AI for writing documentation

hruodr said:
Without understanding mdoc, those .Cm introduce text as bold face, have clearly a wanted effect in
the representation, perhaps their are though to be used in other logical context.
Click to expand...
hruodr said:
Or perhaps use older standards: mandoc -Tman txl.mdoc > txl.man
Click to expand...

My point may not have completely landed. So, this may sound like I'm patronizing you--I'm not--but stay with me a second please.

The equivalent being:
Code: 
mv /etc/rc.conf /etc/file
mv /etc/crontab /etc/file
mv /usr/local/etc/sshd...

then...
find --type f --name "file" /

A 'command modifier' is an argument to the command (sans dash) as far as I understand it. `mandoc(1)` brought the ability to search and thus why BSD now treats `mdoc(1)` as a "*semantic* markup language" (not a "formatting markup language"). The AI is doing you (and anyone that installs your manpage) a disservice by removing a portion of functionality. Meaning: if you want it bold, use bold (which means nothing in the 'semantic markup' scheme other than to make it stand out).

No please. Do not do that. The old "man" format is not guaranteed to render (as a manpage) properly but you have the right idea for my example above. If you take that 'semantic markup' and translate it to "another format" then you either get pretty colors (no 'semantics' involved) or gibberish--because now every file is called "file" for example.
 
JohnK said:
My point may not have completely landed.
Click to expand...
Of course I get your point. I am not an expert on roff, but I have a different sense of the importance of separating structure and rendering.

This is like the story of HTML and CSS. Originally HTML did not strongly distinguished between structure and rendering, later the separation became it stricter, HTML for structure, CSS for representation. I am till at the level
of HTML 1, I like simplicity.
 
drhowarddrfine said:
Thus the problem.

There is an article online that says the free AI tools are not what company programmers use. Those types pay for the AI tools they use and that means hundreds of dollars a month. Even then the generated code has to be gone over. He recommends, for you and me, buy the $20/ month AI and you'll see the difference. I intend to try it for the fun but I'm not having fun right now.
Click to expand...
Install OPENCODE. They have different Tier 1 model every month or so for FREE (some throttling is present tho). Now, I believe, it is the Big Pickle - anonymous, so probably Chinese beast of a model.
I gave it really lousy prompt, told it to set up the agents itself and to not bother me with questions. The process amazed me - it even created simulation of ethernet medium so it could test the program (sending files over UDP using one-directional gateway). Spat out complete code, documentation, how-tos, ASCII diagrams. Needs some tweaking but if I had to code it myself it would be a week at least, assuming I had miraculously learned python on professional level in my sleep.
 
Well, ChatGPT does sometimes generate useful stuff, even patches for Open Source projects. But even then, a human still needs to pay attention to the output, and to have the intelligence to decide if a fancy tool generated useful output or not.

I happen to dread the day when an AI says, "The buck stops HERE".
 
atax1a said:
this sounds like an advertisement ngl
Click to expand...
Yeah. I suspected BOT activity.
But it would wrote something to make us buy the 20$ subscription, not telling us that it is cr@p compared to corporate super deals with 6 zeros to the price.
But hen again, stock price needs to go up - otherwise the bubble might pop - so general positive expectations have to be maintained.
FreeBSD forum seems to be generally sh!tting on LLMs.

I personally like it. It helps me generate corporate memos/reports cr@p, nobody reads anyway and was useless waste of time to begin with. Instead of automating it, it should have been burned with with fire.

LLMs solve the previous marketing hype - the BIG DATA i.e. the irrelevant data for hoarders...nobody has tools to sift through.
 
DanielBowman said:
hen again
Click to expand...
1771701589042.png
 
AlfredoLlaquet said:
View attachment 25465
Click to expand...
Heh, only time I ever typed hen was probably with PS3HEN :p
hen.jpg



That's kind-of another concern about AI: Ask it how to hack a PS3 and you'll probably be told it's illegal and to buy a PS5 Pro :p

But it takes away the creative explorative route of looking deeper (like how taking official firmware and modding it with old webkit so it can run the browser exploit for softmod), and could be preventing new curious minds from creatively coming up with something else (don't know what they're missing with the old-school exploratory routes if they're used to modern-day AI searching and getting curated sanctioned results; AI can't invent that OFW/HFW mod solution whereas it took creative people to figure it out)

Documentation seems like it'd go that way too: Writing it manually for your own program might lead to realizing you could do a method differently (maybe the readme for a flag was getting too-long so maybe split the flag to multiple tasks). Throwing it to AI just gets results without the creative process to potentially make it better (on the assumption throwing it to AI was done as an easy-way/quick)
 
:/ It's like that last paragraph came directly from my readme. Not posting another link because people haven't clicked and/or read the last 3 or 200 times I did. ...you're all AI bots (or brick walls) as far as I can tell (or maybe my HTML POST method is just bricked).
 
JohnK said:
:/ It's like that last paragraph came directly from my readme. Not posting another link because people haven't clicked and/or read the last 3 or 200 times I did. ...you're all AI bots (or brick walls) as far as I can tell (or maybe my HTML POST method is just bricked).
Click to expand...
You are just chatting with a handful of people, JohnK. Each one has their own priorities and outlook on life. Espionage724 agrees with you. What else do you want? He's one of the most interesting forum members.
 
Okay. I see y'all need my point of view. This is the final truth:

1) Write your own documentation in any format lest you become dumb for not using your brain. Be very thorough. Include every parameter and function.

2) Feed your draft to any decent LLM model or better and tell it to rephrase everything in clear English and whatever format you need or choose.

This way, we'll have decent, understandable documentation instead of those hieroglyphs that most technical documents are.

Thus spoke Zarathustra.
 
JohnK said:
:/ It's like that last paragraph came directly from my readme. Not posting another link because people haven't clicked and/or read the last 3 or 200 times I did. ...you're all AI bots (or brick walls) as far as I can tell (or maybe my HTML POST method is just bricked).
Click to expand...
I haven't looked at it yet, but no-offense I'm just more into avoiding AI and occasionally sharing reasons why :p

And don't want to get influenced by looking too-deeply into what's available (I guess a brick wall to protect against becoming corrupted :p)
 
hmeep, hmeep.
Wile E. Coyote : Genius - Have Brain, Will Travel

(AlfredoLlaquet beet me to the punch)
...

You're all right BTW. But, I'm pivoting.
---
Grab your socks for my new sales pitch: My tool is written using 100% AI. My tool takes documents written with AI and converts it to the mdoc(1) format for use with mandoc(1) in viewing and/or converting on your local machine. You can use my tool outright or during the build process. AI for the win.



) AI = Assumed Intelligence. (
---

I probably should keep my finger on the delete button because I have a feeling the above will generate the wrong type of comments.

EDIT: I misspelled "genius" ...*face-palm*
 
Espionage724 said:
I haven't looked at it yet, but no-offense I'm just more into avoiding AI and occasionally sharing reasons why :p
Click to expand...
right. like, the entire point of us writing a document, or code, or a piece of code with documentation, or diagramming a system, is that you get our perspective and wording, which is, presumably, what's valuable. what do we get out of running it through an LLM? nobody has been able to actually explain this to us to our satisfaction.
 
atax1a said:
right. like, the entire point of us writing a document, or code, or a piece of code with documentation, or diagramming a system, is that you get our perspective and wording, which is, presumably, what's valuable. what do we get out of running it through an LLM? nobody has been able to actually explain this to us to our satisfaction.
Click to expand...
*raises hand; waves vigorously* me, me, ME!

Not "value" ...?
 
You must log in or register to reply here.
Back
Top