Miyagi Labs Thinks AI Tutoring Should Feel Like a Tutor, Not a Flashcard App

The Macro: Test Prep Is a Massive Market Running on Old Ideas

SAT and ACT prep is a strange industry. It generates roughly $2 billion a year in the US alone, and the dominant players are still selling what amounts to a book with answers in the back. Khan Academy made it free but not personalized. Kaplan and Princeton Review charge thousands of dollars for group classes that move at the speed of the slowest student. Tutoring marketplaces like Wyzant connect you with a human tutor for $50 to $100 per hour, which works great if your parents can afford it.

The whole system has a core problem: it treats every student the same. You get the same practice questions, the same study schedule, the same explanations regardless of whether you’re struggling with quadratic equations or reading comprehension. Some kids need 200 hours of prep. Some need 20. The current market makes both of them sit through the same curriculum.

AI should fix this. The technology exists to build a tutor that watches how you solve problems, identifies where you’re weak, and adjusts in real time. The question is whether anyone has built that product well enough to compete with the incumbents who have brand recognition, school partnerships, and decades of trust.

There have been attempts. Squirrel AI in China has done adaptive learning at scale. Duolingo proved that gamified learning loops work for language acquisition. But standardized test prep has specific constraints that make it different from general education. You’re training for a specific format, on a specific date, with a specific scoring rubric. The feedback loops are tighter and the stakes are higher.

The Micro: MIT Chess Guys Building a Study Engine

Miyagi Labs is an AI-powered exam prep platform for the SAT and ACT. The product includes diagnostic tests, personalized study plans, over 10,000 expert-reviewed practice questions, and full-length practice exams. An AI tutor provides explanations tailored to each student’s learning patterns. They claim students see skill improvement at twice the rate of traditional methods, and a student testimonial on the site says they improved their SAT score by 200 points in two months.

The founding team is a two-person operation out of YC’s Winter 2025 batch. Tyrone Davis III studied computer science at MIT and previously founded The Gift of Chess, a nonprofit that spreads chess education globally. That is a genuinely interesting background for an ed-tech founder. He’s seen how structured learning can scale through a nonprofit lens, which tends to produce different instincts than someone who came up through pure startup culture. Guang Cui, also MIT computer science, describes the product vision as turning videos into complete and interactive courses. He lists ninja warrior among his interests, which has zero relevance to the product but is the kind of detail that makes you root for someone.

The competitive field is crowded but fragmented. Khan Academy is free and good but not truly adaptive. Magoosh has solid content at a budget price point. PrepScholar built an algorithm-driven approach years ago but hasn’t kept pace with modern AI capabilities. The premium tier is dominated by human-led programs like Manhattan Prep and TestCrafters. Miyagi sits in an interesting gap: more sophisticated than the budget apps, more affordable than the premium programs, and more adaptive than all of them.

The product is available on iOS and Android, which matters because students live on their phones. They’re also running a Discord community and have affiliate programs, which suggests they’re thinking about distribution seriously from the start.

The Verdict

I think Miyagi Labs is positioned well in a market that’s overdue for a real product upgrade. Test prep has been selling the same basic experience for 20 years with minor cosmetic changes. An AI tutor that genuinely adapts to each student and delivers measurable score improvements is the product everyone has been waiting for.

The risk is trust. Parents making a $500 to $2,000 purchasing decision about their kid’s college future tend to go with the brand they recognize. Kaplan has been around since 1938. Princeton Review since 1981. Getting a parent to trust a two-person startup with their teenager’s SAT score is a sales challenge that technical quality alone won’t solve. They’ll need testimonials, score improvement data, and probably school or counselor partnerships to break through.

At 30 days, I’d want to see completion rates on study plans. Are students actually following through, or dropping off after the diagnostic? At 60 days, the metric is score improvement. Not self-reported, but measured through practice exam progression. At 90 days, the question is word of mouth. If students who use Miyagi tell their friends, that’s the growth engine that makes everything else work. The product has 10,000 practice questions and a real AI tutor under the hood. The question is whether high school students will find it before their parents book the Kaplan class.