In the era of large models, is AI helping students "do homework" or "learn how to learn"? Former Google executive Piyush Ranjan has clearly chosen the latter. Recently, he officially launched the AI education platform Fermi.ai, aiming to completely change the logic of how middle school students learn STEM (science, technology, engineering, and mathematics) subjects.
Reject quick answers, focus on "deep learning."
Different from AI assistants that simply provide solution steps, Fermi.ai's core philosophy is to use AI to promote deep thinking, rather than offering cheap correct answers. The platform focuses on math, physics, and chemistry, guiding students step by step through conclusions using intelligent canvas, concept map question banks, and adaptive real-time tutoring. It is more like a patient private tutor, aimed at helping students understand underlying scientific principles and accurately diagnose and correct errors in their thinking.
From a prestigious background, it carries the "next billion users" gene.
The founder Ranjan has an impressive background; he once led several major projects at
Impressive test results: Students are becoming more "independent."
According to feedback from pilot projects, Fermi.ai has delivered impressive results: after using the platform, students showed continuous improvement in learning outcomes. More encouragingly, as usage time increased, students' reliance on AI prompts significantly decreased. This proves that AI can indeed act as a "scaffold," stepping back once it has taught students how to think.
Currently, Fermi.ai is planning to expand into more subjects. In the context of companies like
