concise. On the surface everything looks 'good' whether code or prose, but then if you dig a bit, try and understand the whole text you quickly realise that 80% of it is unecessary and the whole thing could have been re-worded/re-coded into something a fraction of its size and complexity. I asked Sol to reduce the length of some documentation we had by makin…

It's not that it fundamentally isn't able to produce something concise, it's that the business model of the companies developing these models rests on selling tokens...

Re: AI;DR (AI; Didn't Read)

#526

Earlier quoted context omitted.

Good instinct. Fair challenge. This corrects my framing. It makes your point sharp. This is a significant finding. Positively confirmed. That settles it, and it flips the picture. Honest verdict - this is not small. Let me wire it in.

There's got to be a Weird Al song like this coming soon.

[flagged]

Re: AI;DR (AI; Didn't Read)

#527

Earlier quoted context omitted.

Good instinct. Fair challenge. This corrects my framing. It makes your point sharp. This is a significant finding. Positively confirmed. That settles it, and it flips the picture. Honest verdict - this is not small. Let me wire it in.

There's got to be a Weird Al song like this coming soon.

Bless the lord that it's not been presented to us yet, because you just know that it exists already.

Re: AI;DR (AI; Didn't Read)

#528

Earlier quoted context omitted.

It's not the ocean, it's the poster's own team. A simple "AI comments suck" in a sprint retro would be trivially easy and would at least start the conversation.

While I agree with the spirit of what you're saying, it is very similar to whenever anyone complains about things where they work. "Well, why don't you just raise it with management?" You're assuming that management 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…

Seems to me if technical decisions on the level of "no more AI comments" go through management the organization doesn't know what they are doing.

To be fair, I would not be surprised.

Re: AI;DR (AI; Didn't Read)

#529
post #174

Earlier quoted context omitted.

Yes, the original prompt is 10000% better than the AI output, because it shows me exactly how much the author (you?) cares about this topic and how much effort they're willing to put into communication, which is barely any at all.

> which is barely any at all Clearly you either didn't read it, misunderstood it, disbelieved it, or just DGAF :)

Add some paragraphs and yes; your style is more "interesting than what an LLM generated. That's the only think truly hard to read compared to the output.

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.

Re: AI;DR (AI; Didn't Read)

#530
Whenever I'm tempted to base my judgment on how something's been created rather than what the end result is, I think about the scene where Sheldon throws away the french toast Penny made for him, despite smelling good, just because "Monday is oatmeal day."
Post reply on HN