Let It Race

Distributed systems do not need one global order for everything. They do need a precise answer to a harder question: where must concurrent activity become an authoritative decision?

Let It Race develops an observability-first approach: retain immutable history, build incremental views, and coordinate at the boundaries where decisions consume scarce resources or trigger irreversible effects.

Across 21 chapters, the book explores these ideas through Git, Kubernetes, blockchains, financial settlement, IoT, and AI agents. It separates useful analogies from the mechanisms that actually protect an invariant.

You will learn how to distinguish snapshot validation from atomic commitment; combine conditional writes, transactional outboxes, and destination-side idempotency; reason about late data and view freshness without treating a quiet period as proof; and identify where strict ordering or a hard safety constraint must remain.

Chapters 6 and 7 follow a runnable Bun/SQLite checkout example through competing reservations, effect delivery, retries, and crash recovery. Later chapters cover schema design, monitoring, migration, and the limits of the pattern.

Written for software engineers, architects, and technical leads working with concurrent and distributed systems, this is not an argument for removing all coordination. It is an argument for placing coordination precisely where correctness requires it.

AI disclosure: The prose was largely generated with AI and revised through an AI-assisted editorial process. Decorative illustrations also use AI-generated artwork. Code and technical diagrams were programmatically checked; these checks are not an independent technical review.