Rooted

An app that shows you your weak spots, then helps you work on them one habit at a time.

Client: Independent product
Role: Solo Founder & Designer
Team: Solo
Duration: May 2026 – Present
Platform: Web, iOS

App Store: livePlay Store: closed beta

60 second read · Read the full deep dive →

Problem

Self-improvement advice is generic. I wanted people to see their own weak spots, then get small actions to work on them, without handing very personal answers to an exposed form.

v1 · Form + Claude

Manual, and not private enough

Change.: 16 questions written with a psychologist friend. I ran each person's answers through a Claude skill to write their report, then sent tasks by hand.
What broke.: People told me they worried about privacy: their answers sat in a Google Sheet that could be exposed, and privacy is the point of the app. Every report also meant copying and reformatting sheet data by hand.

Rooted Assessment

What changed?

A typical day

What makes it hard to start?

Submit

Sheet

What changed?

A typical day

What makes it hard to start?

Private answers

Copy rowsReformatPaste into ClaudeSend by hand

Someone fills in the form.

v2 · Web app

Secure and liked, but nobody came back

Change.: A web assessment with row-level security and Clerk auth, so the database and API key couldn't be attacked. People said it helped them understand themselves.
What broke.: No reminders: PWA notifications didn't work. WhatsApp reminders took a long Meta verification, then sent people back to a sign-in. They stopped returning.

Rooted web app: choosing how to be reminded

Rooted · Your task is ready

A WhatsApp reminder arrives
You tap it
Sign in, again
You close the app

v3 · Mobile app

Reminders worked, one task wasn't enough

Change.: A native app with daily notifications and a single task a day. I used it myself and could finally do the tasks.
Why.: I wanted three tasks a day. Research said one coached task beats a habit tracker, so I kept one.
What broke.: One task a day left long gaps. A finance task on day 1 and the next on day 5 gave nothing to build a habit on.

1Home: a path of days, one unlocked
2Today's task: just one
3Done. You showed up
4Back home: every other day is locked
One task a day left long gaps with nothing to do.

v4 · One daily list

One list for the day, not separate pages

Change.: Assessment tasks, goal tasks and habits all show up as the day's activities in one list. Before, habits lived on their own page.
Why.: I opened the task table and found 192 tasks that only asked people to notice something, and none that built a system. I rebuilt the task engine, then added habits under the daily task.
Tradeoff.: It risks reading as a habit tracker, and a long list overwhelms. You start with three habits and have to prioritise to add more.

Rooted app home: today's task, goals and habits in one list

Before: separate pages

Tasks page

Financial Clarity · move Rs 50 aside
Run a 5K · 20 min jog

Habits page

Drink water on waking
10-minute walk
Read 10 pages

Daily taskFinancial Clarity · move Rs 50 aside
GoalRun a 5K · 20 min jog
HabitDrink water on waking
Habit10-minute walk
HabitRead 10 pages

Tasks in one place, habits in another.

How I built it

I ran AI agents like a small team

PRs merged: 690+
tickets raised: 380+
commits: 3,300+

Totals across the web and iOS repos, as of 6 October 2026.

Roles. An orchestrator wrote the plan, a worker built it, and an independent senior reviewed it. I owned product, risk and release.
Two gates. Risky work (auth, payments, database schema) needed senior approval on the plan and again on the code. Green tests alone never counted as approval.
A rule against gold-plating. Reviewers could block only for a loss I couldn't recover by hand. Everything else was a note.
Verify, don't trust. I re-checked every agent's "done". A pre-PR manual check caught a bug that crashed the report for every user.
Where I overruled them. The onboarding focus picker shipped twice and I rejected it ("it flopped"). I redesigned it with ASCII wireframes before any code.

How a change ships

Orchestrator · Writes the plan
Senior · Approves the planPlan approved
Worker · Builds it
Senior · Reviews the codeChanges required
You · Check it yourself, then shipShipped

Green tests alone never counted as approval.

Tools

What earned its place, and what didn’t

One person’s verdicts, from one workflow: solo, spec-driven, shipping a mobile app. Yours may differ.

Core

The build stalls without it

Claude Code. Wrote the PRD, TRD and specs, then built milestone by milestone.
Supabase and Clerk. Names and answers kept apart with row-level security, plus sign-in.
agentmemory. Lets different models check what was done and solved last time.

Helped

Useful, replaceable

BB and Codex. BB ran the agent setup for 2 to 3 months. Codex joined so nothing depends on one tool.
Figma. The last 20%: fixing UI by hand after two failed agent rounds.
Playwright and simulators. Caught what local runs hid, since local didn't always behave like live.
Command Code. Helped a little, mostly as a second option.
Vercel. Hosting for the web app.

Dropped

Tried it, moved on

Wrapping the web app. Not smooth enough, so I rebuilt natively in Expo.
Web push (PWA). Reminders didn't arrive reliably.
Antigravity. Didn't help in my setup.
Kimi. Didn't help in my setup.
Keep-awake tool. Needed until BB made it unnecessary.
Windows gaming PC. Moved to a Mac, though 16 GB is tight with simulators and Docker.

Impact

Each version reached more people than the last: from 10 people filling a form to 75+ using the app. Daily reminders brought people back, and habits are too new to measure.

People reached, by version

75+

Form + Claude

Web app

Mobile app

Mobile app v1.2

5.0 from 5 ratings · App Store, India

★★★★★
Game changer
“I had a personal struggle that I was not able to overcome, fear of not belonging and dash of social anxiety, I took as assessment in Rooted Mind, and it started giving me daily tasks that were pretty simple, 25 days in now, I feel better, I started talking to people around me and it felt easier, I know I can get better, Will be updating my 90 days review here.”
App Store review · Sep 29, 2026 · v1.2.0 · India
★★★★★
excellent app
“have clarity and understanding for better growth”
App Store review · Aug 24, 2026

Want the full story?

The deep dive is the rough, honest version, in the order it happened.

Read the deep dive →