Apr 19, 2026
Code Review is Dead. Long Live Spec Review.
At AI velocity, code review breaks because reviewers approve code they don't understand. The solution is to shift review upstream to plans.
Featured essay
The Reliability Ceiling: Why AI Struggles With Transactional Systems
AI can be highly capable and still fundamentally unreliable. The difference between accuracy and reliability is structural, not a scaling problem.
Start here
Hand-picked essays that hold up, the best entry points to my thinking
process
Code Review is Dead. Long Live Spec Review.
Stop reviewing code you can't understand. Review the plan instead.
organizations
The Cognitive Trap Behind Scope Creep
Scope creep is a perception failure, not a discipline failure.
architecture
Agent-Native Architecture
Agents don't adapt to your architecture. They scale it, flaws included.
Recent writing
Apr 5, 2026
The Cognitive Trap Behind Scope Creep: Why Good Teams Still Fail at Scale
Scope creep persists in well-managed programs because teams can't perceive cumulative change. Here's what's actually happening, and the three simple visibility tools that counter it.
Mar 30, 2026
Agent-Native Architecture: Designing Software Systems That AI Can Understand, Operate, and Evolve
Agents navigate your codebase differently than humans. Without the right architecture, their speed amplifies your system's flaws. Here's what to do about it.
Mar 22, 2026
The AI Productivity Paradox: Why Enterprise AI Deployments Fail While Startups 10x
Enterprises see ~10% gains from AI tools while startups achieve 10x velocity. The gap isn't the tools, it's the workflow.
Jan 24, 2026
Why Code Review Breaks at AI Velocity
Nearly half of all code is now AI-generated, but your review process wasn't designed for it. The failure is structural, not a matter of discipline.
Jan 18, 2026
The Intent Gap in Software Requirements
Requirements capture the 'what' while losing the 'why,' creating a Knowledge Transfer Tax that compounds over time.