Field notes from building AI-native systems, teams, and operating models — for the work that still has to ship on Monday.
Four essays, one per anchor. Start anywhere — they all name the same shift from a different angle.
Vibe coding vs. reliable systems
Vibe coding vs. reliable systems. One is typing into a chat window and hoping. The other is engineering.
The harness
If you cannot pinpoint the layer where a failure occurs, your architecture is incomplete.
Specify. Direct. Validate.
AI doesn't misunderstand requirements. It fills in gaps. Silently. Confidently. At scale.
Architect-CEO
The ratio of architects to implementers is inverting.
Four recurring themes run through the series. Each is a lens on the same shift — from treating AI as a clever tool to engineering the systems that make it reliable.
The paradigm shift at the heart of the series: from typing into a chat window and hoping, to engineering the systems that make AI reliable.
All 1 essay →The engineering substrate around a coding agent — specifications, validation, workflow, context — that turns probabilistic output into production-grade software.
The loop that replaces reactive prompting. Write the spec, direct the agent against it, validate the output — then fix the spec, not the code.
All 1 essay →How leadership roles change when agents execute. Engineers define outcomes and validate them; the organization writes software differently.
All 2 essays →