From 1x to 100x: Open-Source AI Tutoring as Evidence Infrastructure for Education Evaluation

Access to learning content is no longer the bottleneck in global education. The real crisis is motivation and engagement. Project P&AI, built on years of scaling Pandai (YC S21) across Malaysia, is an open-source AI-powered learning system that takes a fundamentally different approach: instead of waiting for students to ask questions, it proactively initiates study sessions through messaging apps, adapts via spaced repetition, and sustains engagement through gamification mechanics proven with over 1 million Malaysian students.

What makes P&AI distinctive for the evaluation community is the evidence infrastructure it generates. The platform produces continuous, granular learning data; per-student mastery trajectories, misconception patterns, retention curves, and engagement analytics; available in real time to teachers, schools, and independent evaluators. This represents a shift from periodic, survey-based educational assessment toward continuous monitoring that informs both implementation decisions and outcome evaluation.

P&AI comprises three open-source components: (1) PAI Bot, a self-hostable proactive AI learning agent with teacher monitoring dashboards; (2) Open School Syllabus (OSS), the world's first machine-readable, open curriculum repository; and (3) OSS Bot, contribution tooling that channels student learning patterns back into curriculum improvements. The deployment model removes per-student fees: organisations self-host the system while donors fund AI compute costs, achieving unit economics under $0.50/student/month.

This session will present the P&AI model and early Malaysian school implementation data, demonstrate the platform's real-time evaluation dashboards, and facilitate a discussion on how the global evaluation community can help design rigorous impact assessments for AI-powered education, and on partnership opportunities for piloting, funding, and contributing to this open-source infrastructure.