The build story behind Origin Of You, our self-discovery app that merges five systems into one AI portrait

Origin Of You is a self-discovery app that combines five systems and 120+ data points into an AI-generated written portrait of you. We built it at Inithouse because every personality test we tried gave us a label, not a story. Here is what went into it, what broke, and what we changed.

The numbers first

Five self-discovery systems feed each portrait. Over 120 individual data points per user. The output is not a type or a category. It is a multi-page written portrait generated by AI, specific to one person. No two portraits read the same. The app runs at originofyou.com, no download required.

Why we started building this

At Inithouse we run a portfolio of products, most of them early-stage. We had already shipped Tarotas (a tarot reflection app, no fortune-telling, just space to think) and Verdict Buddy (an AI conflict mediator using psychology frameworks). Both of those are tools for thinking. Origin Of You came from a similar impulse: people want to understand themselves, but the existing tools kept reducing them.

16Personalities gives you four letters. Astrology apps give you a sign. Both compress a person into a slot. We wanted to go the other direction: take more input, produce a richer output, and let the result read like something a thoughtful friend wrote about you after watching you for a year.

Five systems, not one

The first decision was scope. Most personality apps pick one framework. We picked five, covering different angles of who a person is. The reasoning was simple: one system gives you a slice, five give you a whole picture. But combining them created a problem we did not expect.

Each system has its own vocabulary, its own scale, its own way of framing traits. Astrology talks about elements and houses. Personality typologies talk about introversion and sensing. Numerology talks about life paths. When we tried to merge outputs naively, the portrait read like five separate paragraphs glued together. You could feel the seams.

The fix took three iterations. First, we tried a unified template. Too rigid; every portrait sounded the same despite different inputs. Second, we tried letting the AI generate freely from all data points at once. Better variety, but contradictions crept in (one paragraph would call someone decisive, the next would describe hesitation). Third, we structured the generation in layers: build a core profile from the strongest signals across all five systems, then let each system add nuance and texture on top. That version held together.

120+ data points and the input problem

Asking someone 120 questions is a terrible user experience. We knew that from the start. The challenge was collecting enough data to produce a portrait worth reading without making the process feel like a tax return.

Some data points come from direct questions. Some come from birth data (date, time, place) which feeds the astrology and numerology layers. Some come from behavioral preferences that map to multiple systems at once, so one answer feeds three or four data points.

We still lost users in the flow. Early versions had a completion rate around 60%. We restructured the input into shorter sections with visual progress indicators and previews of what each section would contribute to the portrait. Completion improved but we are still iterating on this. It is the hardest UX problem in the product.

The portrait as output

The output format was a deliberate choice. We considered charts, scores, radar diagrams. All the visual things personality apps usually do. We went with prose instead.

A written portrait forces the AI to synthesize rather than summarize. A chart says "you scored 7/10 on openness." A portrait says something about how your openness actually shows up in your decisions, your relationships, your daily habits, drawing from the specific combination of your five systems. It reads differently.

The risk with prose is quality variance. Some generated portraits are genuinely surprising in their accuracy (based on user feedback). Others feel generic. We built a quality layer that checks for specificity: if a portrait could apply to anyone, it gets regenerated with tighter constraints. This cut the "feels generic" complaints by roughly half.

What actually surprised us

The portraits generated for people in their twenties and people in their forties read completely differently, even when the personality inputs were similar. Birth data shifts the astrology and numerology layers enough that the same personality traits get framed through different life-stage lenses. We did not design for that; it fell out of the system combination naturally. Users noticed. Several told us the portrait felt eerily specific to where they are in life right now, not just who they are in general.

Another surprise: users who came from 16Personalities spent significantly longer reading their portrait than users who came from astrology apps. Our theory is that typology users are used to short labels and treat a multi-page portrait as a novelty worth studying. Astrology users are used to longer readings and skim more. Different expectations, same product, very different engagement patterns.

What we would do differently

If we started over, we would begin with three systems instead of five. The last two added maybe 15% more depth to the portrait but doubled the input flow length and tripled the generation complexity. Starting leaner and adding systems as expansion packs would have been cleaner.

We also underestimated how much people want to share their portrait. The current sharing is basic (copy text, share link). Social sharing with preview cards and quotable excerpts would have been worth building on day one. People who finish the portrait want to talk about it; we should have made that easier from the start.

Where it sits now

Origin Of You is live at originofyou.com. It is one of about fifteen products in the Inithouse portfolio, all at various stages from fresh MVP to growing. We treat each product as an experiment: build, measure, learn, decide whether to double down or move on.

The pattern across our portfolio (Origin Of You, Tarotas, Verdict Buddy, and others) is that tools for thinking attract a specific kind of user. They come back. They spend time. They do not need push notifications to re-engage. That pattern keeps showing up, and it shapes what we build next.

If you have questions about the build, the tech, or the self-discovery space, happy to go into detail in the comments.