Visible intent
Show what the system plans to do before consequential work begins.
Code:Emotion Design
We design AI that organizations can trust, inspect, and stand behind.
Interfaces, governance gates, and decision points that keep a human in control of the system.
For thirty years, Code:Emotion has built identity, interface, and editorial systems where design had to do more than decorate. Today the hardest version of that problem is AI: software that recommends, drafts, and acts, often without a clear way for people to see, check, stop, or reverse it.
The same discipline now underpins Arthur's native applications and practical AI teaching: design first, implementation with AI assistance, and human accountability throughout.
Human control, made visible
AI product UX
Governance interfaces
Decision systems
What we design
The human-in-the-middle layer of AI products
Most AI products emphasize what the model can do. We design what the people around it need in order to understand, approve, correct, and remain accountable for the result.
Human decision gates
Place approval, modification, escalation, and refusal at the moments that matter.
Inspectable outcomes
Preserve the reasoning, actions, ownership, and reversal path behind each result.
HITM in action
Three case studies in judgment-preserving interface design
These studies show the method in application. Fully working prototypes are the next stage; the current work should be read as design evidence, not as a claim of validated operational outcomes.
01
Transparent automation
GLASSBOX
An automation interface built around visible intent, live execution, consequence gates, audit, and undo instead of opaque agent magic.
Full studyPrototype forthcoming
02
Crisis-state interaction
Panic Defense
A low-cognition mobile flow designed around human choice, motor imprecision, minimal reading, and a clean exit rather than an engagement loop.
Full studyPrototype forthcoming
03
Human-owned completion
DONE
A case study in making completion, review, ownership, and the final human decision legible inside an AI-supported workflow.
Case studyPrototype forthcoming
One discipline, three expressions
Design becomes working software — and teachable practice
Code:Emotion supplies the product thinking, interface craft, visual systems, and judgment. Augmented Arthur: Applications turns that discipline into native software built with AI assistance and tested in daily use.
Human in the Middle makes the method explicit: what remains human-owned, where review belongs, and who is responsible. Arthur's AI Lab and HITM Labs then make those lessons useful to other people.
Code:Emotion contributes
Design authority
Concept, product logic, interface systems, visual judgment, and interaction models.
AAA proves
The work survives use
Running applications, real data, bugs, corrections, release gates, and named responsibility.
HITM teaches
The method transfers
Tutorials, labs, workshops, and guided decisions grounded in current practice.
Established studio
Thirty years of design craft, pointed at the hardest problem in software.
The studio still works across identity, interface, editorial systems, and unusual hybrid briefs. The commercial focus is now where that experience is most scarce: AI products that organizations need to trust, inspect, and stand behind.