Featured · AI & Automation
AI Is Becoming a Layer in Software, Not the Software Itself
The model is the one part of the system whose output you cannot specify in advance. Good AI engineering is mostly about everything built around it.
Insights
Notes from building systems that have to be right.
Essays on software architecture, applied AI, trading technology, product design and learning. How we think about the work, written to be useful whether or not you ever work with us.
Featured article
01 · Categories
Five subjects we know from practice.
- 01Technology & EngineeringArchitecture, delivery and the discipline of building systems that are still working years after launch.2 articles
- 02AI & AutomationApplied AI as an engineering discipline: grounding, evaluation, permissions and where human judgement belongs.2 articles
- 03FinTech & MarketsTrading and market systems, viewed as engineering problems. Technology discussion only, never investment advice.2 articles
- 04Product & Digital SystemsHow software products are actually experienced, and why the whole system, not the screen, is the product.2 articles
- 05Learning & Future of WorkContinuous learning, practice, assessment and the balance between teachers, platforms and intelligent tools.2 articles
02 · Latest
All articles.
Technology & Engineering
Why Good Software Architecture Is Mostly About Decisions
The boxes on an architecture diagram rarely cause trouble. The decisions behind the lines do, and most teams never write them down.
Technology & Engineering
Building Technology That Survives Beyond the First Release
Launch day tells you less about a system than any other day in its life. What matters is whether it can be run, changed and trusted by people who did not build it.
AI & Automation
AI Is Becoming a Layer in Software, Not the Software Itself
The model is the one part of the system whose output you cannot specify in advance. Good AI engineering is mostly about everything built around it.
AI & Automation
The Hard Part of Enterprise AI Is Not the Model
Choosing a model is the most discussed decision in an enterprise AI project and, increasingly, the easiest to reverse. The decisions that determine success are organisational.
FinTech & Markets
Why Trading Technology Is a Systems Problem
A trading platform spends most of its complexity not on deciding what to do, but on knowing what has actually happened. That is a systems problem, and it rewards systems thinking.
FinTech & Markets
From Market Data to Decision: Designing the Modern Trading Workflow
Charts sit at one end of the workflow and order buttons at the other. Most of what determines whether the workflow is disciplined happens in between, and in the review that closes the loop.
Product & Digital Systems
Software Products Are Systems of Experience
A user's opinion of a product is formed as much by a slow search, a confusing permission error or a late email as by anything drawn in a design file.
Product & Digital Systems
When a Product Needs Fewer Features
Every feature is paid for twice: once when it is built, and then indefinitely, mostly by people who never asked for it. Mature products get better by subtraction more often than their roadmaps admit.
Learning & Future of Work
What Changes When Learning Becomes Continuous?
A degree was designed to load knowledge before a career began. Careers now keep asking for more. That changes what learning is for, how it is assessed and what teachers are needed for.
Learning & Future of Work
The New Learning Stack: People, Platforms and Intelligent Tools
Institutions tend to buy a platform to solve a teaching problem, or an AI tool to solve a content problem. The learning stack works when each layer does its own job and the layers share what they know.
Want to talk about the systems behind these ideas?
The articles describe how we think. If you are building something where the same questions apply, we would be glad to hear about it.