Writing
Thoughts on building software that lasts, AI that actually helps, and lessons from 30 years of production code.
2025
Your AI Pilot Worked Because It Wasn't Production: Here's What You Have to Change
MIT says 95% of enterprise AI pilots fail to deliver ROI. The problem isn't the model—it's that production systems are architecturally different from pilots. Here's what 20 years of running production healthcare systems taught me about that gap.
Context Stacks: Why Real AI Engineers Don't Care About Prompts
The people building serious AI products aren't obsessing over magic words. They're designing context stacks—and that changes everything about how you should think about prompts.
Why AI Agrees With Everything You Say (And Why That's Dangerous)
AI models are trained to be helpful, which means they're trained to agree with you. The research is damning. Here's why this breaks strategic decisions and how to architect around it.
Healthcare AI Is the X-Ray Moment—And Most Companies Will Blow It
Hippocratic AI just hit $3.5B valuation with 115 million patient interactions. After 20 years building healthcare systems that passed state audits, here's why most AI companies will fail the same compliance gauntlet.
Integration Rescue: A Case Study in Stabilizing a Legacy Partner
How we stabilized a failing legacy partner integration—seams, quarantines, and drills—in weeks instead of rewriting the system.