The discourse problem

ENS governance produces volumes of forum text every week — thousands of posts, hundreds of threads, a steady stream of discussion that shapes every major decision. No one reads it all. Stewards skim it, contributors track threads they're personally in, and structural tensions that span multiple conversations stay invisible until they surface as friction in a vote.

What we built

We adapted Talk to the City (T3C) — an open-source AI tool from the AI Objectives Institute — for the ENS governance forum. The original T3C pipeline assumes flat conversational data; ours handles the forum's hierarchical thread structure and extracts deeper meaning from quote-and-reply chains.

The bigger change is conceptual. Instead of mapping discussions to topic labels ("treasury," "stewards," "spam"), the pipeline extracts Themes and Tensions — spectrums between two legitimate positions rather than prescriptive categories. A tension isn't a problem to be solved; it's a structural feature of the conversation that needs to be made legible.

We processed the ENS governance forum corpus through the modified pipeline.

What we found

The tensions surfaced algorithmically lined up closely with those independently identified through qualitative analysis of 22 stakeholder interviews conducted during the Metagov Phase 1 study.

This is a cross-validation worth pausing on: a fully automated pipeline reading forum text recovers the same structural tensions a human researcher arrived at through hours of interviews. It suggests forum discourse, even at its messiest, faithfully reflects the real governance state — and that the bottleneck isn't producing analysis, it's making it continuously available.

What this enables

A prototype is deployed and browseable. The proposal is to run the pipeline continuously as new forum activity arrives, turning governance from a series of snapshot retrospectives into a live surface: stewards and contributors can see what's contested, what's converging, and how positions are shifting in something close to real time.

This research is one piece of a larger thesis at Lighthouse: governance worth defending requires being visible.