The four essays
Confidence Scores · AdTech
How I Design Confidence Scores for AI Products
A recommendation engine that beat human buyers — and got overruled half the time anyway. What I learned about the half-second where a person decides whether to trust a number.
Essay · 8 min read →
Failure States · Due Diligence
Design the Failure State First
A model that demoed at 94% confidence and was wrong 40% of the time in production. Why the happy path is the smaller half of the job, and the failure state is where trust is won or lost.
Essay · 6 min read →
Explainability · FinTech
The Explainability Layer: Making AI Legible
A compliance officer asked one question the system couldn't answer — and a trade went through unsupervised. Why "why this, not that?" is a design problem, not a data-science one.
Essay · 7 min read →
Human-in-the-Loop · EdTech
Human-in-the-Loop: When to Ask, When to Override
A 94%-accurate recommender that students skipped 60% of the time — out of click fatigue, not disagreement. When asking for confirmation stops respecting the user and starts taxing them.
Essay · 5 min read →
Shorter pieces, on Substack
Read more on Substack ↗ (opens in a new tab)The Agentic MVP: Why Your Next Launch Will Be Lovable, Not Just Viable
May 20, 2026 · Read on Substack ↗
(opens in a new tab)
The AI Fight Club: Weaponizing Claude and Gemini for Bulletproof Products
May 17, 2026 · Read on Substack ↗
(opens in a new tab)
The New Renaissance: How AI is Transforming Us from Software Operators to Digital Artisans
May 5, 2026 · Read on Substack ↗
(opens in a new tab)