Jason Mentors The Next Generation + Why AI Is Bad at Teaching (Aristotle Interview) | E2346

Jason’s central advice to students was to use AI to extend their learning, not just complete assignments: identify gaps in understanding, ask questions, and build models of their own. He encouraged Agastya to create AI agents that could teach finance one-on-one, including advanced material. For students considering venture capital, Jason recommended studying a firm’s investments, decisions, exits, and fund returns across decades, rather than relying on general impressions of the industry.

Jason also urged students to read biographies well beyond finance, ideally one a month or at least 10 a year, to understand how industries and careers developed and to find lessons across fields. He assigned David five biographies, asking him to take notes and record two takeaways from each, then return with what he learned in a month or a year. He cited Walter Isaacson’s biography of Einstein and Stephen King’s On Writing as examples, and asked Agastya to report after his first college year on the biographies he had read.

For aspiring entrepreneurs, Jason advised starting a project to learn the full mechanics of building a business, including taxes, negotiation, forming an LLC, and payroll. Freelancers can fill gaps in skills the founder does not have. He pushed back on the idea that a company needs $500,000 to get started: in his view, $10,000 to $20,000 a month could cover expenses and one employee, whereas spending $25,000 or $50,000 monthly for 10 months could become a costly distraction. The point was to learn by building while keeping the experiment manageable.

Jason said students should distinguish between investing in skills, starting a company, and investing in markets or other founders. Looking back to 1996, he said he might have considered funding the first year of his magazine or angel-investing in technology companies. David said he would not build a company around the first idea he encountered, and considered whether moving from Long Island to New York City could save about two commuting hours a day. He estimated a shared Manhattan apartment at $1,500 to $2,500 per month and said he might put $250,000 toward angel investments.

Jason argued that general-purpose AI models are not necessarily getting better at teaching. Models optimized for complex reasoning may deliver slow, verbose answers that finish a task without helping a student understand it. Aristotle co-founder Sean Reddy described effective teaching as creating productive struggle: guiding students through each step rather than letting them delegate the work to AI. Aristotle’s core approach is to predict the next useful teaching move, whether that means explaining a concept, asking a question, offering a practice problem, or using a diagram or whiteboard. Reddy said that requires understanding a student’s knowledge and context, while current language models often lack the multimodal capability and training to interpret cues such as attention, speech, or facial expressions.

Aristotle’s demonstration taught integration through the distance traveled at a constant speed, using questions and an interactive whiteboard. The company serves students preparing for AP and IB courses, the SAT and ACT, and other subjects, with a stated ambition to teach anything to any student. Reddy said most U.S. tutoring demand comes from students ages 13 to 18 and fewer than one in 52 U.S. students has a personal tutor; he pointed to the strong outcomes associated with one-on-one tutoring and Bloom’s Sigma problem. Aristotle is building its own tutoring system, benchmarks, and evaluations, hand-grading thousands of hours of sessions to generate training data that captures spoken and session context missing from text alone. It post-trains specialized models for tasks such as generating math problems, reading whiteboards, and understanding student speech, then combines them in a pedagogical system. The company keeps student-session data proprietary, citing the need to protect information involving people under 18.

Reddy described Aristotle as both an AI lab and a tutoring business, charging per session and offering session packs and unlimited access. Jason cited a $50 monthly package for eight sessions; Reddy put the unlimited plan at about $200, and said some students use the service for more than 60 hours a month. Jason said AI’s educational potential receives less attention than fears about danger and disruption, while emphasizing that outcomes depend on how the technology is used. He also raised concerns about children’s socialization and asked whether group or multiplayer AI learning could support peer learning. He predicted that successful tools such as Aristotle could eventually enter classrooms, potentially creating demand for teacher dashboards.