The Failure
Garden
A causal strategy simulator for the ways a plausible AI plan quietly becomes unsafe, expensive, and politically impossible to correct.
Introduce pressure. Watch the interactions.
Pick any stressors. Three or more reveals coupled effects that single-risk planning misses.
Causal failure map
Click a node or relationship-adjacent node to inspect the mechanism. Line styles distinguish causal roles.
What this configuration tends to produce
Run the model to generate scenario ranges. These are transparent planning ranges—not forecasts, benchmarks, or empirical probabilities.
Replay where correction gets harder
The path is the median-severity story from the modeled configuration—not a claim of what will happen.
Buy down a path, not a feeling
Every selection consumes budget and change capacity. Some slow time-to-value or create a new exposure. Outcomes recompute immediately.
Each actor sees a different failure
These are structured challenges grounded in the active model, not consensus theater.
Assume the program failed in month 14.
Separate seed facts, model assumptions, and dissent. A useful pre-mortem should be uncomfortable enough to change a decision.