Jonathan Haaswritingthemesnowusesabout
emailgithubx
Jonathan Haaswritingthemesnowusesabout
April 28, 2025·2 min read

The Day After: Building a System to Remember What Matters

A weekend project that uses the Limitless pendant and API to capture conversations before they fade, delivering a Monday morning digest of decisions...

#productivity#reflection#personal-growth#conversation-analysis#memory

Filed under Personal systems. The personal operating-system posts: single-serving tools, local context, and software that fits one person exactly.

Most conversations evaporate within 48 hours. Decisions made, commitments spoken aloud, names and context from new people met -- all of it fades unless you write it down. I rarely write it down.

So I built lifelog-email -- a system that does it for me.

The Pipeline

The Limitless pendant captures audio from weekend conversations (personal only, never work, always with explicit permission). The Limitless API transcribes and summarizes. A "Day-After Worker" parses the transcripts, extracts decisions, action items, people met, and conversation summaries. Monday morning, a digest lands in my inbox.

The worker is the interesting part. It sits on top of the Limitless API and targets the specific things that tend to evaporate fastest: commitments made in the moment ("I will send you that," "I will set that up"), names attached to context ("she is working on X at Y company"), and the reasoning behind decisions -- not just what was decided, but why.

What It Changed

"When did we decide that?" stopped being a guessing game. I have receipts. Not just the decision, but the conversation that led to it and the reasoning at the time. This has been concretely useful when revisiting plans weeks later.

Follow-ups became reliable. Remembering someone's name, what they are working on, and what you discussed is the difference between a warm follow-up and a cold one. The digest captures this automatically.

Speech pattern feedback was unexpected. Early digests revealed I overuse "you know" and "like" in casual conversation. Once you see it transcribed, you cannot unsee it.

The Rough Edges

The pendant misses things. Summaries are sometimes inaccurate. The system flags conversations I do not care about. None of the NLP is particularly sophisticated -- it is a weekend project, not a product.

But even a janky version of "review what happened last weekend" beats relying entirely on memory. The bar is low, and any system that clears it is worth having.

Share:
//

More in Personal systems

Previous on this shelf: Building a Developer Environment That Actually Works: My Dotfiles Journey

Next on this shelf: Forget Perfect Data: Building a Usable Voice Profile Extractor

Open the full shelf

emailgithubx