Aether Initiative is an open, async-first study map for people who want to understand frontier AI and think seriously about AGI.
Aether Initiative 是一个开放、异步、低压力的前沿 AI 学习地图。它适合想系统理解 LLM、视觉模型、多模态、世界模型、强化学习、Agents、持续学习、记忆、认知架构、因果学习和开放式学习的人,也适合还在寻找自己研究主线的人。
The repo is closer to a shared navigation system than a fixed curriculum: read what is useful, leave notes when you have something to say, and build your own path as your questions become clearer.
这个仓库更像一张共同维护的路线图:你可以按顺序读,也可以围绕自己的问题跳读;可以安静学习,也可以贡献笔记、问题、论文卡片、idea 或小 demo。
- Learn at your own pace.
- Use the roadmap when you want structure.
- Create a personal path when your goal becomes more specific.
- Share notes, questions, papers, demos, and research ideas when they are useful to others.
- Keep technical claims grounded in sources.
- Keep the atmosphere calm, curious, and welcoming.
中文:
- 按自己的节奏学习。
- 想要结构时使用总路线图。
- 有明确兴趣后,可以写自己的个人路线。
- 有笔记、问题、论文理解、demo 或 research idea 时,分享出来即可。
- 技术判断尽量附上来源。
- 保持讨论清晰、友好、低压力。
.
├── Learning_Areas/ # Reading roadmaps for each direction / 各方向阅读路线
├── Study_Docs/ # Chinese DOCX study notes / 中文讲义型 DOCX
├── Reading_Guide/ # Bilingual guide for using the roadmap / 双语阅读指南
├── templates/ # Note, idea, demo, and path templates / 笔记模板
├── personal-paths/ # Optional personal learning paths / 个人学习路线
├── discussions/ # GitHub Discussions draft posts / Discussion 草稿
├── assets/community/ # Community entry assets / 社群入口素材
├── Code/ # Scripts used to generate study docs / 文档生成脚本
├── GITHUB_SETUP.md # GitHub publishing steps / GitHub 发布步骤
├── START_HERE.md # Community intro text / 社群入口说明
├── INTEREST_FORM.md # Lightweight interest form / 加入表单模板
├── CONTRIBUTING.md # Contribution guide / 贡献指南
└── CODE_OF_CONDUCT.md # Community norms / 社群行为准则
There is one recommended order for people who want a coherent path from foundations to AGI-oriented systems. The starred folders in Learning_Areas/ mark the current main line.
如果你希望按一条相对完整的主线学习,可以从下面的 DOCX 顺序开始。Learning_Areas/ 中带 * 的方向,是当前更贴近 AGI 体系思考的主线方向。
01_LLM_Foundation.docx02_Vision_Models.docx03_Multimodal_Models.docx04_World_Models.docx05_Reinforcement_Learning.docx06_Reasoning_and_Planning.docx07_AI_Agents.docx08_Continual_Adaptive_Learning.docx09_Memory_and_Knowledge_Update.docx10_Cognitive_Architectures.docx11_Causal_Learning.docx12_Open_Ended_Learning.docx
13_AI_Systems_Infrastructure.docx14_Diffusion.docx15_3D_Spatial_AI.docx16_Robotics_Embodied_AI.docx17_Neuro_Symbolic_Reasoning.docx18_Mechanistic_Interpretability.docx19_Social_Learning.docx20_Evaluation_Alignment_Safety.docx
Start with:
- 01_Foundations_and_Infrastructure
- 02_Perception_Generation_and_Worlds
- 03_Agency_Learning_and_Memory
- 04_AGI_Principles_and_Cognitive_Systems
- 05_Evaluation_Alignment_and_Safety
Small contributions are valuable here. A useful question, a clean paper note, a corrected link, or a tiny demo can help the next person move faster.
这里很适合轻量参与。一个好问题、一篇论文笔记、一个修正链接、一个小实验,都可能帮后来的人少绕一点路。
Useful templates:
The shared roadmap gives a common language. Personal paths let different people follow different questions.
总路线提供共同语言;个人路线保留每个人自己的兴趣和节奏。
Examples:
If this repo is paired with a Discord, Slack, WeChat group, forum, or GitHub Discussions space, the suggested setup is in COMMUNITY.md.
如果你想把这个仓库和 Discord、Slack、微信群、论坛或 GitHub Discussions 配合使用,可以从 COMMUNITY.md 开始。
Current community entry points:
- Discord: https://discord.gg/27XQXvxG
- WeChat QR / 微信群二维码: assets/community/wechat-group-qr-2026-05-20.jpg
- GitHub Discussions: https://github.com/RCF-117/Aether-Initiative/discussions
- First Discussion / Welcome thread: #1
- GitHub repo: https://github.com/RCF-117/Aether-Initiative
当前社群入口:
- Discord: https://discord.gg/27XQXvxG
- 微信群二维码:assets/community/wechat-group-qr-2026-05-20.jpg
- GitHub Discussions:https://github.com/RCF-117/Aether-Initiative/discussions
- 第一条 Discussion / 欢迎帖:#1
- GitHub 仓库:https://github.com/RCF-117/Aether-Initiative
For contribution style, see CONTRIBUTING.md. For the lightweight entry text, see START_HERE.md. For an optional signup form, see INTEREST_FORM.md.
贡献方式见 CONTRIBUTING.md。社群入口文案见 START_HERE.md。轻量加入表单见 INTEREST_FORM.md。
Use official paper links in public docs. Downloaded PDFs, rendered QA images, caches, and local build files should stay local; .gitignore already covers the common cases.
公开仓库里建议使用论文官方链接。本地下载的 PDF、渲染检查图片、缓存和构建文件保留在本机即可;常见路径已经写入 .gitignore。
Released under the MIT License.
本项目采用 MIT License 发布。