Education should be open source.

Khanmigo. Pallo. Geniebook. Magic­School. Duolingo Max. Each of them owns the relationship with the student. Each of them owns the UI, the data, the profile, the learning history. Each of them, when a student leaves, takes the entire learning record with it - and the next platform starts from zero.

This is not a moral failing. It's the structure of a venture-backed consumer product. The investor thesis requires lock-in. The lock-in requires data captivity. The data captivity guarantees that the student's mastery map exists in seven incompatible silos, none of which can talk to each other, none of which compound across a learning lifetime that should naturally span twenty years.

Before Stripe, every e-commerce site built its own payment processor. It was slow, it was buggy, and it was insecure. Stripe didn't make payments. It made the layer underneath, so a thousand teams could.Now Stripe doesn't compete with Shopify or Amazon - it powers both.

Before Auth0, every web app built its own login. Auth0 didn't make logins. It made the layer underneath, so a thousand teams could. The shape of the move is identical: identify the rebuilt-everywhere primitive, build it once well, expose it as a contract, get out of the way.

Before Twilio, every product that wanted to send a text message wrote its own carrier integration. Twilio didn't make conversations. It made the layer underneath, so a thousand teams could. Apply the same pattern to student intelligence - to the portable, queryable, citation-grade record of how a learner actually learns - and you have VOSE.

We don't compete with Verdict. We don't compete with Pallo. We don't compete with Khanmigo. We make it trivial to build a hundred Verdict-quality apps, each owned by the team that built it, each compounding into the same portable identity, each carrying the citation chunks and the mastery vectors that no consumer-facing product would ever expose on its own. The infrastructure layer doesn't take share from the products on top of it. It enables them.

VOSE didn't start as a thesis about infrastructure. It started as a 19-year-old building a study tool for himself. Verdict was the first thing - an AI tutor for Singapore A-Level Economics, built by a private candidate sitting his own papers in November 2026, dogfooded daily, written because what existed for serious H2 Economics study was either generic, kid-coded, or too shallow to engineer real learning out of.

Building Verdict in the open made the structural problem visible. The same primitives - portable identity, citation transparency, mastery vectors, source-tagged retrieval - were going to have to be built again by every team that wanted to make a tutor for a different subject, a different curriculum, a different region. The work was real, the work was substantial, and the work was being repeated everywhere. That's the moment a product becomes infrastructure.

VOSE is what got built underneath Verdict so that anyone else who wanted to ship a serious tutor for any other body of structured knowledge - IB Chemistry, the bar exam, AWS certification, a tutoring centre's in-house curriculum, a textbook with a chatbot in the margin - could start from the layer Verdict had already needed, instead of building it from zero. Verdict is the first surface on the layer. The hundredth surface hasn't been imagined yet.