← All experiments
livestarted 2025-06-01 · updated 2026-08-16
Call transcription and retrieval agent
A Whisper pipeline feeding an assistant that answers member questions from what was actually said on the calls, with the source.
stack: Whisper . Node . Claude Code
The problem
Hundreds of hours of recorded coaching calls held answers nobody could find. The knowledge existed and was functionally unsearchable.
What I built
A Whisper transcription pipeline feeding an assistant that answers member questions from what was actually said on those calls, with the source. Not a chatbot bolted onto a FAQ. Retrieval over real proprietary material.
What it taught me
It ran for six months, then a backend billing issue stopped the transcription, and nothing alarmed on that failure. It sat broken for a while before we caught it. That lesson is now the first thing I build into every scheduled job.
What I learned
- Build fault monitoring for scheduled jobs on day one.
- It ran silently broken until a human noticed. Alarms are part of the product.