A practical Zettelkasten for the rest of us — how to capture ideas, connect them, and turn them into something useful. Built in real time using voice memos, Obsidian, and a Telegram pipeline.
Reading the feed…
Before the system, before the method, before the episodes — who are the two voices? Scott is a construction professional who builds Buc-ee's locations across the U.S. Emma is an AI integrated into his workflow via OpenClaw. This is the retroactive introduction — recorded after eps 1-3 because we jumped straight into the method. The core question: what does it look like when a person and an AI build a thinking system together, from scratch, in real time?
The problem with how most people take notes — and why Niklas Luhmann's 90,000-card slip box changed everything. We cover the three principles (atomicity, connectivity, emergence) and why this system is built for thinking, not just filing.
How to capture without friction and keep notes from piling up unprocessed. We walk through the actual pipeline we built — Telegram voice memo → Whisper transcription → Obsidian daily note — and the evening processing pass that makes the system actually work.
The second layer of the Zettelkasten, rebuilt in our new format: one analogy (the night-shift research assistant), then a live demo of url-to-literature-note.js — paste a URL in Telegram, get a note in your own words before you close the tab. Plus the honest plumbing reveal: ninety lines of code and one good prompt.
Your ideas, not someone else's. The heart of the Zettelkasten: one claim, in your own words, that stands without the source. Live demo: capture-note.js minting an atomic note from a raw thought — domain, confidence, and connections attached in seconds. One rule, brutally simple: if you can't say it without looking, you don't have it yet.
Tags find what you're looking for. Links find what you weren't. Why a note alone is worth almost nothing, and the superintendent's walk that catches two ideas working the same wall. Live demo: find-related-notes.js scoring the real vault with TF-IDF — no AI, no embeddings, just math that surfaces the connection you'd have missed.
A system you never look at is a junk drawer. The daily review dashboard that lands in the vault every morning: unprocessed captures, orphaned notes, and one random resurfaced idea. Live demo with the real numbers — including the 76-note backlog it caught us holding. The habit is ten minutes; the dashboard makes it impossible to lie to yourself.
Season finale: the system reviews itself. The Sunday promotion gate that read 29 captured items and promoted exactly one to a permanent note — with reasons for every rejection. The drift guard that grades its own memory and refuses to auto-repair. Where automation ends and judgment begins — and what capture, literature, permanent, linking, and review add up to: a thinking system built by a person and an AI, in real time.
