The Information Layer

Can AI understand the political world?

AI carries embedded ideological biases, struggles to reason about forward-looking political questions, and operates within an information ecosystem that has its own deep structural problems. We're building tools to measure and fix that.

Prototypes

AI Political Bias Tracker

Independent, standardized ideological assessment across every major AI model. We surveyed Americans on their perceptions of AI political bias and found that models across the board lean left — and that Americans across the spectrum want less slant, not more.

AI's Political Architecture →

AI Political Advisor Eval

We tested what happens when people ask AI for voting advice. One in five Americans already do. Every major model recommends fringe parties in democracies worldwide — in Japan, AI tells left-wing voters to support the Communist Party.

AI Is a Shitty Political Advisor →

Forecasting Agent

AI agents that trade political prediction markets on Kalshi. Tiered architecture: cheap models for triage, expensive models for deep analysis, LLM councils for the hardest calls. Key finding: better base models beat clever prompting.

Can AI Reason About Politics? →

Texas Primary Experiment

We beat prediction markets and polls in calling the Texas Republican primaries using AI-driven forecasting. Both traditional information-aggregation mechanisms failed where our agentic approach succeeded.

An Agentic Experiment in Texas →

The 100x Research Institution

A research swarm where AI agents propose, execute, and evaluate new studies. Expert researchers direct dozens of agents on ambitious programs — producing replicable, systematic, and continuously updated research at unprecedented scale.

The 100x Research Institution →

Essays

Building the Truth Machine

Prediction markets promise to be truth machines. The reality is thinner. Our analysis of accuracy across hundreds of political contracts: sharp on elections, fuzzy on everything else.

Read on Substack →

When Predictions Become News

The CNN-Kalshi deal and what happens when market prices stop being information and start being the news itself.

Read on Substack →

Inside the Markets Aggregating Political Reality

How prediction markets work, where they fail, and what it would take to make them reliable infrastructure for political intelligence.

Read on Substack →

Don't Let AI Choose Your Politics

What happens when AI sits between voters and their votes — and why getting the foundations right now is easier than fixing them after they're entrenched.

Read on Substack →

Inside the Roblox Casino

Seventy million children log onto Roblox daily. The rules that govern their experience — monetization mechanics, recommendation algorithms, platform policies — are algorithmic governance in miniature. Nobody voted on it.

Read on Substack →

The Algorithm Behind 'Steal a Brainrot'

Why Roblox has become a casino for kids. The gambling mechanics hidden inside children's games, and what it tells us about unaccountable algorithmic governance.

Read on Substack →