One calm step at a time — a 4-minute walkthrough, narrated by a calm English voice paced line-by-line to the action. The words are also burned in as small captions, so deaf and hard-of-hearing viewers get the same story — nothing depends on sound alone. An optional Simplified Chinese caption track sits behind the CC button. There is no music on purpose: the voice and the words carry it.
Your browser can’t play this video. The raw file is at
stepwise-demo.mp4 .
4:04 · 1280×720 · H.264 · English voiceover · small burned-in captions · optional 中文 track via the CC button.
What you’ll see
The wall a normal to-do app builds — then Stepwise: kind breakdowns with buffered estimates, sessions of at most four steps, the one-step focus screen with a quiet text timer, the trail of wins, and the accessibility panel (five text sizes, high contrast, dyslexia-friendly text, reduced motion), full keyboard flow, and the PWA on a phone.
ADHD-friendly
dyslexia-friendly
screen-reader tested patterns
offline PWA
zero dependencies
no tracking, ever
Transcript
The full transcript (English) is also downloadable as a caption file:
stepwise-demo.en.vtt
(Chinese: stepwise-demo.zh.vtt ).
Show transcript
This is a normal to-do list. For an ADHD, dyslexic, or anxious brain, it isn’t a plan — it’s a wall.
Stepwise is our answer — built for the TaskForge hackathon, whose theme is accessible tools for focus and calm productivity.
The home screen refuses to show you a wall: your recent tasks, one input, and calm copy.
Deciding can wait. Dump a task in exactly as it is — sorting it out is the app’s job, not yours.
Let’s add a real one: “Reply to Prof. Reyes about the thesis draft.”
Stepwise breaks it into small, concrete steps — with kind time estimates: a 35% buffer, because time blindness is real.
Steps are chunked into sessions of at most four, with breaks in between. Every list stays short enough to breathe.
And the last step is always a reward. Finishing should feel like finishing.
Focus mode shows exactly one step. Everything else hides behind a peek — available, never pushed.
The ring timer is quiet by design. Time is always shown as text — never color alone.
It can be paused. And when time runs low, five kind minutes appear — no questions asked.
But you never have to wait for a timer. Press Done the moment it feels done.
The next step never auto-plays. You set the pace.
Every win gets a breath, not a slot machine.
Or skip the step. Skipping is never failure — momentum matters more than perfection.
One more — and the session ends.
Sessions end with a built-in break, because rest is part of the work.
Back on the task, progress is quiet: no confetti pressure, no red badges.
This is the trail of wins: one stone per finished step, lanterns every fifth.
Stats count steps, not streaks. There is no shame anywhere in this app.
Every stone is named and keyboard-focusable — screen readers get the same story.
Your data is yours: export or import as JSON, any time. No accounts, no cloud, no tracking.
The accessibility panel changes everything instantly — and remembers it, locally.
Five text sizes: the whole UI scales, not just paragraphs.
High contrast: black, white and gold, tested beyond AA.
Dyslexia-friendly text: rounder letters, wide spacing, a calmer rhythm.
Reduce motion, for days when even gentle animation is too much.
And back to default in seconds. Your settings, your device.
Fully keyboard-driven: N for a new task, Space runs the timer, Esc always lets you out.
A polite live region announces every step change and timer milestone.
It’s a PWA: installable and fully offline. Your list works on a plane, or at 2 a.m. with no signal.
On a phone, the same calm: one column, big targets, nothing to hunt for.
Under the hood: readable rules, not AI. 24 passing tests, zero dependencies.
Built during the hackathon, with the AI assistance disclosed — as the rules ask.
Stepwise — one calm step at a time. The link is right below this video.
Links
Built during the TaskForge hackathon. AI assistance (GLM, by Z.ai) is disclosed — as the rules ask.