MomBoard turns raw interview transcripts into tagged, searchable evidence — pains, workarounds, budgets, commitments. And it grades you on how well you interviewed.
Open source · runs on your machine · your transcripts never leave itHow do you handle infringing listings today — what happened the last time you found one?
Every Monday I export everything to Excel and clean it up by hand — it's half my Monday, every Monday.
Honestly, I'd totally use something like what you're describing!
I'll set up a session with our legal team next week and share our takedown log.
You do the calls. Then the evidence goes somewhere to die.
One row per call, a link to a doc nobody reopens. Six months of interviews you can't search, compare, or trust.
You remember the quotes that flattered your idea and forget the ones that didn't. Your product decisions inherit that bias.
"I'd totally use this!" feels like validation. It's noise. What people already do, pay for, and commit to — that's signal, and your notes give it no structure.
You read The Mom Test. Your note-taking didn't.
From messy transcript to interrogatable evidence in the time it takes to get coffee.
Paste it, upload a .txt or .vtt, or (soon) pull straight from Google Meet. Metadata takes ten seconds.
Verbatim quotes get annotated with the Mom Test taxonomy. Every tag is a suggestion you accept or reject with one key. Nothing enters your evidence unreviewed.
Filter every quote by tag, company, date. Synthesize themes across calls. Watch hypotheses accumulate evidence — or die.
Most note-taking tools help you file. These features help you get better.
Every call gets a Mom Test critique: did you ask about the past or pitch the future? Did you fish for compliments?
Score 3/10 on your first call, 8/10 by your tenth.
Other tools summarize "positive feedback." MomBoard tracks your compliment ratio and celebrates when it drops.
Flattery is noise. Behavior is signal.
State a falsifiable belief. Every new conversation attaches evidence for or against it automatically.
Decisions point at a meter, not a vibe.
Interview transcripts are the most sensitive documents a founder owns. Treat them that way.
One SQLite file, no infrastructure. Postgres when you grow. Open source on GitHub.
Point it at Ollama and a Qwen MoE model — a 32GB-RAM machine runs the whole pipeline. Zero API calls, zero per-token costs.
Use OpenAI for maximum tagging quality — or mix both, per pipeline stage. Your call, per agent, one env line.
So here's our scoreboard, the same way we'd make you keep yours.
Everything else is in the README.
No. Self-hosted, your database, your models if you want. There is no telemetry — the footer of this site is the whole privacy policy.
No. A CPU with 32GB of RAM runs a local MoE model comfortably. Tagging happens as a background job, so speed genuinely doesn't matter — quality does.
Everything lands as a suggestion until you accept it. Review is keyboard-first — j/k to move, a to accept, x to reject — and takes about a minute per call.
No. It's inspired by the note-taking system at the back of the book, with attribution and admiration. Buy the book. Seriously.
The software is free and open source. You pay your own LLM costs — or nothing, running locally. A hosted version may come later; if you'd commit to that, tell us. That's a 🤝, and we count those.