A new approach to voter guides (human and AI) – CA Prop 50 example

Hacker News

Share card

A new approach to voter guides (human and AI) – CA Prop 50 example

I built a new kind of voter guide you can read directly or explore conversationally with AI. It's a prototype with California's Proposition 50 (special election Nov 4) as the test case. For pragmatic reasons, it uses a Custom GPT for the AI (direct API integration is currently cost-prohibitive at scale for a public-service project). Assume better AI integration scenarios in the future (browser side-panel, better API economics, etc). But the more fundamental concept is, any integration scenario will need a highly curated, issue-specific knowledge source (the guide content and format). Would love feedback on the content format (organization, depth, balance), overall concept, and user experience. Other context for those interested... Audience: People who want to think through issues for themselves at reasonable cost to their time and sanity. Problem: Public debate is dominated by advocacy rhetoric—facts twisted to fit narratives, nuance lost to slogans. No one enjoys sifting through that noise for signal. And AI assistants trained on (or searching) the discourse can end up echoing the distortions, or diluting valid arguments along with invalid ones while attempting balance. Approach: Provide highly curated, balanced, and verified content for informed decision-making. Structure content both for people to read and for their AI assistants to use conversationally. Prompt the AI for strict adherence to guide content in presenting arguments and key facts.