On Building a Second Brain
Everything I make starts the same way: I'm walking somewhere, an idea arrives, and I press record. The pipeline I'm building — slowclaw.social — exists because I got tired of those recordings dying in a folder. The question that focused it:
"What am I building? A content-creation plus content-consumption app. A personal assistant is not a second brain — maybe I'm building a second brain kind of thing."
The insight, honestly held
In January 2026 I watched an app I'd sketched months earlier — an assistant that improves itself through its own interface — get built by someone else and celebrated everywhere. I felt a flicker of disappointment, then something better: confirmation. The trend was real. I'd caught it independently. And my addition still stood: a laid-back assistant, one that isn't in a hurry.
"There are amazing builders out there. I put out ideas, and they can build to completion if it catches their interest. Somehow the universe listens."
The layer-cake
The architecture, as it lives in my head and in the code:
- Layer 1 — Capture. Voice memos and text notes, with on-device transcription. Distraction-free: while writing, everything else hides away.
- Layer 2 — Memory. Vectorized notes compared against what I've already stored, so the raw layer feeds a cleaned layer I call the NAND layer — new ideas in, duplicates out.
- Layer 3 — Expression. Feed the raw material to agents: NotebookLM-style podcasts to share with friends, tweets that grow into articles, articles that grow into chapters of a book that doesn't need publishing soon.
Give everything to the machine
The workflow I'm designing is total: whatever I do, give it to the AI. It drafts, it generates options, it judges its own output — "use an LLM as a judge to weed out bad ideas." What comes back to me is minimal: questions, suggestions, the places where my actual expertise is needed. Full trust in, minimal friction out.
Start with automating what you already do. Then the simple products. Then the apps. That's the whole plan.
Sources: memos Cosimastraße (Oct 2025 – May 2026), Big AI Root Node Project, Future of AI Vibe Coding, 091302 Journal entry. Full transcripts in the archive.