69

Source document

[2025-11-09 05:53] ajaxdavis: https://meet.google.com/tpt-vtjc-cgx think ima work on the slideshow generator and air advertising [2025-11-09 07:11] lisamegawatts: I joined but was only one there [2025-11-09 22:06] traves_theberge: Wild [2025-11-10 07:16] alluring_piglet_29962: https://github.com/opencontainers/runc/issues/4990 [2025-11-10 07:21] alluring_piglet_29962: HN comments are interesting too: https://news.ycombinator.com/item?id=45871531 [2025-11-10 08:09] traves_theberge: nah fuck that guy [2025-11-10 08:14] traves_theberge: I'm seriously sad for these guys. LLM's IMHO write better code and comments and commits than 98% of people if properly used. instead of creating a policy for commits " Lets just ban the use of LLMS" LMAO [2025-11-10 08:15] alluring_piglet_29962: Well if you can't tell it was written with an llm then there's no problem [2025-11-10 08:15] alluring_piglet_29962: I think this is leading to proof of personhood somehow [2025-11-10 08:16] alluring_piglet_29962: Govt id perhaps [2025-11-10 08:18] traves_theberge: perhaps. [2025-11-10 08:19] traves_theberge: POP. [2025-11-10 08:20] traves_theberge: either way, coming from my perceptive and learning how to code in the era of LLM, its seems like a last ditch effort you save classical programming by scared boomers. js [2025-11-10 08:21] alluring_piglet_29962: I don't understand how that guy can be so confident about spotting llm work [2025-11-10 08:21] alluring_piglet_29962: It's the false positives that bother me [2025-11-10 08:21] traves_theberge: because the commit message was more than 6 words lol [2025-11-10 08:25] foxhop.: the leverage from code generation in 2025 means if anyone blocks progress they are routed around. [2025-11-10 08:27] foxhop.: use the word & code calculator to generate solutions, pick proper ones and bypass bureaucracy. [2025-11-10 09:11] ajaxdavis: depends on context like most things, that project `runc` is low level stuff that works with OS and hardware. the codebase is probably littereted with comments that link to the threads of discussion as to why `version 0.3.7 of the linux subsystem kernal blah doesnt like dates when access on an m2 processor when it is a full moon` llms are obviously no where near solving that shit yet. doesn't sound like OP is overly zealous in his recommendation to dismiss them for now. and yeah "llm generated" is too broad, and LLM could very easilly assisting in figuring out why a bug is happening. but his concern is completely warranted in how much time it takes to read PR's when you are a curator, only to figure out the PR only somewhat fixed the issue. i think the solution is to just add more testing and other processes. culturally people should start tagging there PR's with what level of LLM's they used to make the PR and then the curators can discern how they want to spend their time as the space evolves