Archive
Filter decisions by domain, failure mode, time horizon, and detectability. Then read the autopsy with a calm eye.
Filters
Showing 10 of 10
Engineering / InfrastructureCadence lock-in
Fixed-interval offshore pipeline inspections
Primary assumption
Degradation drivers would remain stable (temperature, chemistry, flow regime, CP performance).
Engineering / InfrastructureModel inertia
Static corrosion growth rates in CP models
Primary assumption
The calibrated corrosion rate remained representative until the next campaign.
Engineering / InfrastructureDesign-life as a narrative
Design-life assumptions for subsea components
Primary assumption
Duty cycles would match the design basis (pressure/temperature cycles, actuation frequency).
AI / Data / SoftwareLabel lag + distribution shift
Credit scoring models trained pre-COVID
Primary assumption
The relationship between observed features (income, employment, utilization) and risk remained…
AI / Data / SoftwareChanging error shape
LLM hallucination risk assumed to decrease monotonically
Primary assumption
Model upgrades would monotonically reduce hallucination rates in our domain.
AI / Data / SoftwareConcept drift hidden by superficial equivalence
Anomaly detection trained on “normal” that no longer existed
Primary assumption
The upgraded process remained statistically comparable to the training regime.
Strategy / Product / OpsCohort mismatch
SaaS pricing locked to early adopter behavior
Primary assumption
Mainstream customers would mirror early-adopter usage and willingness-to-pay.
Strategy / Product / OpsOptimization for a world that ended
Supply chain optimization tuned for just‑in‑time stability
Primary assumption
Lead time variance would remain bounded within historic ranges.
Strategy / Product / OpsArtifact ossification
Risk matrices treated as evergreen artifacts
Primary assumption
Likelihood and consequence bands remained representative as operations evolved.
Strategy / Product / OpsOversight decay
Automation rollout assuming constant human oversight
Primary assumption
Exception rates would remain low enough for manual oversight to be reliable.