I disagree with this hot take. While perhaps correct for certain aspects, but I don’t see local ai going anywhere.
This post mentions nothing about the finite cash streams, circular profit structures, or the environmental/social/political impacts of the ever-expanding data centers where these VMs and AIs will supposedly live forever.
Will the cutting edge typically be with the cloud providers? Sure, it’s their existential requirement.
But, will and do local models fill a niche that will never go away in my view? Also yes.
Local AI isn’t going anywhere- it’s just getting started. Ten years from now either we will have dumb terminal glasses that can’t do anything without an internet connection, or local-first AI will be supplemented by cloud providers for the larger needs. I vote the latter, and it would appear some such as Apple do as well.
Not nearly every use case for AI is a coding environment. This person seems to forget there aren’t 8.3+ billion developers on Earth. And not nearly every use for AI requires a cutting-edge model.
As the author stated, he earns money peddling hardware for these large, distributed coding agents so - to put it nicely - he is biased. I fundamentally disagree with almost everything he said. Local models are just starting to have their moment in terms of speed & intelligence. Theyre closing the gap on intelligence to a point where I can see future advancements in both software and consumer-level hardware making these distributed agent workflows kind of pointless.
Yup agreed, I've ran 1 billion tokens through with >3000 steps and multi-hour autonomous output with nothing more than a 5090, Deepseek Harness, and Qwen 3.8 27b Quasar NVFP4. You get 260k context, ~150 token/s continuous performance, and full agentic capabilities.
13 comments
[ 0.27 ms ] story [ 21.4 ms ] threadThis post mentions nothing about the finite cash streams, circular profit structures, or the environmental/social/political impacts of the ever-expanding data centers where these VMs and AIs will supposedly live forever.
Will the cutting edge typically be with the cloud providers? Sure, it’s their existential requirement.
But, will and do local models fill a niche that will never go away in my view? Also yes.
Local AI isn’t going anywhere- it’s just getting started. Ten years from now either we will have dumb terminal glasses that can’t do anything without an internet connection, or local-first AI will be supplemented by cloud providers for the larger needs. I vote the latter, and it would appear some such as Apple do as well.
Not nearly every use case for AI is a coding environment. This person seems to forget there aren’t 8.3+ billion developers on Earth. And not nearly every use for AI requires a cutting-edge model.
Given that Nitter et al are now mostly gone, please stop posting Twitter links to HN, as it appears to be against the site rules.
At this point I just assume the people publishing this stuff are not able to discern the difference between good writing and bad.
English isn't even my first language and its plain as day to me when someone uses AI to generate these walls of text.