Articles
Insights
Notes on AI systems, product thinking and design – from real projects, not from the textbook.
-
UX
When the model is wrong: designing errors as a state
The notice “AI can make mistakes” offers no help when an error actually occurs. What the guidelines from Microsoft and Google provide for the error case and what follows from them for design.
Read more → -
Interaction design
Five patterns for AI interfaces that already exist
Interfaces for generative AI can build on existing patterns. Five patterns that are already described in published design systems and well-known products.
Read more → -
Design Systems
Prototypes from your own system: what Atlassian reports
Atlassian has designers and product managers build prototypes in its own design system with AI and has openly described the problems along the way. What is transferable and what is not.
Read more → -
Governance
Who checks the agents' code?
Which design system rules can be checked automatically, how an agent works with error messages, and what can be learned from automatic merging of dependencies.
Read more → -
Design Tokens
Tokens instead of guessed color values
Hard-coded color values make a design system drift apart. How a three-tier token hierarchy and Tailwind v4 make guessing harder, with a verified example.
Read more → -
Design Systems
Contracts instead of essays: describing components in a machine-readable way
Prose leaves a language model room for interpretation, which a component contract with fixed values, purpose, and prohibitions rules out. Plus the question of how an agent accesses this description selectively.
Read more → -
Design Systems
Three design systems that have opened up to agents
Teams at Indeed, Atlassian, and IBM have described publicly how they make their design system readable for AI agents. What they built and measured, and which conclusions cannot be drawn from it.
Read more → -
Design Systems
Why coding agents fail at the design system
When coding agents fail at a design system, the cause often lies in knowledge that isn't written down in any document. What a design system needs to capture so an agent doesn't have to guess.
Read more → -
AI architecture
Nine building rules for an AI pipeline at a publisher
How does a publisher make sure an AI pipeline publishes nothing it cannot take responsibility for? Nine building rules from a system in day-to-day publishing, with a checklist for choosing a vendor.
Read more → -
Multi-agent systems
What building a multi-agent system requires
Seven specialized AI agents for a publisher's content production. Why the architecture was decisive, more so than the design of the prompts.
Read more →