MasterClass bets AI teaching agents can broaden access to tutoring
SiliconANGLE Jonathan Anthony
MasterClass is building AI tutoring agents that adapt to each learner. It could make one-to-one teaching less rare, if the evals actually hold up.
Based on reporting by SiliconANGLE, Jonathan Anthony — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
MasterClass is trying to turn AI tutoring from a demo into something that can scale. The company is building teaching agents that adapt lessons to the way each student responds, with the pitch that personalized instruction could reach far more people if software takes some of the cost and staffing pressure out of tutoring.
Mandar Bapaye, MasterClass’s chief product officer, says the tech only works if the pedagogy is sound. The system can’t just drop a chatbot into a lesson and hope for the best; it needs a learning framework underneath it. That matters because MasterClass is aiming at actual instruction, not a novelty feature people try once and forget.
The company’s AI-native business program, MasterClass Executive, uses a multi-agent system that watches for signals like cognitive overload and falling motivation, then adjusts the lesson plan. Bapaye said MasterClass runs about 10 agents behind every learner interaction, tracking inputs, outputs, tool calls and the communication between agents. That’s a lot of moving parts, and it makes testing and monitoring just as important as the model itself.
That’s where CoreWeave comes in. Lukas Biewald, senior vice president of AI initiatives at CoreWeave, has been advising the team, and MasterClass has selected W&B Weave to trace, monitor and improve the agents. Bapaye said MasterClass also built its own agent on Weave’s Model Context Protocol interface, and that agent reviews traces each night to flag issues and likely root causes. It’s the unglamorous side of AI, but it’s the part that decides whether the system keeps working once real users show up.
The demand numbers explain why the company is pushing so hard. The first cohort drew 30,000 applications for about 500 spots, and the second is nearing 50,000 applications. Bapaye framed the problem as cost, supply and quality: personal teachers are expensive, scarce and uneven. AI, he argues, can attack all three at once. Maybe. But the real test is whether the system can keep its promises after the lab lights are off.
My take — AI-written commentary, not fact-checked reporting
This is the right place to use AI: boringly, repeatedly, and with lots of trace data. The industry loves talking about “personalization,” then ships a chatbot and calls it education. MasterClass at least seems to understand that tutoring is an operational problem as much as a model problem, which is more than most AI vendors manage before breakfast.
Read more about this at: SiliconANGLE
