Conversation

Notices

1. Embed this notice
   hexaheximal (hexaheximal@mstdn.social)'s status on Friday, 15-May-2026 00:06:21 JST hexaheximal

   https://git.openwrt.org/openwrt/openwrt/log/?qt=grep&q=Claude

   Are you fucking kidding me???

   I knew OpenWrt was slowly falling apart as a project for different reasons, but this is the final nail in the coffin for me taking it seriously going forward.

   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Friday, 15-May-2026 00:06:19 JST :blobcathug:

     in reply to
     @hexaheximal :blobcatgoogly: why is it bad to use claude for reviews?
   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Friday, 15-May-2026 00:34:12 JST :blobcathug:

     in reply to
     @hexaheximal :blobcatgoogly: ok, I haven't read through the links in full detail yet, but when I look through the sources the following thoughts come to mind:
     
     AI as a Fascist Artifact:
     I agree on some points within this Article, on others i dont. Where I would disagree is the big difference when AI is used as a tool to get stuff done. In this specific case AI more or less executes instructions, so therefore the person giving the orders decides actively to use a tool with fascist ulterior motives wich makes the person a fascist but not the tool.
     
     Open slopware - why not llms: I agree when it comes to the arguments of Stolen Training Data (even tho not fully since i, in very simple terms, give a fuck about copyright) and Environmental Impact.
     
     ai-alternatives/llm-afflicted-software & Your Brain on ChatGPT: I wonder if this source is relevant to my question / our situation...
     
     I think i should explain that briefly, I see a difference in how AI is used. As part of a review process, why not?
   • Embed this notice
     hexaheximal (hexaheximal@mstdn.social)'s status on Friday, 15-May-2026 00:34:14 JST hexaheximal

     in reply to

     • :blobcathug:

     @Jain https://tante.cc/2026/04/21/ai-as-a-fascist-artifact/
     https://www.media.mit.edu/publications/your-brain-on-chatgpt/
     https://codeberg.org/small-hack/open-slopware#why-not-llms
     https://codeberg.org/ai-alternatives/llm-afflicted-software#purpose

   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Friday, 15-May-2026 03:48:28 JST :blobcathug:

     in reply to

     • justsoup :asexual_flag:
     @justsoup @hexaheximal When I read through the latest news and statements from devs of well known projects, yes they tie skills to a subscription in terms of pay more for better.
     On the other hand, im rather not sure if LLMs actively deskill people, enable unskilled people or beeing a support to skilled people... I'm simply interested in the latter. If I don't want people to become stupid, then i do have other targets rn, for example tiktok. The quality of the code was never the topic here, and yes, i acknowledge the environmental impact, which I also criticize myself
   • Embed this notice
     justsoup :asexual_flag: (justsoup@mstdn.social)'s status on Friday, 15-May-2026 03:48:30 JST justsoup :asexual_flag:

     in reply to

     • :blobcathug:

     @Jain @hexaheximal One of the major issues of LLMs is that they tie skills to a subscription. Developers have learned how to review code and debate changes effectively, but LLMs actively deskill people and cause them to rely solely on a paid tool. Another major issue is that the code that LLMs output are differently licensed depending on the legal jurisdiction, which can cause legal issues with licensing. Then there is the environmental issue of power and water and land used.