There's enough solid FOSS tooling out there between vLLM and Qwen3 Apache 2.0 models that you can get a pretty good assistant system running locally. That's still in the software creation domain rather than worldwide impact, but that's valuable and useful right now.