---
title: "Local and open source AI has to win"
date: 2026-07-15
canonical: https://solmaz.io/x/2077339203986509903/
x_url: https://x.com/onusoz/status/2077339203986509903
license: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
---

A lot of criticism coming towards local models, and the money people are spending to run them

Some of these criticisms are valid. No, the layperson will not buy a DGX station, nor spend $50k on a rig

They will spend max $3k on a computer, and if their work really necessitates it, up to $10k

But in some of these criticisms, I see a lack of first-principles thinking

Local models might be shittier now compared to the ones you buy from the cloud

The question is, do you WANT to LIVE in a world where you don't have a choice but to RENT your intelligence? From a duopoly that can extort you for the last cent in your wallet, in the long run?

I personally do not like that idea. So open source AI HAS to win and keep on winning, continuously and permanently. There is no other choice

There is also a preferential belief that AGI/ASI will be reached, but then we will NOT be able to use that to make local models work efficiently???

Like, can you believe the rate of optimization and compression that has been happening to models in the last 3 months?

Would you have believed that a o4-mini level model could run in your phone and o4 in your desktop, if we had told you 1 year ago

When we are finally there, we will probably look back at this moment and think, "wow, we had not even started to see the real gains from optimization"

Local and open source AI HAS to win, and it is up to US

So STOP complaining and get to work

*Part 1/2 of a thread; root: https://solmaz.io/x/2077339203986509903/*