One Domain. Three Providers. Three Different Truths.
What multi-provider hotel integrations teach us about normalization, pricing and why a common domain model is much harder than it looks.
From the Architecture in the Real World seriesField Notes
Engineering lessons
from real systems.
A hand-picked collection of practical observations from architecture, TravelTech, HealthTech, enterprise AI and automation—focused on what changes when technology meets the real world.
AI Memory Is Really a State-Management Problem.
Why useful conversational systems quickly become an application architecture problem—not simply a question of context windows or vector databases.
Field NoteSecurity Before AI Architecture.
Before selecting models and RAG stacks, organizations need clear answers about data boundaries, approved environments and who can send what where.
Field NoteThe Best AI Model Depends on the Problem and the Context.
Model selection is an engineering decision involving quality, latency, privacy, cost, failure modes and the workflow around the model.
Field NoteBuild vs. Buy Is Rarely a Binary Decision.
The real decision is often which capabilities create differentiation, which should remain commodity, and where integration cost changes the equation.
Field NoteAI Should Explain Without Quietly Becoming the Decision-Maker.
A practical architecture question: where generation ends, where validation begins, and when a human needs to own the next action.
UpcomingCurated, not content marketing
We publish when there is
something useful to say.
Field Notes are adapted from real engineering work and selected writing. The goal is not volume—it is to document patterns worth carrying into the next system.