All work

Side project2026

Guilt-free AI daily planner

Goalify turns a messy thought into small, timed steps and never punishes a missed day. I took it from Figma to the app store in 2 months, using the Figma MCP and Claude Code to build it on the design system, screen by screen.

View live site (opens in a new tab)
Role
Product design, AI-assisted build
Year
2026
Focus
Figma to production with Claude Code
Status
Live
Timeline
2 months, Figma to app store
Contents

Overview

What Goalify does

You tell Cubi, Goalify’s AI coach, what’s on your mind. Cubi plans it as small steps with times, and you approve what stays.

I led the product from research to release: the design in Figma, then the build with Claude Code. Goalify is live at goalify.dev.

The problem

Why people give up on planners

  • Starting is the hard part. An empty list gives no first step.
  • Time blindness. Tasks have no estimates, so a morning plan falls apart by lunch.
  • Punishing streaks. One missed day resets weeks of progress, and people stop opening the app.

Competitors

How Goalify compares

How Goalify compares with popular planners
FeatureGoalifyTodoistTickTickStructuredTiimoMotion
AI breaks a goal into stepsYesYesNoYesYesYes
A time estimate on every stepYesNoNoYesYesYes
You approve before AI adds anythingYesNoNoNoNoNo
No lost streaks or "overdue" pressureYesNoNoNoYesNo
Focus timer built inYesNoYesNoYesNo
Tasks, habits and calendar in one appYesNoYesNoYesNo
Designed with neurodivergent people in mindYesNoNoNoYesNo

Based on each app's public features at the time of research.

The opportunity: an AI that does the breaking down, a person who stays in control, and a design that never shames a slow day.

Design decisions

Four decisions behind the design

  1. AI plans, you approve. Cubi never adds a task on its own, so people stay in control and trust the AI.
  2. Small steps, each with a time. One move answers both “where do I start” and time blindness.
  3. Guilt-free by design. No red counts, no “overdue” and no lost streaks. Unfinished tasks quietly roll over.
  4. Accessible by construction. Contrast, 44 pt tap targets, screen reader labels, Reduce Motion and 130% text are checked on every screen.

Try the core loop:

Say it

Talk or type to Cubi the way you would tell a friend. One task or a whole messy list is fine.

Cubi chat: the user asks for a meditation habit every morning at 10am and Cubi plans a task for approval.

Cubi splits it

Cubi breaks the goal into small steps, each with a time. Nothing is added until you approve it.

Cubi breaks "Meditate for a few minutes" into three sub-tasks, with Reject and Approve buttons.

Do it

Start the focus timer, tick the task off when it ends, and keep your streak going.

Focus timer showing 30:00 for a meeting, round 1 of 4, with 15, 30, 45 and 2 minute presets.

Design to code

From Figma to the app store in 2 months

  1. Problem framing

    Research

    Mapped why people abandon planners and compared five popular apps feature by feature.

  2. Figma

    Prototype in Figma

    Explored the core flows as interactive Figma prototypes and tested them before any code existed.

  3. Figma variables

    Build the design system

    Defined colours, type, spacing and components as Figma variables and reusable components, mirrored one to one by tokens in code.

  4. Figma MCP · Claude Code

    Hand the design to Claude Code

    Used the Figma MCP to give Claude Code each frame's exact layout, variables and components, so it built screens from the existing system instead of guessing.

  5. Quality gate

    Review against the design

    Checked every screen side by side with its Figma frame in light and dark, against a fidelity checklist and automated contrast tests.

  6. Release

    Ship one screen at a time

    Each screen went out as its own pull request with before and after screenshots, so the app never drifted from the design.

AI did the typing; the design system kept every screen consistent.

Get in touch

Open to new product design roles. Send a note and I'll reply within a day or two.