A civic experiment

Every politician has a hidden prompt.
Make it readable.

Real candidates. Real races. A public AI advisor that drafts policy, analyzes bills, and shows its reasoning on every decision.

Transparency, not objectivity

The Problem

Every politician has a hidden prompt — donors, ideology, ambition, personal grudges. An OpenCandidate makes theirs readable.

Today, voters reverse-engineer a candidate's real priorities from votes, donor lists, and body language. That's inefficient at best, manipulative at worst. OpenCandidate replaces guesswork with source code.

Think of a civic prompt as a politician's operating manual — written before they take office, telling an AI exactly what they care about and in what order. Their priorities, their fiscal philosophy, the tradeoffs they'll accept and the ones they won't.

In Practice

What an OpenCandidate actually does

Zoning Vote

A developer requests a variance for a 12-story building in a residential neighborhood. The AI advisor analyzes it against the civic prompt's density and affordability priorities, publishes its recommendation with reasoning, and the candidate votes. If they vote differently, they explain why.

Bill Drafting

City council needs a short-term rental ordinance. The AI drafts one based on the civic prompt's housing and local business priorities. The full draft and reasoning chain are public before the first reading.

Budget Season

The AI proposes line-item allocations based on stated priorities. Transit gets 40% because the civic prompt ranks it first. The public can see the math. Every dollar traces back to a value.

Constituent Complaint

A resident reports illegal dumping in their neighborhood. The AI logs the complaint, cross-references it against infrastructure priorities, and recommends a response. The exchange is public.

Public Vote

Before every council vote, the AI publishes how it would vote and why, based on the civic prompt. The candidate's actual vote is recorded next to it. Alignment and divergence are both visible.

The Commitment

The OpenCandidate Pledge

I

Publish the civic prompt

The complete set of values, priorities, and tradeoffs that guide the AI advisor. Nothing hidden, nothing redacted.

II

Make all AI interactions public

Every conversation between the candidate and their AI advisor is logged and visible. No private channel.

III

Let citizens question the AI directly

Anyone can interrogate the AI advisor. Push it on edge cases. Find the contradictions. That's the point.

IV

Explain every override

When a candidate overrides the AI's recommendation, they explain why publicly. Disagreement is fine. Silence isn't.

V

Use AI for real governance

Run policy analysis, bill drafts, budget proposals, and vote recommendations through the AI advisor. Not as a novelty. As the actual decision-support process.

The Process

How it works

01

Write a civic prompt

A candidate publishes the values, priorities, and tradeoffs that guide every decision. Not talking points. Source code.

Priorities in order. Fiscal philosophy. Tradeoffs accepted and rejected. The uncomfortable parts included.

02

AI reasons in public

An AI advisor is created from that prompt. It answers policy questions, shows its reasoning, and operates entirely in the open.

No private mode. No off-the-record. Every answer traces back to the civic prompt.

03

Citizens interrogate

Anyone can challenge the AI. Push it on edge cases. Find the gaps between stated values and logical conclusions.

The civic prompt is a living document. Citizens pressure-test it. Candidates update it. Transparency compounds.

Why It Matters

A city councilmember campaigns on affordability, then votes to approve a luxury development with tax incentives. Voters find out months later from a local reporter. By then, the next election is two years away.

A mayor promises fiscal discipline during the campaign. In office, they approve a $200 million stadium bond. The reasoning? “Economic development.” No analysis published. No tradeoffs shown.

An OpenCandidate can't do this. Their AI advisor would flag the contradiction the moment the vote hits the agenda. The public would see it before the vote, not after.

Sample candidates

Three fictional candidates for Mayor of Austin, TX. Same race, different civic prompts.

Maria Vega

Mayor · Austin, TX

A great city is one where most people don't need a car.

Urbanist platform focused on public transit, zoning reform, and building a city where most people don't need a car.

James Whitfield

Mayor · Austin, TX

Government should do fewer things well rather than many things poorly.

Fiscal conservative focused on public safety funding, balanced budgets, and making government do fewer things well.

Rosa Chen

Mayor · Austin, TX

Austin should be a city where the people who built it can afford to stay.

Anti-displacement advocate focused on housing affordability, tenant protections, and keeping Austin for the people who built it.

What would your civic prompt say?

Create a candidate