The Agency-Control Graduation Framework
by Keith Coleman & Jay Baxter • Product Lead & Founding ML Engineer at Community Notes (X)
Keith Coleman and Jay Baxter lead the Community Notes team at X (formerly Twitter). Keith, a former product leader, and Jay, a machine learning engineer, developed the open-source, crowd-sourced context system that uses bridging algorithms to combat misinformation.
🎙️ Episode Context
Keith Coleman and Jay Baxter discuss the origins, principles, and engineering behind Community Notes. They reveal how a small, 'thermal' team used dynamic goal-setting and a unique bridging-based algorithm to build a decentralized fact-checking system that scales effectively where traditional methods failed. The conversation covers the transition from Birdwatch, the impact of open-sourcing the algorithm, and the operational philosophy of 'Continuous Calibration' that allows them to move rapidly.
Problem It Solves
Prevents bad actors and low-quality content from overwhelming a decentralized system by filtering for competence and neutrality.
Framework Overview
A meritocratic user progression system that ensures quality by requiring contributors to prove their judgment before earning the power to create content. Participants must first demonstrate the ability to identify helpful notes (rating) before they unlock the agency to write them.
📊 Framework Funnel
When to Use
Building decentralized networks, community moderation systems, or crowdsourced knowledge bases where trust is paramount.
Common Mistakes
Granting full access to all users immediately; using simple majority voting which reinforces bias.
Real World Example
Community Notes contributors starting as raters and only unlocking the 'Write Note' button after their ratings consistently align with the bridging consensus.
If you write one that people normally disagree find not helpful, you actually will ultimately lose your ability to write and have to earn it back.
— Keith Coleman & Jay Baxter