Build vs Buy the Model

'Build vs buy' for AI is really three decisions: which model (frontier API vs open-weight self-hosted vs fine-tuned), what context layer (your retrieval, tools, data — almost always yours to build), and what workflow (always yours).

The rule that survives: buy the commodity layer — frontier models are leased capability, the cheapest way to be state-of-the-art today — and build the differentiating layer: context, evals, workflow. Own the data loop.