DeepTutor vs How LLMs Work

DeepTutor is an open-source AI tutoring platform from HKUDS that just shipped v1.0.3. Unlike ChatGPT wrappers dressed up as learning tools, DeepTutor is architected around a genuine agent-native philosophy: "not chatbots — autonomous tutors." The system runs five integrated modes within a single continuous thread — Chat (with RAG and web search), Deep Solve (multi-agent problem solving with source citations), Quiz Generation, Deep Research (parallel agents with cited reports), and Math Animator (Manim-powered visual explanations of mathematical concepts). Context flows between modes, so a question in Chat can escalate to Deep Solve without losing thread history. The standout feature is TutorBots — persistent AI tutors that maintain their own memory, personality, and skill sets across sessions. Combined with a RAG-ready knowledge base where you can upload your own PDFs and notes, DeepTutor effectively becomes a personalized learning environment that evolves with you. A Co-Writer feature turns any document into a collaborative editing session with AI as a genuine co-author. An Agent-Native CLI exposes every capability as structured JSON for autonomous agent pipelines, complete with a SKILL.md spec. The platform supports 25+ LLM providers including OpenAI, Anthropic, DeepSeek, Groq, and local models via Ollama or llama.cpp. It ships under Apache 2.0, installs via Docker, and launched v1.0.3 on April 13, 2026 with question notebooks and Mermaid diagram support. For students, researchers, or anyone building on top of a learning platform, this is the most architecturally serious open alternative to closed tutoring products.

"How LLMs Work" is a free, browser-based interactive guide that walks through the complete pipeline for building large language models — from raw web scraping to RLHF-trained conversational assistant. Created by Yash Narwal and based on Andrej Karpathy's technical deep-dive lecture, it's been getting significant traction on Hacker News (214+ points) for turning dense ML theory into something genuinely accessible. The site covers data collection and deduplication, Byte Pair Encoding tokenization with a live demo, pre-training and next-token prediction, inference with a probability sampling simulator, post-training with RLHF, and RAG. Each section uses animated visualizations, clickable pipeline diagrams, and canvas-based graphics — not static explainer images. The progressive narrative structure follows a single piece of text through every stage of the pipeline, making abstract concepts concrete. In an era where everyone uses LLMs but few understand how they work, this kind of high-quality educational resource matters for a different reason than tools: it raises the baseline competency of the entire developer ecosystem. Better-informed builders ask better questions, make better design decisions, and push vendors toward more transparency. This is the kind of project the HN community rewards — and deserves the signal boost.