AI is the fastest-changing content area of recent years — a new model almost monthly, yesterday’s tutorial stale by morning. This guide does not rank by follower count. It groups creators by what you want to do with AI, to help you aim in the flood.
All ten creators below have profiles indexed on Dawn Island; click a name to see the platform, content focus, and any verified public figures.
Want to build AI projects
Lin Yi LYi
A former algorithm engineer who on Bilibili uses AI for hardcore-yet-playful projects — from training an AI to beat fighting-game bosses to running an all-AI company, blending reinforcement learning, generative AI, and startup reality.
Who it suits: those who want to see AI actually “built”, curious about both tech and entrepreneurship. How to use it: as an idea bank and project sample; learn his engineering thinking of “turning an idea into a project”.
Want to study theory and papers
Tongji Zihao
Focused on AI paper reading, computer vision, large models, and robotics; his “AI paper reading” series and Stanford CS231N Chinese walkthrough are go-to resources for Chinese learners.
Who it suits: students and engineers who want to read papers and go deep into algorithms. How to use it: as a systematic textbook; build a CV and LLM framework in his series order.
Gengzhi Ge
An AI PhD who explains large models, papers, and research trends in a straight, no-nonsense style, making the math and ideas behind modern AI accessible.
Who it suits: those who want to keep up with the frontier without being scared off by formulas. How to use it: as a “frontier radar” for quick intuition on LLMs and research trends.
shuhuai008
Known for the “machine learning whiteboard derivation” series, deriving ML algorithms from scratch step by step, rigorous and gradual.
Who it suits: learners who want true theory, not just package-calling. How to use it: when you want to crack an algorithm’s math, follow the whiteboard — more useful than reading conclusions.
Want to use tools and get started
Qiuye aaaki
Widely regarded as the person who brought Stable Diffusion to Chinese creators; his one-click SD / ComfyUI / WebUI packs lowered the barrier for millions, and his model leaderboards remain a staple reference.
Who it suits: ordinary people who want AI art but were put off by environment setup. How to use it: start with his intro and packs, then follow the model leaderboard to stay with the community mainstream.
GenJi
A 2025 Bilibili Top-100 creator focused on AI application education — AI software development, animation, LLM principles, voice cloning; his “90 minutes to use AI” series brings advanced AI to millions.
Who it suits: workplace people who want a systematic, zero-to-usable path. How to use it: as an “application course”, following his series from tool to scenario step by step.
Lao Mai’s Tool Library
Surfaces the newest and hottest AI tools — Sora, Claude, every model drop — with first-hand tests and practical efficiency workflows; among Bilibili’s top AI creators.
Who it suits: those who do not want to miss new tools and prefer real tests before adopting. How to use it: as a “tool intelligence station”; when a new model drops, watch his test first to skip your own pitfalls.
Qiuzhi 2046
A former product manager turned AI creator sharing tool reviews, hands-on teaching, and industry observation; directed Bilibili’s first AI Spring Festival Gala in 2026.
Who it suits: ordinary users wanting both “how to use” and “where the industry goes”. How to use it: tool practice from her reviews, industry judgement from her observation — covering both ends.
Want to patch coding and ML basics
Mofan Python
A long-time Bilibili coding educator whose short, punchy Python and ML tutorials accompanied a generation of Chinese beginners, covering data processing, sklearn, neural nets, and RL.
Who it suits: absolute beginners who want a solid code base. How to use it: as an entry textbook; do Python and sklearn first, then papers and projects above.
Want trends and the founder view
Zhou Hongyi (Red Coat Uncle)
Founder of 360, viral as “Red Coat Uncle”, posting almost daily on AI, cybersecurity, and startup management, and proposing an L1–L5 “agent evolution” framework.
Who it suits: those who want the entrepreneur and industry view of where AI goes. How to use it: as a “trend wind vane” — but remember his content carries a stance endorsing his own AI products; listen broadly.
How to judge whether an AI-teaching channel is trustworthy
Three workable tests:
Does it explain principle, not just show effect? “Watch me one-click generate” is fun, but a channel that cannot say why leaves you unable to adapt.
Does it label timeliness and version? AI tools iterate fast; tutorials without version or date may not run three months later.
Does it separate “can demo” from “can deploy”? Demo environments are clean, real ones messy; a reliable channel states the boundary.
Common mistakes
- Chasing new models, skipping principle. Hype changes quarterly; principle lasts years. Base beats fad.
- Watching without practising. AI is hands-on; running one project beats ten videos.
- Swayed by “quick mastery”. “Master LLMs in three days” is mostly clickbait; systems engineering takes time.
To browse more profiles in this niche, see all indexed AI creators.