Sean McKendry

I make hard concepts playable. Not simplified — playable. The argument stays intact; you just get to hold it in your hands.

What I actually do

Take a concept most people bounce off — how a language model fabricates a citation, why a corporation is legally a person, what a headnote is and why one rewrote a century of law — and build the thing that lets someone feel it in ninety seconds. Usually that's a game. Sometimes it's a room you walk through, or a page that reads itself aloud.

The mechanic carries the argument. In Spot the Lie you don't read about machine fabrication; you get three statements, one of them false on purpose, and you find out whether you can catch it. That's the whole thesis, and nobody had to sit through the thesis.

It's the same skill a good teacher has and a good product designer has: knowing exactly where understanding breaks, and building the handhold right there. I've done it across AI literacy, civics, law, and accessibility — declassifiedbysean.com is the companion project where the same instinct runs on the personal record.

Don't take my word for it — check the line

Here's the anchor, and the whole pitch: every claim on this page is a live door into the work that proves it. No résumé lets you do that. Most ask you to trust a bullet point; this one dares you to click it. The job I'm applying for is human judgment you can verify — argued by a résumé you can verify.

• Say I can catch a machine's bluff? → The Tell · Spot the Lie · the pitch test
• Say I ran a no-lying, human-in-the-loop build? → the one rule · the public commit ledger
• Say I can write an argument that holds up? → the Ethos · the Lexicon
• Say I built the whole thing solo? → the museum — hundreds of rooms, all of it, mine.

America's most checkable résumé. It travels not because it's loud, but because you can't find the seam — viral marketing done wrong: no hook, no growth-hack, just a claim that survives every click.

Selected work — all of it live

Only Humans Score — sole designer, writer, and developer (2026–present) · onlyhumanscanscore.com — free, no login, works offline.

• The reference site. Several hundred pages on identifying AI fabrication and evaluating sources, written and shipped solo, in the open — every claim sourced, corrections dated and published rather than quietly removed, and a public change ledger anyone can audit.
• Verification tools. A daily false-claim game (Spot the Lie), a field guide to how hallucinations read (The Tell) — including the pattern that confidence rises as verifiability rises — and a two-way test for whether a claim survives checking.
• A real-time multiplayer web app. Server-authoritative party game on Cloudflare Workers with Durable Objects and WebSockets: OAuth sign-in, a matchmaking lobby, and shareable invite links that seat a guest with no account.
• Application security, found and fixed. A cross-site request forgery vector in a credentialed endpoint, an unauthenticated resource-creation flaw, and a denial-of-service condition where abandoned sessions could permanently disable a room — each closed and covered by an automated integration suite (30+ assertions, verified deterministic).
• Front-end and accessibility. Full static stack in HTML/CSS/JavaScript; installable PWA with offline support and a strict Content-Security-Policy; accessibility targeted at WCAG AAA with a published audit, including a screen-free audio mode that reads any page aloud on-device.

What I bring to an AI team

• Human evaluation & red-team instinct — spotting fabrication, testing confident output, designing the checks. The whole site is this skill, working.
• Disciplined human-in-the-loop AI use — a person who gets real volume out of the tools without ever letting them speak unverified.
• Writing that survives a check — competitive-debate-grade reasoning, fast synthesis, prose a non-expert can verify.
• Full-stack static web — HTML/CSS/JS, PWA/service workers, CSP, accessibility.
• Brand voice & communication — a distinctive, honest register built to be shared.

Why me, plainly

• Zero execution risk. The proof already shipped. You're not betting on a promise; you're reading a finished thing and deciding you want the hand that made it.
• The background check is the website. Radically legible — the record is public, dated, and checkable, all the way down.
• Remote-native. This was all built solo and remote, on a static stack, in the open.

Open to remote roles in AI human-evaluation & red-teaming, AI-assisted writing and editing, communications, and civic-tech. If the fit is real, let's talk: [email protected].

Education

Bachelor's degree, Economics and Management (major); Political Science (minor) — Albion College, Albion, Michigan, 2006.

Competitive policy debate: research under time pressure, argument analysis, and evidence evaluation. It's where the habit of checking a source before repeating it came from, and it's the same habit this whole site runs on.

This résumé leads with deployed work because deployed work is the strongest credential I have; a traditional work history is available on request. One human's honest page — no lying.