The Problem with Shared Databases

A shared database looks like integration without the hard parts. In production it quietly couples ownership, releases, and failure modes. Architecture Files #10.

Why Eventual Consistency Is Usually Correct

Strong consistency everywhere is often a wish. Eventual consistency with named bounds is usually how enterprise software actually stays honest.

The Hidden Cost of Synchronous APIs

Synchronous APIs look simple until your product's latency and availability become whoever answers the other end of the wire. Production Notes #08.

Why CRUD Stops Scaling Before Databases Do

Production Notes #07. CRUD looks like a clean product model until workflows, side effects, and ownership show up. Databases are rarely the bottleneck. The create-read-update-delete assumption is.

Idempotency Is the Unsung Hero of Enterprise Software

Idempotency decides whether "twice" is safe. Production Notes #06 on making duplicates harmless before retries, webhooks, and humans force the question.

Why Retries Make Bugs Worse

Production Notes #05. Retries feel like free reliability. In production they often duplicate side effects, amplify stale state, and turn a timeout into a reconciliation project.

Designing Software for Partial Failure

Enterprise software fails in pieces, not all at once. Design for brownouts — not only for the outage that never comes.

The Myth of a Single Source of Truth

Every enterprise eventually demands one source of truth. What they need instead is clarity about who owns which version of reality — and when.

Every Integration Is a Distributed System

We call them integrations. Each one is another system with its own timing, failures, and version of the truth — whether we design for it or not.