🦋 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.
tbrowder__ guifa: it was i. sounds like good advice. thank you! 01:35
rf Morning folks 13:26
lizmat rf o/
Voldenet yes, evening 13:47
Xliff_ \o 14:42
lizmat: Remember the question I asked you yesterday and you said $*CU was the only way to get to the AST tree?
lizmat yes 14:43
Xliff_ How about this: Create a role with two methods: .ast and .set-ast. Once an .set-ast is used, it cannot be used again. At compile time, once an object has been rendered into an ast, do the following: $obj does RakuAST::Object; $obj.set_ast($ast) 14:44
This now gives us a more targetted way in and is only available at BEGIN and CHECK.
grondilu Hello all. 17:44
Do you guys keep up with the recent AI craze? I've heard that some LLM are getting pretty good at generating code and I wonder if they could be used for raku. 17:45
Most notably, I hear that some models now have context windows of ten of thousands of words, so I wonder if it'd be possible to put all raku documentation into a single file and feed it to a LLM. 17:46
I could be throwing a pie in the sky here, I'll admitp 17:48
jdv from what ive read they are not good, their ok 18:14
surprisingly ok but not great
Xliff_ grondilu: I've tried both Bard and ChatGPT. They have varying results based on the language used. Many of them use models that will confuse Raku and Perl together, probably because of the Raku/Perl6 link. 18:34
So if you ask either one to write Raku code, it will be a mix of Raku and Perl.
I wonder if this simple rule would help it... 18:35
"If page matches the regular expression "\$\w+\-\>" then it is to be used as a reference"
lizmat: Did you read my question from this morning? 18:36
lucs grondilu: An interesting read: cacm.acm.org/blogs/blog-cacm/27357...s/fulltext 20:48
tellable6 lucs, I'll pass your message to grondilu
jast I think people just read way too much into the (admittedly impressive) recent progress made with large language models 21:04
specifically people tend to think that large language models "think", which honestly would be an awful stretch of the word
but even though the GPTs are impressive at what they do, many of the important problems in AI progress are as unsolved as ever 21:06
jast and even in wildly more popular languages they'll still happily spit out code, or suggestions for improving code, that don't actually work :) 21:09
lizmat Xliff_: was afk and will look after some sleep 22:27
Xliff_ OK 22:37
Voldenet grondilu: I've tried some tasks, in general gpt3.5 was good at generating things that didn't even compile, but for instance using linux apis is a lot faster with gpt3.5 even with some broken code to fix 22:44
tellable6 Voldenet, I'll pass your message to grondilu

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!