We start from your real problem, but we never hand over the answer all at once. We break it into steps so that learning happens while you solve. And we ship those principles as real products.
Compete on efficiency, and the AI that throws answers fastest wins. But answers received that way don't stay with the user. The next time the same problem shows up, they have to ask the AI again. We go the other way. While people solve their own problems, they make the field's sense and knowledge their own. So next time, they do it better.
We move toward our vision while actually shipping. In June 2026, we put our first app on the App Store.
Find blurry and closed-eye photos, right on your device.
An on-device AI that runs entirely on your phone finds blurry and closed-eye shots and groups similar ones. Your photos never leave your device.
The AI never deletes for you. You decide what to remove. (Principle 04)
Every friend we build follows the same principle. Don't throw the answer right away. Walk the steps together. When it's done, the user is left with knowledge and judgment.
Instead of receiving results, make statistics your own tool.
It doesn't hand you results. It walks you through a step-by-step learning flow so the statistical workflow stays in your head.
Purpose first, technique second.
“A statistics tool an advisor recommends: 'just learn stats with Research Friend.'”
It doesn't write for you, it makes you write.
It doesn't write on your behalf. It asks questions so learners turn their own thoughts into sentences.
Book reports · essays · reports · personal statements
“A writing AI an educator can endorse: 'if you're going to use AI, use Writing Friend.'”
Build what you want to build, and make coding your own tool.
It doesn't write the code for you. It breaks a student's project into small pieces so they write each one themselves.
Project = motivation. Breakdown = learning unit.
“A coding friend that has students saying 'I can build it myself' by the end of class.”
Competing to throw answers faster belongs to the giant model companies. We go a different way: AI where people learn while finding answers, AI an educator can recommend, AI with learning built in. It's a narrow path we can design more relentlessly.
Led AI teams for six years at an AI startup and a global cosmetics company. With that experience, he designs and ships products that spark learning instead of just handing over answers.
A CS-grounded product and design lead. With a background in information systems audit and usability analysis, she designs learning-companion UX where users perceive and decide at every step.
Information for adopting in the field: universities, schools, academies, and education labs.
Universities, middle and high schools, academies, education labs. Scoped by student count and subject.
Inquiry → align on environment, subject, and audience → a 4 to 6 week pilot → review → full rollout.
Student data is stored and used only under a separate agreement with the institution, limited to the minimum needed for learning support. Retention, access, and deletion follow the institution's requests.
4 to 6 weeks. We review student and teacher response, learning-flow data, and satisfaction together, with a results report at the end.
If your university, school, or academy wants to adopt the friend series, send us a note. If there's a new friend we could build together, we'd love to hear about that too.