Strategy
Playing to Win
Pressure-test a business idea, offer, or product through Roger Martin and A.G. Lafley's choice cascade. 12-section output, a Proceed / Narrow / Park / Kill verdict, and zero fake encouragement.
THE WORKSHOP · 4 TOOLS
AI tools I build with, open for you to take.
Free, operator-built Claude Code skills and open-source projects, the same ones I rely on inside my own work. No signup. Fork them, adapt them, ship them. If one saves you a meeting, send it to a friend who needs it more.
TOOL 01 · SECOND BRAIN
Second Brain
Give your AI a long-term memory, built for people who talk more than they type. One prompt turns Claude Code plus a plain-text vault into a voice-first, self-organizing second brain; it files, links, and recalls so you don't have to.
TOOL 02 · SKILLS
Callable skills for Claude Code & Codex.
Download, drop into your skills folder, and invoke in plain language. Each one is a framework I actually use, packaged so it argues back instead of agreeing.
Writing
Voice Profile Builder
Build a voice profile from your pre-AI writing so AI drafts in your real voice, not generic AI prose. Point it at your old posts and essays; get a reusable profile that survives every future draft.
TOOL 03 · VISIBILITY
See what Claude Code is actually doing.
Local-first visibility into your own usage: which skills earn their place in context, and where the tokens and dollars actually go. Everything runs on your machine; nothing is uploaded.
Context in use
IDLE SKILLS · TAP OR HOVER TO UNLOAD
Claude Code plugin
Skill Usage Dashboard
Every enabled skill taxes your context window every session, used or not. It reads your local transcripts, shows real usage counts and token cost per skill, and gives each one a Keep / Modernize / Turn off / Prune verdict. Runs entirely on your machine; nothing is uploaded.
CLI + live dashboard
Claude Code Cockpit
Local-first token and cost visibility. A zero-dependency cc-usage CLI prints spend sliced by session, agent, skill, MCP server, or model, plus a one-command live Grafana dashboard with Max Plan Leverage and burn-rate panels that update as you build.
TOOL 04 · LOOPS
Loops are what come after prompts.
The library of repeatable agent loops I actually run. Start with the field guide: what a loop is, the seven parts every good one has, and when one earns its keep. Then copy the 43.
Field guide
AI agent loops, explained
What a loop is, the seven parts every good one has, and where to copy 43 you can run for coding and non-coding work.
Inside the library
43
Copyable loops
10
Categories
7
My originals
Each loop names its use case, cadence, the verification that proves it worked, and a prompt you can paste into a capable agent.
Ch. 10 of the Field Guide shows the check that makes a loop safe to leave running
TAKE IT · 6 ITEMS
The toolbox
Free and open source. Copy a command or grab the file, and make it yours.
-
01
Second Brain
A long-term memory for your AI, filed and linked by voice.
git clone https://github.com/paultaki/second-brain.gitView on GitHub -
02
Playing to Win
Pressure-test an idea and get a straight verdict.
-
03
Voice Profile Builder
Teach AI to draft in your real voice.
git clone https://github.com/paultaki/voice-profile-builder.gitView on GitHub -
04
Skill Usage Dashboard
See which skills earn their place in context.
git clone https://github.com/paultaki/skill-usage.gitView on GitHub -
05
Claude Code Cockpit
Watch tokens and cost as you build.
git clone https://github.com/paultaki/claude-code-cockpit.gitView on GitHub -
06
AI Loop Library
43 loops to copy, each with the check that proves it worked.
Get the operator lessons behind the tools.
Field Notes is where I write up what's working inside operating businesses, the AI fixes I'm testing, and the lessons I'd give a friend. Free, occasional, no pitches.