Science & Education

OpenMAIC41.3k★

OpenMAIC turns a topic, or your own documents, into an interactive class: slides, quizzes, simulations and project work, taught by AI teachers and classmates who talk, draw on a whiteboard and discuss with you. It comes from Tsinghua University's MAIC lab, is backed by a peer-reviewed paper, and is MIT licensed and free. Self-hosting got heavier in October. Version 1.2.0 moved generation to the server, so you need PostgreSQL and a long-running Node process (docker compose up gives you both); serverless hosts like Vercel stop at 1.1.x. You bring the model keys, and lessons export as PPTX or offline HTML. There's no paid tier. Solo learners and teachers: free, you pay only for tokens. Schools and training teams: free, and MIT allows commercial embedding. Budget tokens seriously: the project's own figures put a full agent-workbench course near 3.6 million. The catch is security. From September 6 to October 11 the project published 16 advisories, two critical: unauthenticated SSRF into cloud metadata, arbitrary file writes, stored XSS. All patched, but with ACCESS_CODE unset every route is open to whoever reaches the port. Stay current and keep it off the open internet.

stellarium10.1k★

Stellarium turns your screen into a planetarium. Give it a date and a location and it renders the sky as you would actually see it, naked eye, binoculars, or through a specific telescope and eyepiece you configure. Windows, macOS, Linux, and Haiku, shipped as a normal desktop download under GPL-2.0. No account, no tier, no upsell. Depth is what separates it from a phone star app. It carries multiple sky cultures rather than only the Greek constellations, renders 3D landscapes so the horizon matches your actual site, and its plugin system includes real telescope control. Amateurs use it to plan a session before going outside. Schools use it because putting it on thirty machines costs nothing. Install is a download and a double click. Nothing to host, nothing to maintain, no ops burden whether you are one person or a whole department. The catch is the mobile story. The desktop app is free and open source, but the phone app is a separate product from Noctua Software, not from this repository. Stellarium Mobile is free to download and the Plus upgrade is an in-app purchase at $18.99, with a cheaper recurring option. Free on your laptop does not mean free in your pocket.

ccf-deadlines9.5k★

ccf-deadlines answers the question every computer science researcher keeps asking: when is the next paper deadline? It's a community-maintained list of CS conference deadlines, from NeurIPS and ICML to database, networking and security venues, with each one's timezone, rank and past dates. MIT licensed, free, no account needed. Most people have nothing to install. Browse ccfddl.com, or subscribe to an iCal feed filtered by field or rank so deadlines land in your own calendar. There's also a free public MCP server so an AI assistant can look deadlines up directly, and a Python CLI you install from source. Updates arrive as pull requests, and the maintainers use AI agent skills to check each change against the official call for papers. Grad students, researchers and anyone planning submissions should bookmark it. The older paperswithcode/ai-deadlines site went quiet in 2024 and only covered AI; this covers all of CS and gets updated constantly. The catch: the ranking system is China's CCF list, so the A, B and C badges reflect that list's view of each venue. CORE rankings are there as a filter, but secondary. And when a conference doesn't publish a time or timezone, the site fills in 23:59 Anywhere on Earth by convention, so check the official call before you plan around the last hour.

canvas-lms6.9k★

Canvas LMS is what universities and K-12 schools actually use. It's the open source version of the same platform that Instructure sells commercially to thousands of educational institutions. AGPL-3.0, written in Ruby on Rails. You get the full LMS: course management, a gradebook, discussion boards, a calendar, file storage, video conferencing integration, and LTI (Learning Tools Interoperability) support for third-party tool plugins. Self-hosting is free under AGPL. Instructure sells the hosted version (Canvas Cloud); pricing is per-institution and not publicly listed, but think five to six figures annually for universities. The catch: self-hosting Canvas is a serious undertaking. It requires Ruby on Rails, PostgreSQL, Redis, a job queue, file storage (S3 or local), and real ops knowledge. The AGPL license means any modifications you make must be open sourced if you offer it as a service. And without Instructure's support, you're on your own for updates, security patches, and integration issues. This is enterprise software that happens to be open source, not a weekend project.

human-atlas4.3k★

Human Atlas is a 3D anatomy explorer that runs in a browser tab. It loads the BodyParts3D reference model, 2,234 individually selectable meshes across fifteen body systems, and lets you orbit, hide layers, and search anatomical terms. There is a free hosted demo, and the code is MIT so you can clone and run it yourself. React, Three.js, and Vite, which means the build output is a static site. Deploy it to Vercel, an S3 bucket, or whatever folder your school already serves. No backend, no database, no API keys. The mesh data is about 33 MB compressed, so first load is the only real performance question. Students, teachers, and anyone curious where an organ actually sits are the audience. The author states plainly that it is not a diagnostic or surgical tool, which is the right call. The catch is the license split. The code is MIT but the anatomy data is BodyParts3D under CC BY 4.0, so forking this into a product carries an attribution obligation the MIT code does not cover. The second limit is the model itself: BodyParts3D is one adult male reference, so it shows you an anatomy, not the range of them.

kana-dojo3.7k★

KanaDojo drills Japanese in the browser, Duolingo style, with no account and no paywall. It covers hiragana and katakana, kanji from JLPT N5 up through N1, and vocabulary that goes just as deep, with streaks, achievements and about 35 experimental mini-games and toys to keep you coming back. It also ships a translator, a verb conjugator and an Anki deck converter. AGPL-3.0. There's nothing to set up. kanadojo.com works out of the box. For your own copy there's a Dockerfile and a Vercel guide. The drills run in the browser, but the translator, progress sync and bug reports each need third-party keys (Google or Azure translation, Upstash Redis). Run a modified version publicly and the AGPL means you share your changes. The hosted site is free and now carries Google ads, plus a Ko-fi donation prompt after streak milestones. Solo learners: use the hosted site. Teachers and schools: self-host to drop the ads. Review works by mixing learned sets back in based on how well you know them, not an Anki-style schedule with due dates. The catch: it's recognition and recall practice only. No grammar course and no listening or speaking practice (it can read words aloud with your browser's text-to-speech), so pair it with a textbook or a tutor. And ignore the star count. The repo auto-posts thousands of no-code "good first issues" whose first step asks you to star the repo, so popularity here says nothing about quality.

Mural1.7k★

Mural teaches a language by making you talk. You hold spoken conversations with an AI tutor across 24 everyday settings instead of doing drills. It tracks the words you stumble on and brings them back in later sessions. All learning records stay on the phone, with no cloud sync, analytics, or ads. MIT licensed, with native iPhone and Android clients in one repo. Getting it running is the hard part. There is no App Store or TestFlight listing. iPhone means building it yourself with Xcode 26 on a Mac for iOS 26.1 or later, and a free Apple account has to re-sign the app every seven days. Android is easier: a preview APK for Android 8 or later. Either way you connect your own OpenAI API key with billing turned on. A ChatGPT subscription does not count. Solo learners comfortable with Xcode or sideloading: free, plus whatever OpenAI charges you. It covers eight languages, including Spanish, French, German, Italian, Brazilian Portuguese, Mandarin, and Norwegian. Everyone else should wait for the hosted version the project is planning, which is not active yet. The catch: your OpenAI bill is open-ended. The README says the conversation time limit is not a billing cap. The project also admits fluent-speaker review is still needed before it makes any claim about pronunciation or learning results. It is a practice partner, not a course.

easyread1.0k★

EasyRead translates English academic papers into comfortable Chinese, page by page, with the original alongside and equations re-typeset so they render properly. You can highlight, take notes, ask an AI about the paper, and export citations as GB/T 7714, APA or BibTeX. It can also translate into Japanese, Korean, Spanish, French, German and Italian. MIT licensed. It runs locally: download an installer, or run from source with Python 3.10+. You pick the translation engine: your Claude Code or Codex subscription (no API key), a paid API like DeepSeek or OpenAI, a free tier like Gemini or Groq, or a fully offline local model through Ollama. Paper text goes to whichever model you choose, and API keys sit in a plaintext config file on disk. Chinese-speaking grad students and researchers reading English literature are the audience, and for them it's free beyond model costs, which the project puts at cents per paper on DeepSeek. Readers who want layout-perfect bilingual PDFs should compare PDFMathTranslate, which is AGPL licensed. The catch: it's a student's one-person project that launched in late September 2026 and ships several releases a day. Installers are unsigned on Windows and not notarized on macOS, so expect security warnings, and the Mac build is Apple Silicon only.