AI for writing documentation

[zester]

zester buddy DM me, tell your company to come to the dark side, 25M a year is bullshit honestly that's maybe 1 rack of our hardware but x5-7 years that's a decent cluster (125-175M) ...

Your company might have enough compute to build their own model when not using for inference.

We don’t really get to choose who we work with because we operate under government contracts and must follow DoD regulations. To work with us, you would need to submit a bid and meet specific requirements the contract typically goes to whoever satisfies the most criteria.

For example, if more than 26% of your company’s staff has a significant disability, you may receive priority consideration under the AbilityOne program. However, that would only apply if we were accepting new contracts, and I don’t believe we are at the moment.

 

[loveydovey]

We don’t really get to choose who we work with because we operate under government contracts

Speaking of the elephant in the room. Some dubious entities that infiltrated the US government want to turn all of you into computers.

About 6 years ago I saw a DARPA youtube video that was doing a presentation about a brain-computer interface (BCI) for humans. And this DARPA person just casually had a slide that displayed a continuum from Human to Machine. So, basically, 95% machine and 5% human - still human, folks! That's how crazy some stuff is that's getting pushed forward.

And you know they've done this before. Humans tend to swallow even highly detrimental stuff if it's done via incremental change. So they are redefining what it means to be human with zero pushback. And the reality is that even 1%machine/99% human is already not human.

But that's the goal. It's not AI. Not robotics. It's BCI. Highly invasive BCI. It's even more. It's fully merging with the machine with eventually humans going full silicon and synthetic. Life as hopping between servers, teleporting minds to distant machines on Earth or interplanetary, eternal life.

But you know when something is forced upon you without full disclosure and significant discussion, it is not in the best interests of you, your family, friends, and community.

 

[zester]

Speaking of the elephant in the room. Some dubious entities that infiltrated the US government want to turn all of you into computers.

About 6 years ago I saw a DARPA youtube video that was doing a presentation about a brain-computer interface (BCI) for humans. And this DARPA person just casually had a slide that displayed a continuum from Human to Machine. So, basically, 95% machine and 5% human - still human, folks! That's how crazy some stuff is that's getting pushed forward.

And you know they've done this before. Humans tend to swallow even highly detrimental stuff if it's done via incremental change. So they are redefining what it means to be human with zero pushback. And the reality is that even 1%machine/99% human is already not human.

But that's the goal. It's not AI. Not robotics. It's BCI. Highly invasive BCI. It's even more. It's fully merging with the machine with eventually humans going full silicon and synthetic. Life as hopping between servers, teleporting minds to distant machines on Earth or interplanetary, eternal life.

But you know when something is forced upon you without full disclosure and significant discussion, it is not in the best interests of you, your family, friends, and community.

I’m going to put you on ignore bro. The level of paranoia in your messages is setting off a lot of red flags for me.

 

[astyle]

We don’t really get to choose who we work with because we operate under government contracts and must follow DoD regulations. To work with us, you would need to submit a bid and meet specific requirements the contract typically goes to whoever satisfies the most criteria.

For example, if more than 26% of your company’s staff has a significant disability, you may receive priority consideration under the AbilityOne program. However, that would only apply if we were accepting new contracts, and I don’t believe we are at the moment.

That's exactly what one calls 'manipulation of gutless and uninformed populace'. Just trying to hide behind excuses to say 'no' to an application. The longer the logic chain that leads to a 'no', the better. Can GenAI generate a conclusion like that?

 

[loveydovey]

I’m going to put you on ignore bro. The level of paranoia in your messages is setting off a lot of red flags for me.

Whyyyyy. There's nothing inherently wrong with uploading people's minds to a computer and teleporting people's virtual representations to other places at the speed of light. I mean, one day some people's minds will get sent into a virtual dumpster or a shredder, but relax, it'll be just like adoption of new technologies like washing machine and dishwashers and calculators. It's really the same thing. Let's hope we wake up on a server one day, having zero physical bodies, because it'll be good for us, good for human civilization, don't you get it.

 

[LibreQuest]

Whyyyyy. There's nothing inherently wrong with uploading people's minds to a computer and teleporting people's virtual representations to other places at the speed of light. I mean, one day some people's minds will get sent into a virtual dumpster or a shredder, but relax, it'll be just like adoption of new technologies like washing machine and dishwashers and calculators. It's really the same thing. Let's hope we wake up on a server one day, having zero physical bodies, because it'll be good for us, good for human civilization, don't you get it.

X Files Season 11 Episode 2 is about this topic. But what they don't talk about in that episode is writing documentation, sadly.

 

[astyle]

I mean, one day some people's minds will get sent into a virtual dumpster or a shredder,

I think that already happened. Just look at Discord, Reddit, and other public forums. Actually, even Zuckerberg once said that Facebook really exposes how stupid people really are - and that was nearly a quarter-century ago when he said that.

X Files Season 11 Episode 2 is about this topic.

Ghost In The Shell, the entire anime/manga collection revolves around that topic.

 

[astyle]

Okay.... Here comes the

https://www.businessinsider.com/claude-chat-transcripts-lawsuit-privileged-ruling-2026-2

I don't mess with 3rd party AI Labs.

WE DO LOCAL LLM!!!
WE RUN HPC | High Performing Computing !!!
WE RUN ON-PREM!!!
WE RUN OPEN SOURCE!!!


View attachment 25412


We are the future... Every Enterprise will come around...

Good thing I'm too lazy to be constantly logged into Claude/ChatGPT. I last logged into Claude about a year or so ago, it knows my real info, although by now it's kinda out of date, and there's nothing politically incorrect in my Claude transcripts. I did teach ChatGPT what 'rhino-phyto-necro-philia' is (a fancy combination of partial Latin words, that means sommelier). I don't care who sees THAT. But yeah, the old adage of 'Think before you post' still applies, whether it's AI chatbot transcripts or old-fashioned forum posts, or tweets.

 

[loveydovey]

I think that already happened. Just look at Discord, Reddit, and other public forums. Actually, even Zuckerberg once said that Facebook really exposes how stupid people really are - and that was nearly a quarter-century ago when he said that.

Again, there's this giant chasm between invasive stuff that bypasses/modifies sensory input, the actual biological circuitry, and stimuli that are processed without disrupting normal biology.

And sadly, no one is truly guarding even the normal biological processing, it seems. Because all it takes is a small modification to humans to incapacitate them to be able to say no to things that no humans in their right minds would ever agree to. Humans can be also be rendered incapable of understanding how they are being screwed with as they agree to self-destructive agenda.

Enough of doom and gloom, I guess. Look how great things are - we have Elon Musk who had the balls to demonstrate to the entire world that bidirectional communication with neurons is very easy and possible with good fidelity (remember that demo presentation in a pig barn? 1984 anyone?). BI-directional - meaning, not just read but also feed into the brains.

 

[hruodr]

But, thanks for not choosing kindness (not a lot of that going around here).

You misunderstood me. Of course I do appreciate your program and I made a bookmark to it for the case I need it. Thanks for it.

But as far as the point about AI and/or interns writing documentation:
Isn't technical writing a class a CS has to take?
Do you not write your specifications before you code (who writes code before they know what they want)?

That is what I wanted to say: it is not my workflow. No big plan before beginning coding, but yes, making notes on the way is the right thing. And I never took such CS class, I am not computer scientist.

BTW, I think the term is roff, manual are macros for it, but I learned a little of troff and kept using plain TeX.

 

[drhowarddrfine]

Maybe related by my good friend Jeff Geerling who lives 10 minutes from me and I've never met:

AI is destroying Open Source, and it's not even good yet

Over the weekend Ars Technica retracted an article because the AI a writer used hallucinated quotes from an open source library maintainer. The irony here is the maintainer in question, Scott Shambaugh, was harassed by someone's AI agent over not merging it's AI slop code. It's likely the bot...

www.jeffgeerling.com

Actually, I live closer to Wayne Gretzky but now I'm really getting off topic.

 

[JohnK]

That is what I wanted to say: it is not my workflow. No big plan before beginning coding, but yes, making notes on the way is the right thing. And I never took such CS class, I am not computer scientist.

BTW, I think the term is roff, manual are macros for it, but I learned a little of troff and kept using plain TeX.

Nor am I (different field).

Whatever, use/think what you want but BSD adopted mandoc/mdoc years ago and does not guarantee anything written in other markup languages to render properly (it's your manpage). I've tried to give you (and other readers of this thread) some advice about what I've seen and the potential problems (-i.e. mandoc is not just about rendering manpages, it's about being able to convert to other formats--and searching--and my tool now extends a small portion of that to a files written in a simple markdown syntax). takeit/leaveit.

Maybe related by my good friend Jeff Geerling who lives 10 minutes from me and I've never met:

vote: is related.

I've had a private repository for years but lately I started to push some projects to github which I think could be a help to others; I envisioned someone saying: "nice. what if we did this..." (you know, normal conversation/interaction) but instead, people seem so closed off and rude. An agent bullying (either directed to or not) sort of brings things to a new--scary--level. Makes me want to go ostrich.

 

[astyle]

And sadly, no one is truly guarding even the normal biological processing, it seems. Because all it takes is a small modification to humans to incapacitate them to be able to say no to things that no humans in their right minds would ever agree to. Humans can be also be rendered incapable of understanding how they are being screwed with as they agree to self-destructive agenda.

If you think about it, alcohol has been performing that role since times immemorial. A bit of alcohol is all it takes for humans to start acting irrationally, and agree to stuff they never would when sober. This is why friends don't let friends code when drunk. This is why I have to ask, "What was the programmer on when they did all that crappy coding? Alcohol or something way more potent?"

 

[loveydovey]

astyle Yes, but what I meant was without your awareness.

 

[SirDice]

And you know they've done this before. Humans tend to swallow even highly detrimental stuff if it's done via incremental change. So they are redefining what it means to be human with zero pushback. And the reality is that even 1%machine/99% human is already not human.

Who's this illusive "they"?

 

[astyle]

astyle Yes, but what I meant was without your awareness.

It's actually amazing how much alcohol one can consume without being aware - like in a sweet, easy-to-drink cocktail . Or even in a screwdriver - two women in my my family were having a long conversation, making screwdrivers as they went - by end of the day, they realized that a whole bottle of vodka has gone into those screwdrivers without them noticing a thing! And neither of them is an alcoholic. Just plain did not notice a thing.

And the reality is that even 1%machine/99% human is already not human.

Does that include medical implants like heart/hips/replacement limbs?

 

[loveydovey]

making screwdrivers as they went - by end of the day, they realized that a whole bottle of vodka has gone into those screwdrivers without them noticing a thing! And neither of them is an alcoholic. Just plain did not notice a thing.

What's worse is if you drink alcohol with caffeine (rum and coke, i.e.). Caffeine will keep you wired and awake as you intoxicate, and then you abruptly crash into a delirium. There's zero heads-up, you feel sober and then you are gone. So I hear at least.

Does that include medical implants like heart/hips/replacement limbs?

Are amputees less human?

 

[hruodr]

Too much 'philosophy', few technique, in spite that in spite of all AI hype the results are impressive.

 

[JohnK]

I ran a test: I took a script/program I have (which I also have documentation I created) and asked AI to generate documentation if anyone is interested in seeing/comparing the two documents.
Note: Full credit; this is the same test atax1a suggested (more "in the spirit of", to be accurate, because I'm not sure I would have anything approaching 'complex' per the original proposal).

If interested, please give a thumbs up or say so. I could create another thread or whatever (admin/mod can overrule).

 

[hruodr]

If interested, please give a thumbs up or say so. I could create another thread or whatever (admin/mod can overrule).

You can show it here. But the idea is to use it as assistant, not put there own info, not put and get.

 

[astyle]

I ran a test: I took a script/program I have (which I also have documentation I created) and asked AI to generate documentation if anyone is interested in seeing/comparing the two documents.
Note: Full credit; this is the same test atax1a suggested (more "in the spirit of", to be accurate, because I'm not sure I would have anything approaching 'complex' per the original proposal).

If interested, please give a thumbs up or say so. I could create another thread or whatever (admin/mod can overrule).

Can the AI distinguish between American English, British English, and Aussie English? in Perl haikus, a human can detect sarcasm. Hell, a human can probably detect sarcasm just by figuring out the language of choice. Like writing a subroutine in C#.NET just to stick it to the boss/lawyer who has a marked preference for the elegance of Python 2.7...or use colorful language to name all the variables, routines, etc. Like the Smartypants port (textproc/py-smartypants). If you don't know that Smartypants even exists, the AI-generated documentation may be correct, but look like complete nonsense.

 

[JohnK]

Actually, it's really confusing. As a follow up I just asked it to compare mine with the one it produced and it kept saying mine was using groff-style macros and I'm in the middle of trying to figure out why it's saying that. It's also saying that mine is more "friendly" and "rich" and its version is more "cleaner" (is that just fancy speak for "sparce" or "void")?

Nonetheless, I have the AI output saved and I will post it. I will post a link to mine (to keep me honest but the one I posted to the AI just now has minor fixes).

 

[JohnK]

Can the AI distinguish between American English, British English, and Aussie English?

ha! I have friends all over those other (non American English) places and I can't understand most of what those weirdos say (call me). I think most of them are just poking fun at me behind my back.

 

[JohnK]

You can show it here. But the idea is to use it as assistant, not put there own info, not put and get.

No. This is what I was getting at with my 'semantic markup' and 'searching' comments. -e.g. The previous AI produced docs, I've seen, have had ".Cm" (command modifier macro) sloshed around the output to produce "color" or "highlights" where *I* would not have put one. This IMO causes a potential problem (with searching or 'semantics'). Besides, I'm also testing the validity of my work (delete or don't delete my repo). -i.e. would people prefer to use an AI or have a manpage built like/with make. or do designers need to keep comments to themselves when typing technical documents.

 

[astyle]

Actually, it's really confusing. As a follow up I just asked it to compare mine with the one it produced and it kept saying mine was using groff-style macros and I'm in the middle of trying to figure out why it's saying that. It's also saying that mine is more "friendly" and "rich" and its version is more "cleaner" (is that just fancy speak for "sparce" or "void")?

Nonetheless, I have the AI output saved and I will post it. I will post a link to mine (to keep me honest but the one I posted to the AI just now has minor fixes).

yeah, this is what English classes are for. Learning the basics of spelling, sentence structure, choice of vocabulary, considering the tone, the audience, technical standards for formatting... and practicing your ass off so that those skills are no-brainers that don't require a lot of effort. I guess we're going back to the age of illiteracy and treating inanimate objects (i.e., computers) like gods/demons.