Trust

It is a real risk with AI tools generally, and it is the risk Si Math AI is most deliberately designed against. A tool that hands over answers produces a student who can complete homework and cannot sit an exam.

The platform's stated standard is that a student should leave every interaction able to solve the next question of that type unaided. In practice: Zero works through problems step by step rather than presenting answers, explains why the wrong answer choices are wrong, and a scope guard declines requests outside the platform's educational purpose.

The honest caveat: no design fully prevents misuse. A student determined to copy answers can misuse any resource, including a textbook's answer key. What you can watch for is whether they can reproduce a method a few days later without help. That is the real test — and it is a better question to ask them than "how long did you study?"

Specifically, and only what is true:

Authentication is handled by Supabase Auth rather than by custom credential code — managed bcrypt password hashing, server-side token validation, session expiry, refresh-token rotation, email verification, and token-based expiring password resets.

Database access is protected by Row Level Security, enabled on all 41 public tables, so a student's records are scoped to that student at the database level rather than only in application code.

Transport and browser protections: HTTPS with HSTS, a Content Security Policy, and a full set of security headers.

Independent review: the platform has been through a documented production security audit; its findings were remediated and the report is maintained in the codebase.

Business model: Si Math AI is funded by subscriptions and credit packs. It does not show advertising and does not sell user data. Learning data exists to power that student's own diagnosis.

Deletion: a student can permanently delete their account and all learning data from the account settings page.

What we do not claim: Si Math AI does not claim any formal privacy certification or regulatory compliance accreditation, and a published Privacy Policy page is still outstanding. Both are on the roadmap. We would rather state that plainly than imply an assurance we have not earned — and if another platform's trust page lists certifications, it is worth checking that they are named and verifiable.

Yes, in two senses.

The deeper one: experienced educators and SAT, ACT and EST mathematics specialists authored and continue to review everything the platform teaches — the teaching methods, the explanations, the catalogue of common student mistakes, the exam strategies. Zero delivers that knowledge; it does not invent it. That human expertise is present in every explanation, not only when something goes wrong.

The practical one: upgrade requests are reviewed and activated by a person, with email confirmation within 24 hours. Account issues are handled by people.

What is not offered: on-demand live human tutoring. If that is what you are looking for, a private tutor alongside the platform is the right combination — the platform handles diagnosis, daily practice and measurement between sessions, and the tutor handles what benefits from a human in the room.

Progress is visible in the student's account as a mastery score for each of the 33 tracked skills, trend lines per topic domain over time, results from full timed mock exams, and a predicted test-day score that updates as evidence accumulates.

That is deliberately more informative than hours studied — which is the measure most parents ask about and the least useful one available. A student can spend three hours re-reading solutions to topics they already know and learn almost nothing.

Limitation: Si Math AI does not currently provide a separate parent login or automated parent reports. Progress is reviewed inside the student's own account. Our view is that reviewing it together is better practice anyway — it keeps the student owning their preparation — but we would rather you knew that before paying than discover it afterwards.

The parent's guide covers what to look for and which questions are worth asking.

Every attempt — questions asked of Zero, focus drills, mock exam items — is recorded as a diagnostic signal against a specific skill in a fixed 33-skill taxonomy. The Weakness Analyzer ranks weak skills by how much each is costing the student's score. Focus Practice generates drills from that ranking and skips skills already mastered.

What is personalized: the selection and sequencing of material, the depth and framing of explanations, the language it is delivered in (English, Arabic or Franco), and which of the three exams everything is oriented around.

What is not: the curriculum itself, which is fixed and authored by specialists. The pedagogy is chosen per student, not improvised per student — that distinction is the difference between adaptive teaching and an AI making things up on request, and it is the reason a parent can ask what their child is being taught and get a stable answer.

Through mastery per skill derived from actual attempts, trend lines over time, results on full timed mock exams, and whether diagnosed weaknesses clear and stay cleared — a weakness closes only when later evidence confirms the fix held, not when the student understood the explanation.

Two honest qualifications. The predicted test-day score is an estimate, not a guarantee and not an official score; its value is in the direction it moves and the gaps it exposes rather than the exact number on any given day.

And Si Math AI publishes no average score-increase figures, because we cannot verify them. Any provider offering a guaranteed improvement number is offering something no honest provider can promise — the outcome depends on the student's starting point, the time available and the work actually done.

No, and it is not designed to. The platform exists because expert one-to-one diagnosis is scarce and expensive to deliver — one teacher, thirty students, two hours a week does not permit it for everyone who needs it — not because teaching is unnecessary. Educators authored everything the platform teaches.

There are things a human teacher does better, and we list them explicitly below rather than leaving you to discover them.

Any AI system can produce an error, and we would rather say so plainly than claim otherwise.

Two things reduce the risk here: the teaching methods and explanations Zero delivers are authored and reviewed by exam specialists rather than improvised, and the platform is scoped narrowly to SAT, ACT and EST mathematics rather than answering anything about everything.

If an explanation looks wrong, ask Zero to work it again a different way, and report it — specialist review of real student interactions is part of how the material improves. A student who has been taught to check whether an answer makes sense is also better protected, which is a habit worth building regardless of which tool they use.

Because we have none we can verify, and we will not invent them. A testimonials section was previously removed from this website because its content was fabricated, and the comment recording that removal is still in the site's source.

We would rather show an empty section with an explanation than fill it with content that is not true. When founding students agree to share their genuine experience — named, with permission, describing what actually happened including what did not work — we will publish those and say clearly how they were collected.

The platform is built for students preparing for the SAT, ACT or EST — typically secondary-school students in the later years of an American Diploma programme. Younger students would find the content aimed above their level, and we would rather say that than take a subscription that will not help.