Long-form post-mortems from live deployments. Each writeup walks through what the system predicted, what subsequently happened, and what we got wrong. No cherry-picking.
Cascade ran continuously for 23 days during the Iran-Hormuz crisis, producing autonomous strategic assessments at sub-five-minute intervals. Ten of ten major predictions confirmed by IMF, IEA, and UNDP. Four predictions preceded the confirming events.