On-Device Socratic Coding Tutor for Middle School CS Classes
An offline AI tutor that runs on school Chromebooks and helps students debug Scratch and VEX VR projects using Socratic questions instead of answers — so kids learn to think, not copy.
middle school CS teachers and their students using block-based programming
- Socratic Q&A mode that asks guiding questions when a student is stuck on a bug
- Teacher dashboard that flags students who have lingered on the same concept for >15 minutes
- Rubric-aligned feedback grounded in classroom learning objectives
- Air-gapped deployment that keeps student code on the device, satisfying district privacy rules
Schools are rejecting cloud LLMs over privacy, and recent benchmarks show small language models can handle tutoring if prompted carefully.
Validated demand: Code.org AI Tutor, Khanmigo, Socratizer.ai, and Stanford SCALE offline tutor are all live products targeting this exact Socratic CS-tutor use case, and CSTutorBench exists as an active research benchmark for VEX VR tutoring.AI Tutor - Code.org ↗Khanmigo pricing: Free for teachers, $4/month for parents & learners ↗CSTutorBench: Benchmarking Small Language Models as Tutors for Block-Based Programming ↗
Crowded: at least four direct Socratic-CS-tutor competitors (Code.org, Khanmigo, Socratizer.ai, smolpc-codehelper offline assistant for ages 11-18), and on-device Chromebook inference has a reference implementation already — only the on-device + VEX-VR/Scratch niche looks defensible.GitHub - SmolPC-2-0/smolpc-codehelper: Offline AI coding assistant for secondary school students ↗Socratizer.ai — AI Pedagogy for Modern Education ↗
Khanmigo proves $4/mo and free-for-teachers school-licensing works, but K-12 district procurement cycles are long and Chromebook Management consoles (Google Admin) already gate distribution — WTP is proven but slow and price-sensitive.Khanmigo pricing: Free for teachers, $4/month for parents & learners ↗Khanmigo — AI Tool Review & Pricing ↗
Block-based CS is durable — Scratch and VEX VR remain central to K-8 CS standards and CSforAll; on-device AI is reinforced by the same privacy/regulation wave that bans cloud LLMs, but cloud tutors may reclaim space once privacy frameworks mature.Scratch - Educators ↗AI Tools for Education in 2026: A Practical Guide ↗
Technically demonstrated: Ithy article and smolpc-codehelper (Tauri + local Ollama) prove a small LM can run on Chromebook offline, and CSTutorBench codebase provides an eval harness — but the benchmark shows even GPT-4o struggles with realistic Socratic tutoring, so prompt/reliability work is the real blocker.Ithy - Running Small AI Models Locally on a Chromebook Without an Internet Connection ↗GitHub - Yunfeng-Wan/CSTutorBench: A benchmark to evaluate LLM Tutors ↗