I build judgment infrastructure for human-led AI: portable agent skills that help people organize messy context, pressure-test thinking, deliberate consequential decisions, and test ideas before trusting them.
20+ years driving enterprise revenue and GTM strategy at Google, Meta, and Twitter. Heavy AI user long before it was a job requirement. Now building in the open.
Explore the catalog: Judgment Infrastructure for Human-Led AI
Start with the workflow: Guided judgment workflow
These are portable behavioral systems for agents, not one-off prompts. The bundle is focused on judgment infrastructure for human-led AI work: organize messy context, pressure-test the thinking, deliberate consequential decisions, and turn ideas, claims, and decisions into evidence-seeking action.
| If you have... | Use... |
|---|---|
| Messy notes, transcripts, research, or context | The Briefing Room |
| A claim, plan, draft, or idea that needs honest critique | Ground Truth |
| A consequential decision with real trade-offs | The Quorum |
| An idea, claim, or decision that needs a small test before commitment | Test Drive |
- One scenario across all four skills
- Practical use cases
- Awesome Judgment Infrastructure
- Why judgment infrastructure matters
Run one quick demo:
- Download The Briefing Room
- Download Ground Truth
- Download The Quorum
- Download Test Drive
- Browse the full catalog
briefing-room — Turns messy context into a structured brief a human can think with. It separates facts, claims, assumptions, tensions, open questions, and next use.
ground-truth — Calibrated honesty and anti-sycophancy for plans, reviews, and decisions. It challenges the premise before validating the answer.
the-quorum — A five-member expert council that pressure-tests consequential decisions. Each member reasons independently, challenges the others, and a chairman synthesizes a recommendation with a pre-mortem.
test-drive — An evidence-design skill that helps people test an idea, claim, or decision before trusting it. It identifies the evidence type, smallest credible test, needed artifact, connector path, approval gate, and learning loop.
Each repository includes Claude installation instructions and Codex installation instructions.