About

What is RepCheck?

RepCheck is a civic tech platform designed to help U.S. citizens understand how their elected legislators vote relative to their personal political interests. It analyzes Congressional bills using LLMs, tracks voting records, and produces alignment scores showing how well each legislator represents a given user's stated preferences.

What is this blog?

This blog documents the development of RepCheck — specifically, an experiment in building production-grade software using AI coding agents. The core question: if you invest heavily enough in design documents, code patterns, and compile-time enforcement up front, can AI agents implement individual components independently and consistently?

We don't know the answer yet. This blog is the honest account of finding out.

The stack

RepCheck is built with Scala 3, Cats Effect, FS2, Http4s, Circe, Doobie, and Firebase. It runs on Google Cloud (Cloud Run, Firestore, Cloud SQL, Pub/Sub, GCS). The coding agents are Claude Code.

The author

Written by Eli Taveras. All source code, design documents, and CI configuration are available on GitHub: votr (main monorepo), repcheck-shared-models, repcheck-pipeline-models, repcheck-ingestion-common, repcheck-g8, and tf-repcheck-infra.