Site Selection: Paying for Zero
Stage × AI, issue 6 of ~38. One stage per issue: what the stage really does, where AI helps, where it must not, and one buildable pattern.
Confidence: hypothesis. The failure modes here are drawn from published industry data, economics and audit research, and peer-reviewed studies of algorithmic bias, not firsthand deployment. The buildable pattern at the end is not: it is an unbuilt design. Argue with it.
What this stage really does
Issue 3 drew the map; this stage picks the team. From the feasible footprint, someone must choose the actual institutions: which hospitals and clinics get contracts, budgets, ethics submissions, training, drug shipments, the full activation treatment Issue 8 priced, on the bet that they will convert their patients and their attention into enrollment.
Under the Hood · 5
5 capability-angle deep-dives generalizing this essay's pattern for AI and software engineers.
Self-description is the weakest evidence in the room. The track record is what actually happened.
A predictable fraction of every portfolio returns nothing. Price it in instead of discovering it at closeout.
Reward the raw promise and you train your sources to inflate it. Discount by track record and honesty starts winning.
A score that can't explain itself is the most comfortable place in the system for bias to hide.
In any market for scarce, in-demand participants, the best ones are scoring you back.