RocketPrepAI: AI Powered Test Prep

Over the past couple of months, I have been developing an AI powered SAT preparation service called RocketPrepAI. It has been an exciting journey for me and the rest of my team, given this is our first shot at the startup scene. Our goal is to provide low-cost, high quality SAT practice and assitance for students who don't have the resources to pay for private tutors. Today, I thought I'd go over how RocketPrepAI works and what makes us unique.

Our Team

Leo McClintock

Head of marketing

Nathan Vuong

Head of Finance

Arthur Johnston

Swiss Army Knife

Aidan Busby

Founder, lead engineer

How does RocketPrepAI work?

RocketPrepAI uses LLMs (Large Language Models) to generate quality mock SAT practice questions, and offers these to users through adaptive and manual study features. Our adaptive study sessions utilize a statistical framework called Item Response Theory, which uses bayesian probability to give users questions based off of how helpful that question is likely to be for their study journeys.

As of December 2025, we are looking into ways to further improve our adaptive study features and offer an automated way for users to generate day-by-day dynamic study plans that optimize the learning process by accounting for the "forgetting curve", the arch nemesis of SAT preparation. Our Item Response Theory framework already provides a solution to this for short-term practice, but for practice over long periods of time a more high-level solution would be nice, and certainly revolutionary!

Question Generation

Contrary to what you might believe, generating SAT practice questions is not as simple as opening up a ChatGPT window and asking it to generate you a hard level rhetorical synthesis question. If you want high quality questions that aren't immediately obvious, you have to go deeper. This is because the way Large Language Models like ChatGPT are trained, they don't understand conceptual difficulty. They only understand semantics, because they don't think like humans do.
So, how do you get a chatbot to think like a human question writer and not like a fancy autosuggest?
Turns out, you first have to come up with a generation formula. Here is what ours looks like:

Generate rough draft -> Analyze rough draft and come up with AI generated feedback -> Apply AI generated feedback -> Format question -> Human review

The next thing you have to do is give the model a bunch of data. Where do you get that? unfortunately I can't disclose our exact sources for intellectual property reasons, but if you take around you will find there is a limited but somewhat helpful number of free SAT practice questions out there- it turns out just enough to train an AI model to generate more of them!

What makes RocketPrepAI unique?

A lot of AI startups right now share the paradigm that all you have to do to create a successful company is vibe code a bug-infested ChatGPT wrapper, hook up stripe, and voila, you are the next Bill Gates.
This mentality has created a lot of pushback towards AI, because a lot of people are rightfully fed up with this proliferation of low quality AI services! But that's where we are different

Unlike some of our competitors, we don't just give you a vibe coded frontend masking a ChatGPT wrapper that uses prompts you could easily write yourself and use for free; No. We have spent months developing, testing, and improving our prompts, and the thousands of lines of Python code behind our generator- reviewing individual questions, giving the generator more and more data to incrementally improve itself with each question generated. And we aren't finished yet- Because quality takes time, and that's what we are here to give you.
In addition, we are now free, though for obvious reasons we do have a "standard" (AKA premium) tier for five dollars a month, for our more generous users.