IRC Log - 2025-08-06

00:02:30 ##forth <cleobuline> forthBot: S" un monstre affreux avec des yeux globuleux des grandes dents et des tentacules "S TEMP-IMAGE

00:02:40 ##forth <forthBot> https://tinyurl.com/29s3yr3x

00:03:27 ##forth <ajhidd> \3

00:19:35 ##forth <cleobuline> forthBot: S" un monstre affreux avec des yeux globuleux des grandes dents et des tentacules "S IMAGE

00:19:50 ##forth <forthBot> https://i.ibb.co/bRvSD06h/mforth-image-w1-J1-W1.png

02:19:50 ##forth <forthBot> Environment for cleobuline inactive, freeing...

11:34:35 ##forth <veltas> I'd be interested to see some tech documentation written with AI that someone considers good, because I think it would be too fluffy to read

11:35:00 ##forth <veltas> Unfortunately my experience is that the source code is the only useful documentation for most projects

11:35:38 ##forth <veltas> But I think that's unavoidable and maybe that should be reflected in our understanding of 'maintainability'

15:46:09 ##forth <xentrac> I don't think it's unavoidable, unless you include things like "TeX: The Program" in your definition of "source code"

15:46:22 ##forth <xentrac> in which case there's nothing unfortunate about it

15:52:02 ##forth <xentrac> a specific thing I did last year was http://canonical.org/~kragen/sw/dev3/smolhershey.md.html#addtoc_1, which is documentation I wrote myself

15:52:37 ##forth <xentrac> but part of that documentation is http://canonical.org/~kragen/sw/dev3/smolhersheyexample.c, which is mostly by GPT-4, although I've edited it

15:53:42 ##forth <xentrac> I did check in the original GPT-4 code in commit e0eab4643c603d2e9d57540d3a4b6026846cfddd

15:54:27 ##forth <xentrac> basically I just gave GPT-4 the header file and asked it to write an example program. but the example program it came up with was kind of dum

15:55:46 ##forth <xentrac> and also badly formatted

15:57:00 ##forth <xentrac> later I added a GPT-4-generated Bresenham line drawing routine in 03044f425bba5a3373d13f322b0052e3f95c158a

15:57:30 ##forth <xentrac> and then cleaned that up because it was also badly formatted

15:59:27 ##forth <xentrac> I felt like GPT-4 was a real boost to my productivity there, but it might have been sort of an illusion

16:01:01 ##forth <xentrac> so I don't know if that's really what you're asking for :-)

16:01:55 ##forth <xentrac> Hmm, I think I wrote http://canonical.org/~kragen/sw/dev3/didcomp.en.md largely with GPT-4

16:03:30 ##forth <xentrac> basically I asked GPT-4 if it could explain the code in http://canonical.org/~kragen/sw/dev3/didcomp.c, and it wrote what seemed like a correct and lucid explanation

16:04:37 ##forth <xentrac> so I just checked in its explanation unmodified. Also the translations into French and Spanish (which I verified were okay) and Chinese (which I did not)

IRC Log - 2025-08-06 - ##forth