Blog
Articles
Notes from the field
The blog follows the life of the lab: learnings, failures, difficulties and new capabilities on the flagship and its outer islands, with regular detours into architecture theory.
Testing an AI analyst on a real incident
Every service was up and every check was green, and some customers still could not buy anything. Three AI analysts had to work out that Canada was the market affected, then trace it to what had caused it. We ran that 162 times.
Notes from the DFW Pythoneers meetup
The slides, and the argument behind them: why a map is not enough to audit an information system, and what a simulated company shows once you put AI employees to work inside it.
What Archipel Labs is for
One simulated company, three instruments: a dataset that contradicts itself, an integration layer you can actually watch, and a way to find out what a company must be for AI to work inside it.
Introducing Archipel Labs
Why a public laboratory for enterprise architecture, and how observing a simulated company beats explaining architecture from above.
LinkedIn posts
Shorter signals
Between field notes, I try to post on LinkedIn: the lighter side of the lab, and open questions to think through together. If that sounds interesting, come follow along.
I'm speaking at DFW Pythoneers on August 6th Archipel Labs tests architecture decisions by running them rather than arguing about them. Right now that's TimberWorks: a small simulated e-commerce company with real orders and real analytics, and synthetic customers that a Python runtime drives…
The store is real. The company isn’t. TimberWorks, the simulated company behind Archipel Labs, is now open online. The storefront runs on PrestaShop. Simulated customers visit it in real browsers, browse Minecraft inspired products, abandon carts and place orders. This produces real traffic, order…
I'm a hands-on architect: give me a problem and my instinct is to test what I think is right against reality. Across a lot of systems, that habit taught me far more than any grand principle, and it shaped every recommendation I later made to clients. But testing ideas properly needs a living…
Archipel Labs started with a lakehouse use case. The original goal was to help a client explore how Snowflake and Databricks could improve customer understanding through medallion layers, star schemas, data cleansing, external tables and slowly changing dimensions. But to move beyond a naive demo…
For the past few weeks I've been quietly building Archipel Labs, a public laboratory for enterprise architecture. It has become one of my biggest projects yet, and one I hope to keep growing for a long time. Over the years I've learned that the real difficulty in architecture rarely lives in the…
Good automation isn't about connecting tools. It's about what holds the platform together. In my last post, I kept it simple: ServiceNow governs, Ansible executes. But between them sits a real platform. And a platform has needs. Here, we need authentication. Secrets. Inventory. Events. Config. A…
Should ServiceNow govern while Ansible executes? That’s the question I wanted to explore. Connecting ServiceNow and Ansible is not the hard part. The hard part is designing the operating model around it. Who owns the approval? Who owns the execution? Where does the audit trail live? How do you…