Open source AI is performative: skeptics note almost nobody uses it as their daily driver

flowersslop · x · 2026-09-13

A contrarian thread questions the open-weights obsession: open models aren't SOTA, aren't clearly cheaper, and the normal downloads are aligned anyway — uncensored finetunes barely get used. Finetuning is rare, so labs could offer the same pipeline anytime. The author observes that genuinely cool AI projects on Twitter almost always build on Claude or ChatGPT. Conclusion: open source advocacy is largely performative — people defend it in theory but few use it daily; the only guaranteed benefit is that nobody can take the weights away.

Original post →

More from AGI Musings

AGI Musings channel →