Building software with agents
How understanding, clear decisions and short feedback loops help us work effectively with coding agents.
Notes on building software
Notes on understanding the work, building useful software, and choosing where AI helps.
11 articles
Building software alongside the people who use it
How observing daily work shapes interfaces, data models and useful product changes.
UX is part of the system
Choose where a capability belongs, then design for clear actions, prompt feedback and interrupted work.
What technical discovery should answer
Investigate the question that could change the build, compare credible options and know when to stop.
Every AI system is still a software system
The interfaces, data, permissions and tests that turn a model into a useful product feature.
When an agent earns its place
Compare adaptive tool selection with simpler designs, then define tools, review and stopping.
Evaluating an AI feature before release
Define the task, compare a baseline and test the failures that matter to the release decision.
When systems use the same word for different things
Separate identities, relationships and source meanings before automating the connection.
Building accessible visual software
Preserve meaning and complete tasks across diagrams, canvases and different ways of interacting.
Choosing an AI provider around your data requirements
Compare model quality, provider commitments and application controls within the actual data path.
Staying with the build
Make support, recovery and knowledge transfer practical parts of looking after software.