A whiteboard where people and AI work on the same canvas.
A speech-bubble head with two eyes. The tail sits bottom-left, toward the person it is talking to. Outside the product the face is always neutral.
The neutral face reversed out of the brand colour. Corner radius is 23% of the icon size.
Colours come from the active theme. The head is always ink. The AI colour is only ever a detail inside an expression, never the whole mark.
The head and the wordmark.
App icon ground and primary actions.
Small AI details inside expressions.
The Error expression only.
Eyes, and the ground behind the logo.
Outer page ground.
A shared type system across the website and product.
IBM Plex Sans 600. The wordmark. Zero tracking.
Newsreader 300. Editorial lines on the website and in documents.
IBM Plex Mono. Interface text and controls. Martian Mono for counts, timestamps and metadata.
Inside the product, the same face tells you what the AI is doing. The head stays recognisable at every size.
Only the eyes and small details move; the head stays still. Movements are short and never bounce. The splash screen uses the Thinking expression.
Every change of expression happens behind a blink. The eyes close in 120ms, the face changes while they are shut, then they open on the new expression in 180ms. The head never jumps.
“Group these stickies by theme”
Hold for 600ms. Show each expression for at least 600ms so the face never flickers.
Done and Error don’t last. They return to Ready after 2 seconds; the label stays until the next action.
One state, everywhere. Status chip, cursor, avatar and prompt always show the same expression.
The face appears where the AI is present. All live placements follow the same state.
Always visible. Shows the live expression and a short label.
A standard arrow in the AI colour with a white keyline. The face sits in the name tag, never as the pointer.
The Suggestion ready face sits on the card with Accept and Discard.
Past messages use the still Ready face. While working, the send button becomes the live face.
Opening the app wakes the face up. Inside the app the wait time decides what you see.
Under 400ms, show nothing. Long tasks get a progress bar and elapsed time, and can be cancelled. Panels that wait on AI content show placeholder lines next to the Reading face.
Use the logo as drawn.