unfold

HYPERBLOOM OCTOBER / DESIGN CASE STUDY

The list is long.
The next step should not be.

Unfold is a capacity-first planner for people managing everyday tasks around limited time and attention.

The design question

What if planning began with what you have available, rather than everything you still need to do? Our working hypothesis is that a bounded plan with clear reasons is easier to act on than an unbounded list. This has not yet been tested with recruited users.

Research into design

Show where you are.

Three numbered stages remain visible: gather, choose, begin. Each stage has a clear heading and a route back.

Give one instruction at a time.

Focus mode shows one task and its small steps. The broader list stays available through a named control.

Explain the outcome.

Excluded tasks remain visible with a reason. Effort limits and time estimates come from the user.

Keep mistakes reversible.

Undo for task changes, confirmation before replacing data, validated imports, and honest storage feedback.

These choices draw on W3C cognitive accessibility guidance: Make Each Step Clear, Use Clear Step-by-step Instructions, and Help Users Focus. Applying guidance is not proof of usability or WCAG conformance.

One complete user journey

Gather tasks → choose time and effort → review what fits → focus on one task → complete or undo.

Honest boundaries

Completed task estimates consume the current time window. Selecting a time button starts a fresh window. The planner uses a deterministic rule: time-sensitive tasks first, then lighter effort, preserving entry order for ties. It chooses whole tasks that fit, rather than estimating partial completion. It does not promise the optimal schedule, sync devices, send reminders, or offer health advice.

Task text stays in browser storage. Storage is not encrypted, and clearing browser data removes it. The app offers private JSON backups. Step checkboxes are session-only. The hosting provider still handles normal web requests.

Inclusive by design

Semantic controls, native dialogs, visible keyboard focus, reduced motion support, optional larger task text, optional higher contrast, and responsive layouts. No drag-only interactions. No forced sign-up, streaks, timers, or invented testimonials.

Next validation

Recruit 5–8 participants, including keyboard and screen-reader users. Ask them to add a task, make a thirty-minute plan, explain why a task was left out, undo a completion, and recover a backup. Record completion, errors, and their own explanation of the plan. Future work, not claimed results.

Team responsibilities

Proposed ownership: Adeel Falak Sher, product direction, frontend and release review. Rauf Khalid, interaction design, accessibility review and demo storytelling. These are proposed responsibilities, not a claim that either teammate performed work they have not reviewed.

Try the working experience ↗