julianmatos97/mom-test-sidekick — explained in plain English
Analysis updated 2026-05-18
Get live suggested questions during a customer discovery call based on Mom Test principles.
Receive real-time warnings when a call is drifting into pitching instead of listening.
Track a coverage checklist for problem, current solution, cost, and budget during a call.
Generate a markdown debrief after each call with the transcript, facts, and a pitching scorecard.
| julianmatos97/mom-test-sidekick | 0xallam/my-recipe | 0xhassaan/nn-from-scratch | |
|---|---|---|---|
| Stars | 0 | — | 0 |
| Language | Python | Python | Python |
| Last pushed | — | 2022-11-22 | — |
| Maintenance | — | Dormant | — |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 3/5 | 2/5 | 4/5 |
| Audience | pm founder | general | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires an Apple Silicon Mac, Xcode command line tools, and granting microphone and screen recording permissions.
mom-test-sidekick, called momtest in its own README, is a terminal based coaching tool for people running customer discovery calls, built around the ideas in Rob Fitzpatrick's book The Mom Test. The problem it addresses is that founders often ask questions that lead a prospect to say polite, encouraging things that sound like validation but are not, such as claiming they would definitely buy a product. The tool sits next to a video call window on a Mac and listens in real time to coach the caller away from those habits. It works by capturing two separate audio channels, the user's own microphone and the computer's system audio, which picks up the other person's voice from apps like Zoom or Meet. Both streams are transcribed on the Mac itself using a speech recognition model called Parakeet, so raw audio never leaves the machine. Only the resulting text is sent to a language model, which the user chooses, to generate live coaching. Every few seconds the tool updates a text based dashboard in the terminal showing what question to ask next, warnings when the user is pitching instead of listening, a checklist of topics not yet covered such as the problem, current solution, and budget, and specific facts extracted from what was said. When the call ends, it writes a markdown summary with the full transcript, gathered facts, coverage gaps, and a count of how many times the user pitched. Setup requires an Apple Silicon Mac running macOS 13 or later, since the speech recognition relies on Apple's MLX framework, along with Xcode command line tools and the uv Python tool. A small Swift helper program needs to be built to capture system audio, and the user must grant microphone and screen recording permissions on first run. A coaching model is required too, either the free Codex CLI with a ChatGPT login or an API key for a provider such as Anthropic or OpenAI. The README lists honest limitations: it only supports English, it treats all non-user audio as one speaker so group calls get confusing, and it checks in only every eight seconds so it cannot interrupt a sentence mid-thought.
A terminal coaching tool that listens live during customer discovery calls on a Mac and nudges the caller toward better Mom Test style questions instead of pitching.
Mainly Python. The stack also includes Python, Swift, MLX.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly pm founder.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.