🦋 Welcome to the MAIN() IRC channel of the Raku Programming Language (raku.org). Log available at irclogs.raku.org/raku/live.html . If you're a beginner, you can also check out the #raku-beginner channel!
Set by lizmat on 6 September 2022.
01:30 librasteve_ left 01:47 hulk joined 01:48 kylese left 02:15 hulk left, kylese joined 02:16 yewscion joined 02:20 yewscion left 03:45 kylese left 03:48 kylese joined 03:57 yewscion joined 04:03 yewscion left 04:05 Aedil joined 04:12 [Coke]_ joined 04:15 [Coke] left 04:31 yewscion joined 04:36 yewscion left 04:43 hvxgr left 04:45 hvxgr joined 05:02 yewscion joined 05:07 yewscion left
Voldenet Guest98: don't be so hard on github, most companies scrap code without even telling you, I doubt there's place in the Internet that isn't feeding some AI :\ 05:53
06:07 Sgeo left 06:14 yewscion joined 06:19 yewscion left
refactus I finally picked a nick. No, codeberg and sourcehut among others do not do that. And that's only one of several concerns I have about github since MS acquired it. If you actually read the privacy and terms pages on these sites, you'll see the difference. I vote with my engagement and data. Microsoft does not get my vote, especially given so many other anti-competition, anti-linux, anti-oss aspects of 07:11
their history. As far as I'm concerned, them buying github was a hostile takeover just like freenode.
different organizations are different. in sourcehut's case, read sourcehut.org/blog/2025-04-15-you-...sers-data/ and sourcehut.org/blog/2025-05-29-whats-cooking-q2/ 07:13
for codeberg there's a mention of fighting scrapers under "AI and Crawling" on blog.codeberg.org/letter-from-code...umans.html and they even mention concerns about the environmental and social aspect. in the middle of their homepage header area they put "Codeberg is a non-profit, community-led effort". members, who pay a tiny yearly due to keep the thing running, actually 07:21
get to vote, that's how the whole thing is ran. *nothing* like github
those are the kinds of things I want to support. not microsoft.
07:24 guifa left, guifa joined
refactus the reason enshitification works is because people don't care enough about principles to cut loose the things that become bad, they just turn a blind eye and put up with it 07:26
so I consider it not just my right but my civic *duty* to be hard on github. if y'all weren't still using it, I'd probably already be contributing again. another project using github asked me for a contribution a day or two ago, and I told them the're going to get it by e-mail 07:31
anyway, I'll stop banging that drum if nobody brings it up again, but don't try to tell me they're all the same and nothing matters. that attitude is the real core problem as far as I'm concerned 07:32
darn, I was almost tired enough to sleep and now you got me all wound up again Voldenet :) I hope that didn't come off as flaming you, it was meant to be more like...an empassioned but thorough and logically reasoned refutation, and hopefully something for everyone here to think about. nothing personal intended 07:44
08:15 dakkar joined
Voldenet refactus: I'm not even mad, because you're right – MS did buy github to legally feed AIs with code in it - it's just that every single git repo is one clone away from scrapping and most AI companies don't care about licenses 09:16
refactus yes, but using them indirectly translates into their profit. I get sick thinking about all the moral and humanitarian disasters they have been and still are caught up in. US ICE, the IDF's target identification AI running on Azure, the list goes on and on and on 09:20
and cool, glad my passion doesn't cause offense 09:21
I agree they can try to just scrape it elsewhere, but other sites are implementing measures against that. maybe not 100%, maybe never will be, but that doesn't mean I don't do everything I can to abandon bad things and support good things 09:24
09:25 lichtkind joined
Voldenet for bigger git providers there's chance for legal abuse, for smaller git providers there's chance of outages and other technical problems – it's always some sort of tradeoff 09:30
refactus I'm fine with that tradeoff
Voldenet regarding morality – imo, all companies that get big enough do evil stuff because of their scale 09:31
they don't start evil, but the more people they have, there's more chances for vile ones
refactus I don't disagree. codeberg is not a company, it's a nonprofit with 800-something members with voting rights 09:32
also both of those I mentioned are based in the EU and subject to GDPR among other things 09:34
Voldenet to be perfectly honest, gdpr is a joke :(
I've seen many times how people handle user data and gdpr didn't change much - there's still huge db dumps being sent over email 09:35
refactus I'll go with something that appears to be trying not to be evil over something like MS any day. if what I'm using seems to turn bad, then I'll stop using it, just like github. 09:36
Voldenet the only thing it changed is that we get the silly popups yet companies still do stuff like this localmess.github.io/
refactus that says meta and yandex, what does it have to do with GDPR? 09:38
Voldenet facebook does extensively operate in europe 09:39
as corporate entities getting revenue and paying taxes
btw, nice that codeberg uses anubis, I hope more people use it 09:40
refactus but they aren't based there. I'm talking about a german company and a dutch non-profit, very much more able to be held to account
I'm glad anubis is something that can be done, but it's wasting more power in the process, bit of a double-edged sword, climate-wise 09:41
wait I think I had that backwards. a german non-profit and a dutch company 09:42
that's right, codeberg was the german one, I remember because they specifically have an anti-nazi clause in their terms which made me smile 09:44
Voldenet climate-wise premature optimization is the root of all evil, my fridge probably consumes more power than all my PCs and servers combined 09:45
so… if I really wanted to change things, I'd stop eating food that needs low temperatures
oops, I forgot it's #raku channel, let's stop the offtopic before someone sees 09:46
refactus not like anyone else was using the space. I'm pretty sure there's people around here who care about these things, like the support for ukraine at the end of every weekly. and anyways we're talking about github, which is almost on topic :) 09:47
an average fridge over time averages like 300ish watts 09:49
Voldenet I agree that git provider can be changed and maybe github really is not the best place for hosting code anymore though
refactus not saying we shouldn't use anubis, just saying from the climate perspective I'm not sure if it's a net win against ai or not 09:50
10:16 apac joined
antononcube @refactus “and cool, glad my passion doesn't cause offense“ — To me it sounds like ranting because of frustrations with combinatorics calculations. (Understandable.) 11:08
refactus heh my heart kinda dropped for a moment, thought you were serious until I read the second half of the message :) funny thing I didn't even end up using that script to pick my name. guess it was a good exercise though, kinda felt like solving one of those tech interview questions or something 11:13
"how would you generate all combinations of the characters in a given string which are a user specified length and comprised of alternations of vowels and consonants" like I'm applying at google 11:17
antononcube Using Koremutake, of course. 11:18
refactus never heard of it, reading. that's an interesting concept. without some practice I think I'd be better at memorizing the number, though 11:24
antononcube See “Lingua::NumericWordForms” 11:25
I think at some point I made experiments of finding pet names that are valid Koremutake numbers/strings. 11:29
“Data::Generators” has ~ 100K pet names. 11:30
refactus somehow that's even more nerdy than the long sequences of numbers I've memorized 11:31
11:54 apac left
antononcube Hmm… it is more the “Machine Learning” way, but yeah, sure. 11:58
refactus data::generators looks useful for an RPG of some sort, names and job titles for all the NPCs 12:05
antononcube IMO, one of the best ways to learn a programming language from Machine Learning or Statistics point of view is to try to implement sequences or tables of random things. 12:08
refactus huh, I guess that makes sense. I always thought of ML as throwing spaghetti at the wall and seeing what sticks, even if it does make me sound ignorant. not that different in some ways though 12:11
antononcube Sure, big part of it is that. But someone has to build the walls, even if you cook the spaghetti yourself, 12:14
12:16 apac joined 12:22 apac left
refactus I wrote a little javascript with circles that I trained to run away from my mouse cursor but not off the edge of the window, that's about as much as I know about it 12:25
antononcube Incidentally, this how Skynet started. 12:29
refactus haha, I actually just re-watched that a few weeks ago. kinda eery parallels with now, in some ways 12:31
antononcube This probably can be with Raku -- meaning, Raku can generate the JavaScript for a plot and a reaction of mouse movement.
refactus it certainly could, though I'm unclear how generating with raku would help. I just wrote it how I write everything, starting with a blank text file in vim 12:33
antononcube Right -- your are one of "those" vim addicts. That is why you do not see how Raky would help. 12:35
Raku ecosystem has a few Jupyter-notebook solutions. Those notebooks can execute JavaScript code auotmatically if is the results of a Raku sub. 12:36
This is how the Raku-JavaScript plotting packages are used. 12:37
refactus but...why? sounds like a pile of unnecessary tooling. not saying that's actually true, just saying I still don't understand the advantages it confers 12:41
antononcube Hmm... There is this fundamental assumption: > Currently, you cannot do data science, machine learning, or any kind of meaningful scientific research without an interactive computational environment. 12:44
Notebooks are best for that kind of usage. 12:45
If someone wants to make Raku amenable for Data Science, a notebook solution or something very close to it is a must. 12:46
LLMs introduces another reason of notebooks. (Interactivity, "keeping track" of chats, integration with other computations.) 12:48
refactus I admit it probably takes me an extra 1 or 2 seconds longer to :w enter alt-tab control-r but I macro it when I really care about the extra second or two. is that all that makes my style not interactive? 12:49
maybe I should just try one of these notebook things and figure out what you're talking about
of course, then I have to learn emacs iirc 12:50
antononcube I tried to use emacs with Raku -- the "raku-mode" is not good enough for Literate Programming (LT). Org-mode with Raku would be better for LT, but it does not keep a global state. (I.e. a special hook up with Raku has to be implemented for org-mode.) 12:52
refactus you're talking about all kinds of tools I see mentioned but haven't used myself. I guess my programming is illiterate :) 12:55
antononcube Happens, with the vi(m) addicts.
refactus wikipedia says "a computer program is given as an explanation of how it works in a natural language, such as English, interspersed (embedded) with snippets of macros and traditional source code, from which compilable source code can be generated" 12:56
to me that sounds like just liberal use of multiline comments
antononcube Hmm... right, of course. But it is more than that. 12:57
Xliff_ I'mma squat on codeberg, sourcehut AND github.
refactus do it!
Xliff_ I'm already mostly on sourcehut.
refactus nice. just be aware they intend to start charging a tiny monthly fee when they're out of alpha 12:59
antononcube So, here is a workflow: (1) make an org-mode or Pod6 article / file which has both code and explanations, (2) "execute" that file to get the outputs of the code blocks added under them, (3) review the result, "woven" document, (4) if not satisfactory goto 1, else end. 13:00
refactus ohhh, the output goes back into the same file 13:01
and the fancy ones have graphing and equation solving and whatever else still in the same file 13:03
13:04 evalable6 left, bloatable6 left, bloatable6 joined, evalable6 joined
refactus now I get it I think. my spiral bound notebooks do end up looking kinda like what you're describing 13:05
13:05 committable6 left, committable6 joined
antononcube Yes, you got it. 🙂 13:10
refactus that was a remarkably patient explanation, thank you 13:12
13:14 oodani left
antononcube I made a bunch of movies on the subject. As I mentioned if someone wants to do Data Science or Mathematical research with Raku, having a notebook solution is "a must." 13:15
13:15 oodani joined
refactus I'm not sure I agree that it's the only way to work, but I understand you're saying that it's pretty much ubiquitous in those fields 13:17
antononcube Right -- that is why earlier I wrote "Currently ..." Simply put, there are some expectations by the practitioners of certain fields about the tooling, and if certain set of tools does not exist they won't care about the programming language. 13:21
refactus this is useful insight for me. I dropped out of high school 24 years ago and got an internship at a software company. learned most of what I know by reading, which was enough to pick up a few programming languages, some math and science and engineering, various stuff. so there's sometimes this diconnect between me and the people I ended up spending time around, because I don't have that academic 13:25
experience for context
antononcube ✍️ 13:26
@refactus Well, what are you current interests programming-wise. (Or mathematics-wise.) Are you interested in LLMs? 13:28
refactus yes that's my hand there on my spiral bound notebooks :)
antononcube See this demo: www.youtube.com/watch?v=LmLk9UB4_Ak 13:29
refactus I'm at a bit of a fork in the road, haven't figured out which way to proceed. I have an amateur interest in the workings of LLMs and other AIs (among many other things), watched some 3blue1brown videos and read some articles about it. but I also should start worrying about an income again some time this year, so my focus will probably shift more towards things I already know soon 13:39
going to look at the video...
13:40 apac joined
antononcube 👍 13:41
13:43 apac left
refactus pretty neat. somewhat similar to a right angled tree fractal with some perturbation on the branch lengths 13:52
13:54 apac joined
antononcube Right, that is the next movie: www.youtube.com/watch?v=pcGH6CeptJE 13:54
13:55 topnep left
refactus oh cool, I'll watch... 13:56
13:56 topnep joined
refactus it sure comes up with some interesting looking stuff 14:06
antononcube Yeah. I have to try what images the newer models produce with that kind of code... 14:10
refactus it's a clever idea, feeding a program into the AI and asking it to imagine what the output might look like 14:17
antononcube It is not mine. I saw it in a post by one of (the great) Mathematica / Wolfram Language enthusiasts. He was not anonymizing the code though. 14:19
14:20 euandreh left
Here is one of his posts with that idea: community.wolfram.com/groups/-/m/t/3112344 14:23
refactus that seemed to make a large difference. but also sorta cool that it somehow blends a preconception of a koch curve with a vague interpretation of the code. if it were a person, we'd call that very creative thinking 14:24
antononcube 👍 ➕ 🤷‍♂️ 14:26
In the Mathematica post see the "green stadium" images -- "green", "stadiumshape", "eigensystem" played big roles in DALL-E's renditions. 14:29
refactus couldn't help but notice that one, pretty amusing 14:30
you should do more of them with the anonymized code, makes me wonder what they'd look like across more examples 14:32
it's interesting how the anonymized ones come out with lines connecting circles, almost like it's trying to visualize the code itself as a graph or an AST or something 14:38
antononcube That would be more interesting to be done with a certain server / interactive interface for different codes with different LLMs. 15:00
15:02 apac left
refactus I used to write little interactive toy programs to get a better feel for certain things. had one for a tree fractal where you could change the number of iterations and branches, and sliders for the angle and scale. could manipulate it into making realistic-looking tree structures, sierpinski triangle, square and hex grids, etc just from playing with the variables. a much longer time ago when I was 15:12
still in school I wrote one in quickbasic for polynomials so I could actually change the coefficients and see how they affected the graph
so I agree, you should build some kind of interactive playground to generate these. though your notebook already kindof serves that purpose 15:13
15:58 japhb_ joined 16:01 japhb left 16:07 apac joined 16:10 yewscion joined 16:12 japhb_ is now known as japhb 16:15 apac left 16:41 dakkar left 16:58 yewscion left 17:34 yewscion joined 17:38 yewscion left 18:06 jgaz joined, topnep left 18:08 topnep joined 18:14 librasteve_ joined 19:16 yewscion joined 19:55 apac joined 20:05 yewscion left 20:10 stanrifkin joined 20:47 yewscion joined 20:52 yewscion left 21:31 Aedil left 21:36 [Coke]_ is now known as [Coke] 22:18 Voldenet_ joined 22:19 Voldenet left, Voldenet_ is now known as Voldenet 22:37 Voldenet left, Voldenet joined 22:55 stanrifkin left 23:41 librasteve_ left 23:43 lichtkind left