Engineering Blog
Technical deep dives into how we parse, process, and serve the Congressional Record.
All Posts
7 articles
Field Notes
We sued Anthropic for $20 and won (kind of).
A broken login, two months of support requests, and a small-claims case in San Francisco.
Engineering Deep Dive
Our Updated Media Processing Pipeline
How we cut timestamp-matching costs 15x and made it 40% faster.
Reference
A Dictionary of Congressional Record Formatting Problems
A running catalogue of the typos, printing errors and layout conventions that make the Record difficult to parse.
Data Pipeline
Parsing the Congressional Record Ain't Easy
A field report on the cases that broke our pipeline.
Document Memory
Compressing Documents into LoRA Adapters
Testing whether a small per-document LoRA adapter can act as compact document memory with citation-aware answers.
Engineering Deep Dive
Timestamp Identification for Clip Cutting
How we locate the exact start and end times for Congressional Record statements in floor video.
Data Pipeline
Parsing the Congressional Record: Speaker Identification
Identifying speakers in the Congressional Record—and why a simple regex wasn't enough.