Probability of Success: The Number That Moves the Money
Stage × AI, issue 4 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 research on forecasting and clinical development, not firsthand deployment. The buildable pattern at the end is not: it is an unbuilt design. Argue with it.
What this stage really does
Probability of success, PoS, and its siblings PTRS, POL, PRS: probability of technical and regulatory success, of launch, of regulatory success, is the number a development organization attaches to each program's future: will this trial read out positive, will this asset reach approval, will it launch. It drives the decisions no other number in this series touches: fund or kill, accelerate or park, partner or go alone, which Issue 1 designs get built at all. It is a forecast, this series' recurring species, but this one allocates capital.
Under the Hood · 5
5 capability-angle deep-dives generalizing this essay's pattern for AI and software engineers.
The reasons a project feels exceptional are rarely the reasons that matter to its odds.
A probability of success with no seams can't be argued with. Split it into factors and the argument becomes possible.
Editing the prompt until the model agrees with you produces your own opinion wearing a neutrality badge.
A forecast that hasn't changed in years isn't stable. Nobody is checking it against what actually happened.
Teams that rate probabilities for a living almost never check whether their 60% calls come true 60% of the time.