Case study

Building a Better Way to Capture Loose Ideas

Loose ideas and commitments were too easy to lose. I built a capture and routing system that turns quick notes into reviewable actions while preserving the original context.

  • AI workflow design
  • Workflow automation
  • Product iteration
  • Operational documentation

The situation

My personal and professional life generates a steady stream of commitments, ideas, notes, and follow-ups. The problem was not a lack of places to write things down. It was the friction of deciding where each thought belonged, remembering to process it later, and turning important captures into action without creating another administrative burden.

The approach

I treated the project as a product and workflow-design problem rather than a note-taking exercise. The goal was to make capture easy in the moment while preserving enough structure to support review, retrieval, and follow-through later. I designed the system in bounded versions, beginning with a manual review loop to establish trustworthy data structures and decision rules before adding automation.

The deployed version accepts lightweight captures through Discord, normalizes and routes them through n8n, creates structured work items in Linear, and appends durable processing receipts through GitHub Actions. A lightweight forwarder hosted on an OCI virtual machine keeps the capture flow available when my laptop is offline.

What I built

  • a versioned capture contract for actions, ideas, notes, and follow-ups
  • routing rules that separate actionable work from items requiring review
  • normalized Linear issues that preserve the original capture and processing context
  • an append-only GitHub receipt ledger for inspecting system behavior
  • agent-facing documentation, runbooks, tests, and validation scripts

Why it matters

The system reduces the mental overhead of holding loose commitments and ideas in my head. It gives me a low-friction capture surface, makes actionable work more visible, and preserves context that would otherwise be easy to lose.

It also demonstrates a practical approach to AI-assisted operations: automate repetitive processing while preserving human review, provenance, and inspectable failure modes.

Current state

The first automated operating version is live. Additional memory and correction-loop capabilities remain under development.

What I learned

The best automation projects begin with a trustworthy operating contract. A small, inspectable loop creates a stronger foundation than adding complexity before the system’s decisions can be reviewed and explained. For a personal system, convenience matters too: if capture is not easier than keeping the thought in my head, the workflow will not become part of daily life.