Use AI This Election

A demonstration post: Scott prompts Claude as a centrist-abundance-YIMBY voter to research his June 2026 California primary ballot, and reproduces long, high-quality outputs (the Superintendent field, Measure A parcel-tax) as evidence that AI is an effective research aid for down-ballot races few people study. He scores Claude's picks across 10 hard races (5 A / 3 B / 2 C), noting failure modes (it took his stated abundance lean as single-issue; ignored unviable fringe candidates), and an 'ignore my beliefs, who would YOU pick' experiment that proposed switches he liked. Frames it as part of his hope that post-AGI democracy improves if people use AI advisors well.

Quality 62 · Strong. Strong/Solid edge. Useful, timely, and honest about the method, but most of the wordcount is Claude's output; Scott's own analytic contribution (the scoring, the failure-mode reflections, the forward-looking framing) is modest. A demonstration rather than an essay.

Claude’s paradigm shift 42 · Moderate. Moderate. 'Use AI to research your ballot' is a timely practical demonstration, not a paradigm move; mild novelty in the AI-advisors-improve-democracy framing.

Real-world impact 2 · Minor. Minor. Could nudge readers toward AI-assisted voting, but at publication it is a within-readership demonstration with no measurable material effect. Within-niche.