I build tools for working with AI, checking what it actually did, and carrying the learning into the next run.
My work connects three questions: What should the agent do? What does the evidence support? How should we explain it? Start with the project closest to your work.
Subscribe on Substack · LinkedIn · X · Book 1:1 time · Learn with me on Maven
| Project | Use it when you want to… | First useful step |
|---|---|---|
| Company Stack | Operate a small digital-product business with bounded agent tasks and checked outcomes | Run the local company demo |
| ChatData | Analyze data, challenge the conclusion, and keep reusable evidence | Install the skills and try synthetic data |
| Your Voice | Write blogs, guides, updates and explanations that preserve meaning and sound like you | Read the skill |
| World-Class AI + Data Operator | Lead AI and data work: strategy, trusted metrics, operating cadence and adoption | Explore the public operator portfolio |
Open one of the four repositories and give your agent this prompt:
Read the README and the linked getting-started guide.
Tell me what I can run locally, what needs setup, and what success looks like.
Help me complete one small example and inspect the result.
Keep my data private and ask before taking an external action.
Start with the synthetic examples where provided. Bring your own context once you understand the workflow and its checks.
Company Stack structures the work. ChatData helps establish what the numbers support. Your Voice helps turn those facts into useful writing. The AI + Data Operator portfolio brings those practices into leadership decisions and team operating habits.
For example: investigate a checkout drop with ChatData, choose one bounded experiment in Company Stack, then use Your Voice to draft a guide or customer-facing explanation. Review the claims and approve the destination before publishing. A good draft, a delivered message and a paid order are different results.
- Run a company experiment: five operating judgments, worked example, first pilot.
- Build trustworthy analytics: ChatData capabilities and AI + Data Operator playbooks.
- Improve the writing: writing briefs, editing workflows, your own voice profile.
- Learn by doing: try the synthetic analytics example.
Break a broad goal into a system. Fix the brittle execution path. Check the destination instead of trusting a success report. Make a new tool earn its switching cost. Turn repeated corrections into checks the next run actually uses.
I write about these experiments at Insight Extractor. The tiny AI company essay is the background to Company Stack.
Follow along as I build, test, and write about AI and data. If you want help with your own work or career, we can spend time on it together.
| Find me | What you'll find |
|---|---|
| Substack / Insight Extractor | Subscribe for essays on agentic analytics, data leadership, and lessons from running these projects. This is my blog and Substack in one place. |
| Follow my posts on AI, data, and careers, and join the discussion. | |
| X | Follow my shorter updates and conversations about what I'm trying. |
| Topmate: book time with me | Book a live 1:1 session for data and AI advice, career questions, or mock interview practice. Bring a specific challenge and we'll work through it together. |
| Maven: learn with me | Explore my courses and lessons on AI and analytics. |