LegitOrFullOfShit

PayPal gives independent civic research a way to stay afloat. A reader pays $1 through PayPal for a sourced AI evidence file, and our server verifies the payment before unlocking it. Tips, certificates, shirts and Donkey Tracker give readers other ways to support the work. All five payment flows are demonstrated in PayPal sandbox: test money, real server-verified checkouts.

At a glance, by the judging criteria. Technological Implementation: PayPal Orders v2 checkout in the sandbox environment, Anthropic's Claude with web search behind Ask the Eagle, AG Grid Community tables, 18 states of official records. Design: a working product with accessibility controls, a mobile-tested journey and a corrections form on every page. Potential Impact: 5,400+ officials and candidates with sourced records, built for accountability after the election, not ballot help. Innovation: said-vs-did with receipts and no scores; the visitor, not the AI, makes the call; when records are missing, the site tracks the public-records request filed to get them. Presentation: live URL, free to use, no account.

Inspiration. A campaign claim travels faster than the evidence behind it. I love Salt Lake City and wanted to truly know who I was voting for. Legit or Full of Shit puts what politicians said next to what they did: their exact words beside their official roll-call votes, every line linked to its source. No scores from us. The visitor makes the call.

What it does. Pick your state or type your ZIP and see the people who represent you. Each one gets a said-vs-did file built from primary records. Ask the Eagle, an AI researcher, digs up sourced leads on any official when a visitor asks, and every lead stays labeled unchecked until a human verifies it against the record. The Eagle never issues a verdict, by design.

Where PayPal sits. Any visitor can have the Eagle assemble an evidence file - every statement, record and source on a person and an issue in one sortable table - and unlock it for $1 through PayPal. The checkout runs in PayPal's sandbox: the server creates the order, the buyer approves it at PayPal, and the server verifies the capture before the file unlocks.

The code also includes a 'Wear your call' T-shirt checkout using PayPal and Printful, designed to create sandbox fulfilment drafts only. Shirts are shown as coming soon in the live shop.

How it is built. Vite + React front end, Node/Express and Postgres on Render, PayPal Orders v2 in the free sandbox, Anthropic's Claude with web search behind Ask the Eagle, AG Grid Community for the evidence file and roll-call tables. What was built during the hackathon window, Oct 1 to Nov 12: 18 states of official data, the tallies service, the Eagle, the shop and the paid evidence file. The dated log is in HACKATHON_CHANGES.md.

Challenges. Keeping every quote word for word with its source. Labeling AI leads as unverified without making them useless. Making the paid path prove itself: cancel, decline and tampered sessions must stay locked.

What I learned. The verdict had to go. Early designs had the Eagle rate politicians, and cutting that made the product more trustworthy and the AI's role clearer.

Traction. In its busiest week, LOFOS recorded 477 pageviews from about 155 browsers. In the 30 days to November 10, it recorded 71 ballot lookups, 16 quiz starts, 8 quiz completions and 10 shares. These are analytics events and browser counts, not a count of individual people.

Accessibility. The site includes larger-text, high-contrast and reduced-motion controls, saved in the browser, plus visible keyboard focus outlines, a skip link and screen-reader labels on controls.

What's next. Planned additions include Agree to Disagree, a moderated community space with usernames, and a sandbox merch shop for T-shirts using PayPal. Donkey Tracker, for per-candidate email alerts, exists in the code as a sandbox-only demo.