Insights
Engineering, written down
What we learn building intelligent systems, published as working notes: no hype, no filler, positions we are prepared to defend in a review.
Production AI is an engineering discipline, not a demo
The distance between an impressive AI demo and a dependable AI system is measured in evaluation harnesses, guardrails, and grounding. Here is what that engineering actually looks like.
Read articleAutomate where the judgment lives
Moving data between systems was never the expensive part of a process. The decision inside it was. AI finally makes that judgment automatable, with conditions.
Read articleModernization without the big bang
The riskiest sentence in enterprise technology is 'we will replace everything at once.' Incremental modernization ships value every quarter and never bets the business on a cutover.
Read articleA cloud platform is a codebase
If your infrastructure cannot be reviewed, tested, and rebuilt from code, it is not a platform. It is an accumulation. Here is the discipline that changes that.
Read articleData platforms earn trust or they earn nothing
Dashboards are cheap. Numbers the whole organization believes are engineered. The difference is modeling, testing, and ownership.
Read articleSecurity is an architecture property
Findings lists after launch are the expensive way to learn what threat modeling would have told you before the first line of code.
Read article