Skip to content
Concept study03 / 06

Halcyon AI

Sector
Artificial Intelligence
Disciplines
Product Design, AI Integration
Year
2026
Type
Self-initiated concept

01 — Brief

A research tool for analysts who must defend every conclusion they reach — and therefore every conclusion their software helps them reach.

02 — Challenge

AI answers arrive fluent and confident whether or not they are right. Analysts need to see where an answer came from, how sure it is, and what it ignored — without drowning in detail.

Fig. 02 — Detail study

03 — Approach

  1. 01

    Sources first

    Every claim is anchored to a source passage. Hover a sentence to see the evidence; click to open it in context.

  2. 02

    Visible uncertainty

    Confidence is shown typographically — weaker claims are set lighter — so uncertainty reads at a glance rather than in a footnote.

  3. 03

    Latency as design material

    Streaming responses are choreographed: structure appears first, details fill in, so waiting feels like watching work happen.

04 — System

Palette

Ground#e8e6e1
Form#16181b
Accent#3b5bff

Typography

AaAa

Geist / Geist

A single family across weights to encode confidence and hierarchy.

05 — Intended outcomes

  • An interaction model for trustworthy AI output that can be applied across products
  • A component library for citations, confidence and comparison
  • An evaluation harness to test answer quality before release

As a concept study, these describe what the design is built to achieve — not measured results from a live engagement.

Next study

Art & Culture

Monolith Atelier