Buttercup

The starting question

People keep learning long after school ends, but the tools around them assume they haven't. Courses assume a fixed curriculum. Note-taking apps organize information without teaching anything. Chat-based assistants answer a question well but forget you the moment you close the tab. Nothing held the shape of what someone was curious about across weeks or months.

I wanted to design for the person who has ten half-formed interests and no structure to hold them — not by giving them a syllabus, but by helping the structure emerge from what they were already doing.

Research

I spent around 150 hours on user research: interviews, persona-building, and testing early concepts with a small sample across the US, UK, EU, and China. Three kinds of people kept showing up.

Lifelong learners with more curiosity than organization. One, a medical student studying Arabic on the side to read poetry in the original, put it simply: “I wish I started getting into my interests sooner.” The problem wasn't motivation. It was permission — a sense that scattered interests needed to add up to something before they counted.

People mid-transition, usually career changers, who wanted structure but got overwhelmed by the number of directions available. One described it as needing “a feeling of where I could get to, of being good at something” before she could even start.

Builders with too many project ideas and a habit of stalling on the first one. What they needed wasn't more ambition. It was permission to break a big idea into a step small enough to start today, without treating the smallness as a failure of vision.

What I built

Three connected surfaces, plus a model underneath them.

Cosmos — a knowledge map that grew as someone learned, so progress looked like a constellation forming rather than a percentage bar filling up.

Glia HQ — a home screen built around the single thread most relevant that day, instead of a dashboard trying to hold everything at once.

Rabbit Hole — a place for ideas that hadn't found a shape yet. Half-formed questions. Tangents worth remembering but not yet worth acting on.

Underneath all three sat a five-level model — Explorer, Enthusiast, Devotee, Virtuoso, Trailblazer — meant to describe how casual curiosity turns into depth, and a “slow thinking” AI layer that responded to what someone was actually learning instead of just answering isolated questions about it.

Where it landed

The concept traveled further than I expected for something built solo and tested with a small beta group. One user used it to plan a shift into data science. Another used it to structure a move toward fluency in Spanish before relocating. We ran a handful of in-person workshops off the back of it — “Design Personal Growth” and “Mapping Your World,” both built with collaborators and run out of a design academy in China — where people used the framework on paper before ever opening the app.

But the case that stayed with me longest didn't fit the “lifelong learning” frame at all. A user was recording conversations with her mother — not to study anything, but so the two of them could understand each other better through what they remembered together. It kept surfacing in interviews and beta feedback, even though nothing in the product was built for it.

That thread is the one that eventually became Glia in its current form: a place to keep the record of a relationship, not a curriculum.

What I'd do differently

The five-level model was more scaffolding than most people needed. A habit like this benefits more from removing friction than adding structure to climb.

I also over-invested early in market sizing and B2B revenue projections — the kind of forecasting that looks rigorous in a pitch deck and means very little before you have real usage data to build from. It was time better spent talking to the fifteenth user than modeling the fifteen-thousandth.