TL;DR
Open-source models are reshaping the US-China AI rivalry by uniting all non-closed forces, yet a quick win is impossible; it will be a protracted struggle.
The author uses Mao Zedong's 'protracted war' theory as a lens on the US-China AI race: American closed models currently lead, but open models are changing the rules.
How open source turned the tide
The 'theory of defeat' says America's lead in compute and data is unassailable, yet a timeline shows Chinese open models closing in fast on top closed ones. K3's release even drew open letters from US firms like Nvidia and Microsoft defending open source — proof it has forged a 'united front'.
Intelligence as cheap as cabbage
Qwen 3.8 27B, with just 27 billion parameters, performs at Opus 4.6 level and runs on consumer GPUs; one enthusiast even deployed it on a modded 3080. Intelligence is becoming dirt cheap, which could spur more innovations like Xiaozhi AI.
Protracted war and base areas
Ninety percent of future token demand will be met by open, smaller models; only a handful of complex tasks will need pricey closed ones. Meituan's LongCat, the first model fully trained and run on domestic chips, is AI's 'base area'. The author concedes the road is tough, but final victory is achievable.
Curated from high-quality sources, with concise summaries and key takeaways.
