Eight Dimensions of AI-Native Design
The cleanest break with the old frame: components become conversations, states become spectrums. My demos stream and blend instead of flipping for exactly this reason.
lives at /sources ↗ source ↗POSITIONS · AI-NATIVE DESIGN
What I believe,
argued next to the work.
Not an essay wall - every position here is already published on this site, next to the reading or artifact it argues for. This page gathers them in one scan. The quotes are the notes as published; the links go to where they live.
01
Beyond the chatbox
The interface is not a conversation with a text box bolted on. States become spectrums, surfaces assemble themselves, and what stays fixed is grammar, not layout.
Designing Generative UI in an Agent-Native World
Generative UI is mainstream now. The design question is what stays fixed when the surface assembles itself - and the answer is grammar, not layout.
lives at /sources ↗ source ↗Smashing argues the button is dying
Smashing Magazine makes the case that the point-and-click contract is ending: intent-driven design collapses the ten-click flow into one stated goal, and the designer’s job shifts from drawing screens to guiding transparent agent behaviour. A mainstream design publication saying the best interface is no interface moves the idea from niche to canon.
lives at /fresh ↗ source ↗OUI-1 puts generative UI on consumer GPUs
Thesys ships OUI-1, an open-weight DiffusionGemma finetune that writes interfaces in OpenUI Lang and scores 71.7% on the Generative UI Benchmark, 5.5x its base model. Diffusion decoding writes 256-token blocks at once, so interfaces stream in under a second on an RTX 5090. Generative UI stops needing a datacenter.
lives at /fresh ↗ source ↗Agent to UI: Designing Interfaces for AI Agents
An agent needs to know what a thing is, not what it looks like. The semantic contract under the pixels - most sites, mine included, are still catching up.
lives at /sources ↗ source ↗Dual-Native Design for the AI Future
Intuitive for people, structured for machines. Automated traffic has passed human traffic; be legible to both or illegible to one.
lives at /sources ↗ source ↗02
Non-determinism, latency and failure
The model will be slow, wrong and confident - sometimes in the same answer. The interface carries that honestly: reasoning visible, uncertainty admitted, evidence attached.
People + AI Guidebook
The textbook. User needs, mental models, trust - less fashionable than the agent discourse and more useful when you actually ship.
lives at /sources ↗ source ↗AI UX Patterns for AI-Native Products
A pattern playbook with the right spine: show the reasoning, admit the uncertainty, hand control back. Closest to how I grade my own demos.
lives at /sources ↗ source ↗Honest motion is a research trail
Most streaming UI fakes liveness with shimmer. BoardUI’s web search draws each source as it opens - the motion is the evidence arriving, not decoration over a wait. That is the difference between animation as progress and animation as theater.
lives at /atlas ↗ source ↗How we built Linear Agent
Linear designed boundaries instead of paths: system prompt, tool design, run scope and the harness underneath are where the behaviour is shaped. Script it too tightly and you dilute the flexibility that makes an agent useful - the agent-design trade in one line.
lives at /sources ↗ source ↗03
Human agency over autopilot
Automation is a dial, not a destination. The design work is deciding where the person steers, approves, interrupts and recovers - then building exactly those surfaces.
The Agent is the Interface
Users evaluate an agent the way they evaluate a colleague - reliable, restrained, honest about what it did. Restraint as a design material. That one idea rewires a roadmap.
lives at /sources ↗ source ↗UX Patterns for Agentic AI: 16 Essentials
Sixteen patterns across delegate, steer, approve, interrupt, recover. The control-surface checklist alone is worth the read.
lives at /sources ↗ source ↗Discover
AIAI clusters research and retrieves patterns.
MEI verify sources, separate signal from repetition, and decide what is trustworthy.
Frame
AIAI generates competing problem frames.
MEI choose the frame that connects user behavior to the business constraint.
Explore
AIAI widens the option space.
MEI set principles, reject weak patterns and select a direction deliberately.
Prototype
AIAI accelerates variants and working prototypes.
MEI decide what the prototype must prove and preserve interaction quality.
Validate
AIAI helps synthesize evidence.
MEI inspect contradictions, resist false certainty and own the final recommendation.
04
Prototypes over screens
A static screen of an AI product proves nothing - the behaviour is the product. So the evidence on this site runs.
05
The stance as a system
The positions hold together as a map and a ruleset, not a pile of takes.