Pawsa

Screen-time budgeting, not screen-time bans.

Concept + system design · iOS · In progress

The short version

Most screen-time apps work like telling a heavy smoker to quit cold turkey: strict limits, no learning, and people give up within days. Pawsa treats screen time like a calorie deficit. It reads your real usage, suggests a budget, and steps it down gradually, with a companion to keep you going. I interviewed 10 people, then designed 30 features and mapped every dependency between them before writing app code.

Role: Solo: research, product, system design
Timeline: September 15, 2026 – present
Outcome: 30 locked feature specs and a 387-connection dependency map. No app code yet.

The problem

I was the user. Six hours a day, all doom-scrolling. None of it was for school. Even my "work" scrolling, checking competitors for RivalFed, ended up as Instagram. I was losing focus and pushing schoolwork back until everything piled up.

I tried the usual fixes. With iPhone Screen Time limits, I kept tapping "Ignore limit for today" until I gave up. Brick makes you carry a physical object around, and like Opal, it only blocks apps. It doesn’t help you use your phone less.

Cold turkey fails.

So I tried my own calorie deficit: a fixed 25 minutes less every week, not a percentage.

About 10 weeks later I was down to about 2 hours a day. It worked.

Then I stopped tracking, and it crept back. Today I’m at about 6½ hours, past where I started.

Screenshot of my iPhone Screen Time: 4 hours 26 minutes so far today, with a weekly average line near 6 and a half hours. — My screen time: about 6 hours a day, cut by a fixed 25 minutes a week for 10 weeks to about 2 hours, then untracked, and now about 6½ hours, past where I started. Pawsa automates the cut in 3% steps. Past points are approximate.

Before building anything, I wanted to know if other people were stuck in the same loop.

Listening

These weren’t formal interviews. When a coworker looked tired, I’d ask what happened last night. One answer: "Slept at 3, shift at 9, I was scrolling and lost track of time."

Then we’d pull up our Screen Time side by side. Mine was bad too, so nobody felt guilty showing theirs. Then I’d ask what they’d tried.

I also tried posting on Reddit, but the post was removed.

Interview log

People: 10 coworkers and friends
Method: What happened, not what they’d do
What I asked: What happened last night, then what they’d tried
What I heard: People who’d tried blocker apps quit, especially the ones that need a physical object
What surprised me: One person averaged 12 hours a day, and none of it was work: games and TikTok

The insight

Cutting 25 minutes a week worked for me because I fill the time with soccer, the gym, and walks. Most people don’t have that. So Pawsa does the cutting for you, in 3% steps small enough that you barely notice.

My screen time· by week

+ Pawsa model

Pawsa model · 4 steps of −3%

≈ 5h 45m

↓ 12% from now

avg

≈6hStart

≈2hWk 10

untracked

≈6½hNow

Pawsa: −3% per step · illustrative

Start (≈6h) and low point (≈2h) are from memory; the weeks between follow the 25-minute rule. Time isn’t to scale, untracked weeks have no bars, and the dashed line averages the tracked weeks. Pawsa steps after runs of under-budget days, not weekly.

The companion came from my coworkers. They kept talking about plushies, cats, anything cute. Around the same time, AI apps started giving their tools cute characters. My bet: people stick with things they find cute. So the shield shows Tofu, not a plain block screen. It’s a bet, not a result yet, and novelty wearing off is one of the risks I track.

How Pawsa works

Skip the walkthrough ↓

Concept: built from Pawsa’s specs. The app isn’t built yet.

Step 1 of 5: I start by reading your last two weeks.
WhyThe median ignores binge days; the first budget is 90% of it.
Phone: Example: 14 days of Instagram use, about 2 hours 12 minutes a day; Pawsa suggests a 2 hour budget.
Step 2 of 5: Then it stays quiet.
WhyNo nagging: iOS tracks time; Pawsa steps in only at your line.
Phone: Instagram used 1h 40m of a 2h budget.
Step 3 of 5: At your line, the shield comes up. Turning back is the win.
WhyFriction, not blocks: people who’d tried blocker apps quit them.
Phone: Shield: Limit reached. your call. I'm just a cat. I’ll wait pressed: turned back, day stays clean.
Step 4 of 5: Going through costs a short wait.
WhyOne push-through resets the streak to 0; “I’ll wait” never locks.
Phone: A 15 second wait: 15 seconds left before Let me through. I’ll wait stays available.
Step 5 of 5: Clean days add up. After a week, Tofu asks.
WhyPawsa asks before cutting, and only 3% at a time.
Phone: Seven of 7 clean days. “a whole week under. wanna go lower? no pressure” Sure chosen; budget now 1 hour 55 minutes.

Tofu, modelled and animated in Blender for this portfolio.

Step 1 of 5: I start by reading your last two weeks.
WhyThe median ignores binge days; the first budget is 90% of it.
Step 2 of 5: Then it stays quiet.
WhyNo nagging: iOS tracks time; Pawsa steps in only at your line.
Step 3 of 5: At your line, the shield comes up. Turning back is the win.
WhyFriction, not blocks: people who’d tried blocker apps quit them.
Step 4 of 5: Going through costs a short wait.
WhyOne push-through resets the streak to 0; “I’ll wait” never locks.
Step 5 of 5: Clean days add up. After a week, Tofu asks.
WhyPawsa asks before cutting, and only 3% at a time.

Tofu, modelled and animated in Blender for this portfolio.

Step-down rulesFrom the specs

Starting point: Each app’s baseline is the median of 14 days of real usage, rounded to 5 minutes
When to step: After a run of clean, under-budget days (7 or 14, depending on whether a step was offered before), a companion offers a cut
How much: 3% per step, and the user can say no
Floor: Budgets never go below 15 minutes
Too tight?: Pawsa can suggest raising a budget back toward recent use

An idea that simple still needed a lot of parts to work.

Building the system

Each of the 30 features sits in one layer: data comes in, logic decides, the UI shows it.

30 features · 387 cross-feature connectionsFrom the map’s model

The hubs: Settings Screen (53), Threshold + Shield (52), and Monetization Mechanics (48). Change one of these and a lot of other features change with it.

Data7 features
Logic11 features
UI11 features
Cross-cutting1 features

Show all 30 features

Data

F59Monetization Mechanics48
F25Line Pools30
F22Bibles (Companion personalities)28
F19Dopamine Facts18
F60Rolling Per-App Usage Statistics16
F02Baseline Read13
F05Boring Activities9

Logic

F13Threshold + Shield52
F39Companion Switching40
F24Mood Selection Algorithm36
F34Streak Calculation36
F42Custom Apps33
F23Mood States31
F58Free Trial21
F15Deep-Link Handoff13
F26Step-Down Logic11
F28Step-Up Logic (Simplified)10
F30Manual Budget Override10

UI

F11Settings Screen53
F31Dashboard34
F16Gauntlet UI33
F56Paywall32
F03Suggested Budgets31
F01Companion Selection28
F14Shield UI19
F07Family Controls Authorization17
F06Setup Guide12
F20Re-Arm UX12
F57Subscription Management9

F35Privacy Posture, a rule every other feature has to follow.19

Numbers count each feature’s connections to other features (consumes, triggers, references).

The map below is the one I worked from. Every line is a place where one feature reads, triggers, or points to another.

Pawsa's feature connection map, layered view, top half. Data features (teal) sit in the left column, logic features (violet) in the middle, UI features (amber) on the right, and F35 Privacy Posture (red, uncertain) on its own at the far right. Thin lines run between them for each dependency, densest around F13 Threshold + Shield, F23 Mood States, and F34 Streak Calculation in the logic column. Other features shown include F22 Companion Bibles, F19 Dopamine Facts, F24 Mood Selection Algorithm, F14 Shield UI, F16 Gauntlet UI, and F11 Settings Screen. — Part of the map: the logic and UI layers.The top half of the map’s layered view: data, logic, and UI. 30 features and 387 connections in all.

Thirty features was already a lot. The next job was deciding what could wait.

What I cut

The hardest cut was the companion animations.

People attach to characters. I want a hand-made, slightly wonky character, because nobody’s perfect and a perfect character feels less human. Hand-made animation would cost thousands, so v1 ships one character and the rest move to v2.

In v1

30 locked features, from the 14-day baseline read to the step-down offer
One hand-made companion character
Privacy promises: data stays on the phone, no accounts, no tracking
A paywall with a 7-day free trial

Deferred to v2

The other companion characters and their animations
One character first.
Pre-limit notifications
Partly because asking for notification permission has a cost.
Fetching content remotely
Waits on a privacy change: in v1, the only network traffic is StoreKit (Apple’s in-app purchases).

How I’d know it works

Three numbers would tell me if Pawsa is doing its job. These are targets, not results. There’s no app yet.

Weekly screen-time change
Is usage on watched apps going down, week over week?
TargetAt least 6% below the starting average by week 4 (two 3% steps)
4-week retention
Are people still using Pawsa a month in, or did they quit like they quit blockers?
Target25% of users still active in week 4
Budget adherence
On how many days do people stay under the budget they agreed to?
TargetUnder budget on at least 5 of 7 days

v1 has no analytics or tracking, by design. So I’ll measure the old way: a TestFlight beta with the 10 people I interviewed, plus a weekly side-by-side Screen Time check-in with each of them.

To be continued

Where it stands: The specs and the connection map are done. No app code yet.

What happens next: I’m saving up for one hand-animated character and the Apple Developer Program, then applying for Apple’s Screen Time permission. The target is February 2027, working more hours over winter break.

Will people put the phone down and go outside?