The article argues that the golden age of AI applications is underway, backed by three developments: the US government shutdown of Fable access (underscoring regulatory risk and the need for open-source/local models), Satya Nadella’s thesis that the moat lies in human expertise and the system around the model, and Salesforce‘s $3.6B acquisition of Fin (formerly Intercom), which used open-source models for price/performance.

Building AI applications requires mastering three new disciplines: selecting the right models (each with trade-offs—Kimi K2 is fast but less precise, Qwen 3.6 27b is small but can stall, GLM 5.1 is a good coder but slow), designing the hill-climbing loop for agentic improvement, and evaluating model+loop performance—a labor-intensive task best handled by vendors to amortize costs.

Companies that master these three disciplines will own the golden age.