Historical Behavior

Andrew Ng co-founded Coursera in 2012, driving the MOOC (Massive Open Online Course) revolution, with Coursera's users exceeding 150 million and revenue reaching $800 million by 2025. However, he has repeatedly stated that "online courses have not solved the problem of personalized learning." In 2023, he proposed that "AI will reshape education," and in 2024, DeepLearning.AI launched an AI tutoring tool, leading to the official establishment of LearnVector in 2026. His historical actions reflect the evolution of educational philosophy from "scaling" to "personalization."

Capital Path

Coursera's investment in LearnVector is an extension of its "AI-first" strategy. In 2025, Coursera launched an AI tutoring feature (Coursera Coach), but its effectiveness was limited. In Q1 2026, revenue was $230 million, an 18% year-on-year increase, but user growth slowed (up 8% year-on-year). Funding is shifting from "content procurement" to "AI R&D"—in 2026, AI-related investments will account for 40% of Coursera's R&D budget (up from just 15% in 2024). The $100 million investment in LearnVector represents 12.5% of Coursera's annual revenue, demonstrating its commitment to "betting on AI learning."

Analogy + Industry Position

Similar cases include Khan Academy's launch of Khanmigo (AI tutoring) in 2024, reaching 50 million users by 2026, though its profit model remains unclear. The difference between LearnVector and Khanmigo lies in their commercialization paths—LearnVector targets B2B corporate training (with over 10,000 corporate clients on Coursera), while Khanmigo focuses on free K12 users. The industry position is characterized by the "explosion of AI education infrastructure"—compared to online courses in 2020 (Zoom, Google Classroom), the potential market size for AI personalized learning could reach $500 billion by 2026 (global education spending of $7 trillion × AI penetration rate of 7%).

Structural Judgment

This fundamentally represents a technological substitution leading to a reconstruction of educational models. Traditional education (schools, online courses) follows a "one-to-many" model, where teachers/content are fixed, and students adapt to the content. AI personalized learning adopts a "one-to-one" model, where content adapts to the student. This explains why Ng emphasizes that "chatbots are not enough"—simple Q&A leads to "cognitive offloading," where students rely on AI to complete tasks rather than mastering skills. LearnVector's "path planning + continuous adaptation" model essentially replaces the role of a "private tutor" with AI, reducing costs to 1/1000 (human tutors at $100/hour, AI tutors at $0.1/hour). This promotes "educational democratization"—personalized learning, once exclusive to high-net-worth individuals, will become accessible to the masses.

ABAB News · Cognitive Laws

Scaling addresses access, while personalization addresses effectiveness.
AI does not replace teachers; it replaces private tutors.
The second half of edtech focuses on learning paths rather than content libraries.