Portreto
A self-reflection app built on personal psychological portraits.
Why I built it
Horoscope apps give everyone the same vague text. I wanted the opposite: an app that feels personal and turns into a daily habit, with a morning question matched to your personality type and an evening look back at the day. People are anxious, and a few minutes of reflection a day can help.
For me it was also a full product challenge, built alone: AI writing good enough to pay for, a choice of business model, content people want to share and reasons to come back every day.
How it works
- Enter a few personal details and get your psychological portrait.
- Each morning, read your tip and answer a question based on your portrait.
- Each evening, answer a personal question that makes you reflect on your day.
- Over time, your notes show what really drives you.
- Get psychological portraits of people close to you and see how you fit together: why you keep having the same fight and how to work it out.
Screenshots
Features
Portraits and relationships
- Your psychological portrait: your character and the main areas of your life, from relationships and work to health and money.
- Psychological portraits of people close to you: a partner, parent, child, friend or colleague.
- A relationship report for nine types of relationship: how you fit together, why you keep having the same fight and how to work it out.
Today
- A tip of the day and a morning question based on your portrait.
- An evening question about your day.
- A monthly focus: the main theme for you this month.
- Question series on themes: failures, money, self-care, calm, relationships, work, blind spots and more.
- A journal of your notes, which stays readable even if you cancel.
Practice
- 22 meditations of about 15 minutes.
- Choose by how you feel, such as "Ready, but can't start".
Monetization
- Free forever: your portrait, one portrait of someone else, one relationship report, the monthly focus.
- Seven days of full access, then a subscription for the daily routine, series and meditations.
- One-off purchases for more portraits and reports.
- No ads.
Sharing, feedback and data
- Share cards from your portrait.
- Rate each part of your portrait: "Spot on" or "Not really".
- Export or delete your data at any time.
Key decisions
- First version scope: value, virality, retention. Without any one of the three, an app like this won't take off, so each had to be covered. The portrait gives instant value, share cards and the relationship report spread the app, and the daily routine brings people back. Everything else waited, including detailed texts on each trait and advice for specific situations.
- Invest in content up front. Each portrait is written for a personality type, and the number of types is limited, so I generated thousands in advance. People get their portrait right away, and it costs less than writing on demand. The risk: a couple of thousand dollars in AI costs per language, spent before the first user arrived. Relationship reports are the exception: there are millions of possible pairs, so they are generated on request.
- Balancing quality and cost. I tested many pipelines and combinations of LLM models on real portraits. The strongest model wrote best but cost far more, so I took a cheaper one and raised quality through the process. Several different models judge the text, because each one catches different problems, such as weak passages or lines that could fit anyone. A stronger model then edits the finished library, and people proofread it. I read samples at each stage and made the final call.
- Free portrait, paid habit, no ads. The portrait is free because it shows the value at once. The daily routine needs a subscription after seven days, and reports on other people are one-off purchases. I rejected ads: they wouldn't bring real money at launch and don't fit such a personal product. People's notes stay readable after they cancel, because locking them would destroy trust.
What's next
First, a store launch in one country to test the app on real users. I'll track three things: do people feel the portrait is about them, based on their "Spot on" ratings, do they come back, and do they pay? Then I'll fine-tune the app. Each new language is a real investment, so the worldwide rollout comes only if the numbers look good.
Tech
- App
- Expo, React Native, TypeScript
- Backend
- Render
- Database
- Supabase, with sign-in
- Payments
- RevenueCat
- AI texts
- Claude
- Analytics
- PostHog, Sentry
- Audio
- Cloudflare R2