## Notes from 16 March 2026 [[2026-03-15|← Previous note]] ┃ [[2026-03-17|Next note →]] I read an [interesting piece](https://rebootdemocracy.ai/blog/experimentation-public-infrastructure) by [[Cassandra Madison]] about treating **experimentation as public infrastructure**. Her argument is grounded: governments are adopting technology faster than they can learn how to use it well. The core idea is that we need safe, structured spaces where governments can test tools like AI at small scale. Madison points to a major flaw in the current system: public procurement is built for large, finished products, but that model breaks down when the technology is iterative or not yet well understood. She draws on her experience with the 2013 health insurance exchange failure in Vermont to show that you can’t outsource risk; when the tech fails, the government still carries the responsibility. She is now the Executive Director of [[Center for Civic Futures|Center for Civic Futures (CCF)]], whose Public Benefit Innovation Fund supports local governments in the US in using AI to improve social-service delivery and reportedly received 450+ proposals from 45 states. The fund is testing projects such as open-source AI to streamline work-verification processes in Maryland and efforts to simplify housing applications.