Notes
Short-form insights that highlight specific aspects of my analysis. Quick reads, focused takeaways.
What 510 Contracts Taught Me About Training Data
My first ML model predicted the same 9 labels for every input. The fix wasn't a better model — it was a better data pipeline. The difference between 510 training examples and 15,700.
When Consistency Beats Intelligence
A 70,000-person field experiment proved AI doesn't need to be smarter than humans to outperform them. It just needs to be more consistent. Better inputs, not better decisions.
The 60-Point Drop: How Structure Recovers What Chat Loses in Medical AI
Medical AI identifies conditions 94.9% of the time. Give people access to that same AI, and accuracy drops to 34.5%. Worse than Googling it. The bottleneck isn't the model.
The Expertise-to-Agent Strategy: Why Purpose Precedes Architecture
McKinsey, MECLABS, and MinistryAI all encoded expertise into AI agents. Two critical filters—Intent and Implementation—determined their vastly different architectures.
The 7% Tipping Point
Most transformations fail because they create users, not owners. Research shows 7% employee ownership—not participation—doubles success rates.
Westerman's Risk Slope
Technology changes quickly. Organizations change slowly. Why AI adoption must climb three levels—and why skipping validation gates makes the wheel fall off.
February 2026
What 510 Contracts Taught Me About Training Data
My first ML model predicted the same 9 labels for every input. The fix wasn't a better model — it was a better data pipeline. The difference between 510 training examples and 15,700.
When Consistency Beats Intelligence
A 70,000-person field experiment proved AI doesn't need to be smarter than humans to outperform them. It just needs to be more consistent. Better inputs, not better decisions.
The 60-Point Drop: How Structure Recovers What Chat Loses in Medical AI
Medical AI identifies conditions 94.9% of the time. Give people access to that same AI, and accuracy drops to 34.5%. Worse than Googling it. The bottleneck isn't the model.
January 2026
The Expertise-to-Agent Strategy: Why Purpose Precedes Architecture
McKinsey, MECLABS, and MinistryAI all encoded expertise into AI agents. Two critical filters—Intent and Implementation—determined their vastly different architectures.
The 7% Tipping Point
Most transformations fail because they create users, not owners. Research shows 7% employee ownership—not participation—doubles success rates.
Westerman's Risk Slope
Technology changes quickly. Organizations change slowly. Why AI adoption must climb three levels—and why skipping validation gates makes the wheel fall off.