The author argues that frontier-level AI has effectively become free and universally accessible, following Alibaba's release of Qwen3.8-27B, a powerful model that "local AI" enthusiasts have since shrunk to run on ordinary laptops and, soon, smartphones. This matters because it removes the cost barrier that previously limited who could deploy advanced AI, meaning capable models will soon run on hundreds of millions of everyday devices at no charge, for good and bad actors alike.

The piece traces how unpredictable token costs in early 2026 made users cautious "tokenmisers," a mindset shattered when Qwen3.8-27B proved runnable on high-spec laptops, and developers are expected to fit it onto smartphones within roughly six weeks to six months. The author likens this to the early-1980s home computer era, seeing genuine benefits such as a "cognition surplus" for tasks like accounting or photo sorting, but warns the bigger consequence is that malicious actors will gain equally powerful, always-available AI to deceive and attack at unprecedented speed and scale. The proposed defence is fighting AI with AI, using pooled resources and defensive swarms to keep people safe, succeeding largely unnoticed unless they fail.