2026.10.26 · The agentic operating model, part 3Four Levels, Three Cliffs
Teams do not adopt AI. They climb four levels, and each step moves the context and the blast radius somewhere new. Most companies stand on three levels at once.8 min
2026.10.19 · The agentic operating model, part 2The Model Grades Its Own Homework
Sorting changes by risk is the right idea. Letting the agent's confidence set the class is the one input that turns every gate behind it into decoration.7 min
2026.10.05 · Adoption, part 2Nobody Resisted
We planned for skeptics and got 1,372 skill files in eleven weeks. Notes on an AI adoption that went too well.6 min
2026.09.28 · The agentic operating model, part 1Every Approval Is a Missing Rule
If the architects do not write the rules the agents read, the person who writes the template does. Notes from reviewing an operating model for agentic delivery.7 min
2026.09.14 · Diagrams, part 3Turn the Sequence Diagram on Its Side
The people who can tell you a boundary is wrong were never going to read UML. Turn the sequence diagram on its side and they do not have to.8 min
2026.09.07 · Reviewability, part 2We Deleted the Wrong Half of Agile
Reviewability is something you design. Extreme Programming already knew how, and we made exactly the wrong half of it optional.8 min
2026.08.24 · Decisions, part 3Design the Stopping Point
Target state is the easy half of a solution design. The hard half is the first step, because that is usually where the money runs out.6 min
2026.08.17Count Your Rejections
A friend passed every round and still did not get the job. Nothing was wrong with him. The company simply had no seat that quarter.6 min
2026.08.10 · Adoption, part 1Adoption Is the Deliverable
The ugliest service in production beats the most elegant one nobody calls. So where exactly does an architect's job end?6 min
2026.08.03 · Reviewability, part 1The Bottleneck Moved
AI made writing code cheap. Our squad reports show where the queue went next: to the people who have to read it.6 min
2026.07.27 · Invent the least, part 2A Feature Is Not a Service
The ticket said "saved filters". The real decisions were which service, whose identity, and how little to invent.5 min
2026.07.20 · Invent the least, part 1The Model Is the Easy Part
I put an MCP server in front of a production API. The hard parts were tokens and state, not the model.6 min
2026.07.13Mathematicians vs. Lawyers
Two thinking archetypes I keep meeting in architecture rooms, and why your system needs both.5 min
2026.07.06Pricing the Invisible
Your best architecture work produces nothing observable. Backlogs only rank observables. Let's fix the mismatch.5 min
2026.06.29Multi-Agent Is an Org Chart
I designed a multi-agent AI system and accidentally rebuilt the org chart, failure modes included.5 min
2026.06.22 · Tools That Scale, part 3Prompt-Driven Architecture
Part three of the "tools that scale" series: what happens when the daily practice of a Solution Architect starts running on AI.5 min
2026.06.01 · Tools That Scale, part 2Probabilistic Metrics
Why the most useful numbers in our AI transformation were never in a database.5 min
2026.05.18 · Tools That Scale, part 1From Signals to Blips
How a tech radar became my favorite tool for driving an AI transformation across several hundred developers.5 min
2026.05.04Grinding Dijkstra at Senior Level
Twelve years into my career, there's a repo on my disk where I practice graph traversals like a student. Here's why that's not humiliating.5 min
2026.04.20A Company of One, Staffed by Agents
Four shipping products, one person, and a field report on what AI actually changes about building software.5 min
2026.04.13 · Diagrams, part 2The Diagram That Cannot Lie
Why I'm building an architecture modeling tool where the diagram is physically incapable of drifting from the code.5 min
2026.03.30Pricing Is Architecture
The pricing page is an architecture diagram wearing marketing clothes.5 min
2026.03.16Anti-Personas: The Users I Refuse to Serve
The most important architecture document in my side project contains no architecture. It's a list of people I will never build for.5 min
2026.03.02Edge Cases Written by Parliament
What integrating with mandatory government systems taught me about specs you cannot refactor.5 min
2026.02.16ADRs for an Audience of One
My hobby project has eleven Architecture Decision Records and zero users. That ratio is correct.5 min
2026.01.26Questions Are an Architect's Sharpest Tool
The artifact of mine that pays off most is not a diagram and not a decision record. It's a list of questions.5 min
2026.01.19The Anti-Corruption Layer Is for People Too
My favorite pattern from Domain-Driven Design protects codebases from foreign models. I keep prescribing it to departments.5 min
2026.01.12 · Decisions, part 2The Migration That Never Starts
Every organization has one. Yours does too. It has a folder, a deck, and a birthday.5 min
2026.01.05 · Diagrams, part 1Diagrams Are Conversations, Not Documentation
The most useful diagram of my career was ugly, partially wrong, and deleted within a week.5 min
2025.12.22Nobody Owns the Sum
Every squad owns its piece. The customer buys the whole. Guess which one has no owner.5 min
2025.12.15 · Decisions, part 1Decisions Have an Architecture Too
We design our systems obsessively, then let the decisions about them happen by loudest voice.5 min
2025.12.01Data-Driven Theater
Everyone wants decisions backed by data. Almost nobody asks what the data went through to get there.5 min
2024.02.19Stop Calling It Legacy Code
With AI in the picture, it's not legacy anymore. It's a Knowledge Container.3 min
2020.07.09The Automerger
How we taught our repositories to merge themselves, and what two years of running it taught us.3 min
2020.05.26Technical Leader: Identity Disorder
Splitting one corporate title into its two words, and what each of them demands.13 min