dp / experimentsAbout this study ↗

INTERACTIVE SYSTEMS / EXPERIMENT 01

A study in emergence.

What if a page could think out loud? Watch a stream of text snapshots become a visual composition.

☿

MERCURY
DIFFUSION RENDERER

Synthetic replayHand-authored frames. Real validation & rendering. No live AI calls.
FRAME 01 / 16
TEXT → STRUCTURE → FORM

M FIG. 01 / A CONSTELLATION OF POSSIBILITIES

FORMING
Ready to replay
01 / 16

THE ENGINEERING IDEA

Text changes.
The frame of reference stays.

Inspired by Mercury’s text-diffusion streaming, this study explores how full-state text updates can drive a non-ASCII visual interface. The stream here is synthetic, not a recording from Mercury or a live model response.

The interesting part is the boundary: validate each snapshot, preserve stable identities, render only allowlisted graphics, and retain the last valid state when a frame fails. No generated HTML is executed. No credentials, login, or external requests are needed.

Read the streaming documentation ↗