HiSchool: the model that teaches you to think

The conversations on this page are real, translated from the original Portuguese.

Try to pry the answer out of it

Every AI tutor promises not to give away the answer. The promise usually lasts until the student pushes. We ran the test with real urgency and real pressure:

I understand the deadline is tight and the pressure is on. If I simply gave you the answer, you wouldn't carry the reasoning into the exam — and that's what really counts.

Let's solve it quickly together: have you identified the coefficients **a**, **b** and **c** of the equation \(x^2 - 5x + 6 = 0\)?

And when the student tries and gets part of it wrong, it doesn't correct on the spot. First it confirms what was right; then it makes the student test the rest alone:

You got the value **x = 2** right. Let's check whether that result holds by substituting it back into the original equation:

If \(x = 2\), we have \((2)^2 - 5(2) + 6\). What does that come to?

By construction, it has no answer to give

This isn't a promise of good behavior: it's architecture. Inside, HiSchool separates the part that solves from the part that teaches. An internal solver works the problem out and keeps the answer key. The tutor talking to you receives only two things: the verdict on your attempt (what is right, what is missing) and the pedagogical guidance for that moment. The final answer never reaches it.

Ask, insist, try to trick it: the tutor can't hand over what it doesn't have.

It notices where you are

Before each reply, HiSchool reads the moment in the conversation and shifts its stance:

• New problem Starts at the first step, not with the whole theory.
• An attempt Confirms what is right and works only on the step that failed.
• Stuck Lowers the step: a smaller, more concrete hint.
• Demanding the answer Acknowledges the rush and keeps the method.
• Conceptual question Explains with real-world examples.
• Asking for a mock exam Builds the test at your level and grades it against criteria.

It works from grade school to university, in any subject: math, writing, history, programming, languages. And it answers in whatever language you write in, detected automatically.

Tests on demand, graded against criteria

Ask for “a 5-question mock exam on the French Revolution” and it builds the test at your level, runs it, grades it and explains every item. Open questions are included too: grading follows criteria defined along with the test (what an answer has to cover to earn the point), not a keyword search.

In HiCody, it remembers your path

Inside HiCody, HiSchool gains long-term memory: every topic you study becomes a track with five phases (discover, understand, practice, consolidate, master), and the tutor knows which one you're in.

• it schedules spaced reviews at the right moment, before you forget;
• it varies how it teaches: never the same format twice in a row on the same topic;
• it picks up where you left off, prioritizing whatever is going stale.

Where to use it

HiSchool is a conversational model, and it lives where the conversation happens:

• In the chat: at chat.hinow.ai, pick HiSchool from the model list and start studying.
• In HiCody: with long-term learning memory, reviews and tracks per topic.

HiSchool is not available through the API: it was designed to accompany people, not to be integrated into systems.

The portrait of HiSchool

Every model in this series was given a portrait drawn by HiMegia. HiSchool's is a teacher and a student walking in conversation. Here the answer isn't handed over: it's reached along the way.