Writing
Pieces on AI architecture, context engineering, building production multi-agent systems, and the craft of engineering.
The engineering system as People, Architecture and Systems/Processes executing roadmap bets, and three properties to observe it by: Efficiency, Accuracy and Growth.
Coding agents let repositories evolve faster than teams can absorb the decisions being made. Architecture becomes a set of governed decisions that move through spec review, implementation and drift detection at the same rate as the code.
If engineering can implement a feature 3x faster but it takes the same time to learn whether customers want it, we haven't captured the gain. Why the dev loop needs to extend from delivery to learning.
Skill-based interviewing gives us a way to change hiring as engineering itself changes. How I am updating the loop (coding with agents, feature design through rollout) now that coding agents are part of the job.
RAG answers retrieval questions. Enterprise-wide intelligence needs reasoning over relationships, state, lens and access control, and that requires a knowledge graph.
When your AI handles 25% of U.S. mortgage closings, the error rate in the critical path has to be zero. Why that constraint is architectural, not a tuning problem.
Staying close to the craft is not the same as doing it. What it took to close that gap as a founding CTO.
Prompts held to production software standards: automatic optimization against curated datasets, guardrails for edge cases, and evals with runtime observability.