AI With Amirthan

← All guides

Workflows · July 11, 2026 · 2 min read

Stop tool shopping: pick one tool per layer

Productivity tool debates are really capture UI debates. Split your stack into three layers, pick one tool per layer, and let agents move work between them.

The problem

Every few months the same debate restarts. Obsidian or Notion? Notion or Linear? People migrate everything, then migrate back. I ran this debate for my own second brain until I noticed it was never really about the tools. It was about capture UI: which app feels nicest for getting a thought in. Everything else those apps do was along for the ride.

What I do instead

The moment capture moved to a chat box, the debate dissolved. My quick notes go into a Telegram chat with a small intake agent that drops them into my vault, and suddenly the big apps had nothing left to compete on. That freed me to pick one tool per layer instead of one tool overall:

  • Capture: a chat box. Fastest input from anywhere, zero friction.
  • Organise: markdown files in a git repo (my Obsidian vault). Agents read and write the files directly, and git keeps the full history of every change.
  • Execute: GitHub Issues plus Claude Code. Issue, branch, PR.

A scheduled agent files everything between layers, so I never pay the organising cost myself.

Stop asking which tool wins. Ask which layer each tool serves.

Why do markdown and git beat app databases?

Because agents reorganise constantly, and in an API-based app that costs one API round trip per block: slow and token heavy. A markdown file is one read and one write, and git gives you free undo for anything an agent gets wrong. So the working layer, the one agents touch most, should be plain text under version control. Notion still has a place as an optional publish and share surface. Linear is a tracker, so it only ever competes for the execute layer, where GitHub Issues wins because the issue to branch to PR loop is native to Claude Code.

Run the three-layer test on your stack

Paste this into Claude:

Here is every productivity tool I use or am considering:
[list them].
Sort each into one of three layers: capture (getting
thoughts in), organise (storing and linking knowledge),
execute (turning items into shipped work). Flag any layer
with more than one tool. Recommend the single tool to keep
per layer, preferring plain text formats an agent can read
and write directly.

Keep one tool per layer. Cut or demote the rest.

Do this now

  1. List your tools and run the three-layer test prompt above.
  2. Pick one tool per layer and write the choice down.
  3. Move your organise layer to markdown in a git repo, then let an agent do the filing.

Claude Code

Turn this into a Claude skill

Paste this into Claude Code. It reads this guide and builds a skill shaped to how you work. 

Read the guide at https://aiwithamirthan.com/guides/stop-tool-shopping-one-tool-per-layer. Then build me a Claude Code skill from it, personalized to how I actually work.

1. Fetch and read the full guide at the URL above.
2. Look at my setup so the skill fits me, not a generic template: skim my repos, my CLAUDE.md / AGENTS.md files, my existing skills, and my stack.
3. Create a new skill (a SKILL.md with a clear name, description, and steps) that lets me do what this guide teaches, adapted to my tools and goals.
4. Tell me where you saved it and how to invoke it.

If anything about my setup is unclear, ask me one or two questions first.

Want to build this yourself?

Join Claude Code Academy: courses, build alongs, and builders turning AI into real systems. And I post a new build on Instagram nearly every day.

Join Claude Code Academy → Follow @aiwithamirthan