Reading progress of “Agent steps users can follow”: 0%
Course · Shipping AI interfaces
Agent steps users can follow
An agent that works invisibly is a mystery with a cursor. The step rail turns a run into a story — plan, act, verify — and the honest version labels its timings as what they are.
Intermediate1 min read (computed · recorded 3)updated 2026-09-14aiagentsprogresscourse
by Motif Editors
revised 2026-09-14 — First publication in the bank-4 AI-interface curriculum.
— A step rail is a story: each step has a name, a status and a duration, read in order.
— Timings that are demo figures must say so — a fake millisecond is a fake fact.
— Reduced motion completes the steps without stagger; the information survives without the animation.
A run is a story with steps
Agents are opaque by default: input goes in, something happens, output comes out. The step rail opens the middle — plan, act, verify — each step a named row with a status (queued, running, done) and a duration. Read in order, the rail is the story of what the agent did, which is the minimum a user needs before they can trust the result.
Timings are data
A step that claims 'ran in 42 ms' is making a measurement. If the figure is illustrative, the rail says 'demo figure' next to it — this site's tool-call card prints exactly that. And under reduced motion the steps complete without stagger: the sequence is the content, the stagger is the decoration.