Christine Lemmer-WebberSubmitted without commentary: "AI Might Be Our Best Shot At Taking Back The Open Web" by Mike Masnick https://www.techdirt.com/2026/03/25/ai-might-be-our-best-shot-at-taking-back-the-open-web/
this is just slopaganda
Mark Shane Hayden@cwebber I am seeing more of this lately with the themes of "but watabout Locally Run Open Models" and "I'm TOTALLY not the ass of the Reverse Centaur REALLY" and it makes me kinda sad because I have cone to the realisation that #GenAI is still a misapplication of #LLM technology that rots your brain no matter how you slice it.
Dude misses the days of Web 1.0 when "view source" was actually useful and informative then advocates for the addition of another layer on the Poop Parfait that is the Modern Web to separate us further from The Source?
No. NO. How can someone so respectable with such a deep history with the WWW miss the point SO BADLY? The solution is to make The Source easy again, not to add Agentic AI on top of front end frameworks that talk to back end frameworks running in containers on VMs hosted in a cloud of corporately owned computers!
You CAN still write simple, easy and useful web pages with a subset of modern HTML, CSS and JavaScript. The solution is #SmallWeb not this garbage!
Marcus@puzzled @cwebber it has already been observed that when two LLM agents interact with each other long enough they (d)evolve into using a sort of machine code style jibber-jabber, even when they do manage to stick to their intended tasks, perhaps because it is more "efficient".
So yeah in some possible hellish future timeline we could end up with software being developed by a herd of pet LLM agents yammering amongst each other like a bunch of Furbies to create jibberish for another set of LLM agents to render and execute apps but don't worry we'll only need a few dozen gigs of RAM and a couple hundred GPU cores for our everyday tasks this is FINE it'll be OK.
passngrin@msh @cwebber
Could someone inform me regarding the "Local models" talking point? My understanding is that it would involve the NPU and some amount of storage space. I'm not even sure if it would be entirely local or if it would involve downloading some data from corpos.
As for GenAI:
1. "Local LLMs" might he useful for word-prediction when typing. I really doubt it would have a positive effect for communicating, storytelling, or even generating prompts
2. "Locally generated" visuals or audio seems even less likely. Maybe a studio would use it to make soundfonts?
3. Local code generation or analysis is the one that I have trouble predicting, despite my experiences in coding.
Could someone inform me regarding the "Local models" talking point? My understanding is that it would involve the NPU and some amount of storage space. I'm not even sure if it would be entirely local or if it would involve downloading some data from corpos.
As for GenAI:
1. "Local LLMs" might he useful for word-prediction when typing. I really doubt it would have a positive effect for communicating, storytelling, or even generating prompts
2. "Locally generated" visuals or audio seems even less likely. Maybe a studio would use it to make soundfonts?
3. Local code generation or analysis is the one that I have trouble predicting, despite my experiences in coding.
@msh @cwebber
Anyways, if Local AI does involve Big Corpo data, then I want to repost this article talking about how exploited African workers are used to support AI: https://www.404media.co/ai-is-african-intelligence-the-workers-who-train-ai-are-fighting-back/?ref=daily-stories-newsletter
Anyways, if Local AI does involve Big Corpo data, then I want to repost this article talking about how exploited African workers are used to support AI: https://www.404media.co/ai-is-african-intelligence-the-workers-who-train-ai-are-fighting-back/?ref=daily-stories-newsletter
