BaileyLi2021 (@higher2021) header
BaileyLi2021 (@higher2021) profile picture

Founder of NaPa Academy l Speaker l Author

Cupertino, California
Joined February 2021
Photo by BaileyLi2021 (@higher2021)Photo by BaileyLi2021 (@higher2021)Photo by BaileyLi2021 (@higher2021)Photo by BaileyLi2021 (@higher2021)Photo by BaileyLi2021 (@higher2021)Photo by BaileyLi2021 (@higher2021)Photo by BaileyLi2021 (@higher2021)Photo by BaileyLi2021 (@higher2021)Photo by BaileyLi2021 (@higher2021)Photo by BaileyLi2021 (@higher2021)
Stanford's TEN FREE Courses on AI & ML: ❯ CS336 - LLM from Scratch ❯ CS221 - Artificial Intelligence ❯ CS229 - Machine Learning ❯ CS230 - Deep Learning ❯ CS234 - Reinforcement Learning ❯ CS236 - Deep Generative Models All links inside:
17
115
1
771
33,381
Naval Ravikant: The gut is the ultimate decision maker.
4
61
2
488
13,444
Elon Musk's advice on how to learn anything:
37
466
20
3,306
236,485
BaileyLi2021 retweeted
整理了一份从零自学 AI 的路线,全部用斯坦福、伯克利、MIT 等名校的免费公开课,YouTube 上都能看。 分 6 个阶段,大约一年: ① 数学和 Python 基础 ② 机器学习:CS229 ③ 深度学习与视觉:Karpathy Zero to Hero → CS231n ④ NLP 与 Transformer:CS224N → CS336(从零训练一个语言模型) ⑤ 强化学习:CS234 + 伯克利 CS285 ⑥ Agent:斯坦福今年秋天新开的 CS329Z + 伯克利 Agent 系列公开课 最重要的建议:一定要做作业,光看视频等于没学。 阶段 0:编程与数学基础 Python - [ ] ⭐ Stanford Code in Place(CS106A 的免费线上版)— 官网 (codeinplace.stanford.edu/) - [ ] ○ Harvard CS50P:讲得非常好的 Python 入门 YouTube (youtube.com/playlist?list=PL…) 线性代数 - [ ] ⭐ 3Blue1Brown《线性代数的本质》:先建立直觉 — YouTube (youtube.com/playlist?list=PL…) - [ ] ⭐ MIT 18.06(Gilbert Strang):经典中的经典 — · YouTube (youtube.com/playlist?list=PL…) - [ ] ○ Stanford ENGR108(Boyd,就是上一版提到的 EE103,现在改了编号),更偏应用 — · YouTube (youtube.com/playlist?list=PL…) 概率与微积分 - [ ] ⭐ Stanford CS109 — · YouTube 2022 (youtube.com/playlist?list=PL…) - [ ] ○ Harvard Stat 110(Joe Blitzstein):公认最好的概率课之一 — · YouTube (youtube.com/playlist?list=PL…) - [ ] ○ 3Blue1Brown《微积分的本质》 — YouTube (youtube.com/playlist?list=PL…) 阶段 1:机器学习 - [ ] ○ Coursera Machine Learning Specialization(吴恩达):适合零基础,先建立直觉 — 官网 (coursera.org/specializations…) - [ ] ⭐ Stanford CS229 — · YouTube 2018 吴恩达版 (youtube.com/playlist?list=PL…) - [ ] ○ Cornell CS4780(Kilian Weinberger):直觉讲得最清楚的机器学习课之一,和 CS229 互补 — · YouTube (youtube.com/playlist?list=PL…) - 验收项目:用 NumPy 从零实现线性回归、逻辑回归和 k-means,再用一个 Kaggle 数据集完整走一遍流程 阶段 2:深度学习 + 计算机视觉 - [ ] ○ 3Blue1Brown 神经网络系列:用动画讲清反向传播和 Transformer — YouTube (youtube.com/playlist?list=PL…) - [ ] ⭐ Karpathy《Neural Networks: Zero to Hero》:从手写自动求导一路写到 GPT。这是理解深度学习最好的入门方式,后几集正好衔接阶段 3 — ·YouTube (youtube.com/playlist?list=PL…) - [ ] ⭐ Stanford CS231n — · YouTube 2025 (youtube.com/playlist?list=PL…)(2026 年春季的录像只对选课学生开放) - [ ] ○ UMich EECS 498/598(Justin Johnson):相当于 CS231n 的加长版,作业质量很高,可以替代 CS231n — · YouTube (youtube.com/playlist?list=PL…) - [ ] ○ fast.ai Practical Deep Learning:先跑通,再深入原理 — · YouTube (youtube.com/playlist?list=PL…) - [ ] ○ MIT 6.S191:轻量入门课,每年更新 — · YouTube (youtube.com/playlist?list=PL…) - [ ] ○ Stanford CS230 — · YouTube Autumn 2025 (youtube.com/playlist?list=PL…) - [ ] ○ 生成模型:Stanford CS236 — · YouTube 2023 (youtube.com/playlist?list=PL…);MIT 6.S184 讲流匹配和扩散模型 — · YouTube 2026 (youtube.com/playlist?list=PL…) - 验收项目:完成 CS231n(或 EECS 498)的作业,再微调一个 ViT 或检测模型 阶段 3:NLP 与 Transformer - [ ] ⭐ Stanford CS224N — · YouTube Spring 2024 (youtube.com/playlist?list=PL…) - [ ] ○ Stanford CME 295:Transformer 和 LLM 概览 —· YouTube Autumn 2025 (youtube.com/playlist?list=PL…) - [ ] ⭐ Stanford CS336 — · YouTube Spring 2026 (youtube.com/playlist?list=PL…) - [ ] ○ Stanford CS25:用来跟进前沿 — · YouTube (youtube.com/playlist?list=PL…) - [ ] ○ CMU 11-711 Advanced NLP(Graham Neubig):覆盖面很广,包括 RAG 和 Agent — · YouTube (youtube.com/playlist?list=PL…) - [ ] ○ Hugging Face LLM Course:工程实战 — 官网 (huggingface.co/learn/llm-cou…) - 验收项目:完成 CS336 作业 1,从零写一个 BPE tokenizer 和 Transformer 语言模型,并把它训练起来 阶段 4:强化学习 - [ ] ○ Sutton & Barto《Reinforcement Learning》教材 — 免费在线版 (incompleteideas.net/book/the…) - [ ] ○ David Silver RL 课(UCL × DeepMind 2015):经典的理论入门 — · YouTube (youtube.com/playlist?list=PL…) - [ ] ⭐ Stanford CS234:打 RL 基础 — · YouTube Spring 2024 (youtube.com/playlist?list=PL…) - [ ] ⭐ Berkeley CS285(Sergey Levine):深度强化学习最权威的课 — · YouTube 2023 (youtube.com/playlist?list=PL…) - [ ] ○ Stanford CS224R:和 CS285 内容重叠,但用 RL 训练 LLM 的部分讲得更多 — · YouTube (youtube.com/playlist?list=PL…) - [ ] ○ OpenAI Spinning Up:给算法实现配上代码 — 官网 (spinningup.openai.com/) - [ ] ○ Hugging Face Deep RL Course — 官网 (huggingface.co/learn/deep-rl…) - [ ] ○ Stanford CS329H:RLHF 和偏好学习,Autumn 2026 正在开课 — 官网 (web.stanford.edu/class/cs329…) - [ ] ○ Stanford CS221:MDP 和博弈部分可以用来打基础 — · YouTube Autumn 2025 (youtube.com/playlist?list=PL…) - 验收项目:在 Gymnasium 上实现 DQN 和 PPO。想进一步的话,用 GRPO 给一个小模型做推理能力的 RL 微调 阶段 5:Agent - [ ] ⭐ Stanford CS329Z Engineering AI Agents — · YouTube 第 1 讲 (youtube.com/watch?v=DGI4vucR…)(目前已公开 1–4 讲,后续讲座在官网陆续发布) - [ ] ⭐ Berkeley 的 Agent 公开课三部曲,按顺序看: - LLM Agents(2024 秋)— · YouTube (youtube.com/playlist?list=PL…) - Advanced LLM Agents(2025 春)— · YouTube (youtube.com/playlist?list=PL…) - Agentic AI(2025 秋)— · YouTube (youtube.com/playlist?list=PL…) - [ ] ○ Stanford CS329A Self-Improving AI Agents:偏研究 — · YouTube 第 1 讲 (youtube.com/watch?v=6YnLB0Xb…) - [ ] ○ Stanford CS222 AI Agents and Simulations:只有阅读材料,没有公开录像 — 官网 (joonspk-research.github.io/c…) - [ ] ○ Hugging Face Agents Course:动手做 — 官网 (huggingface.co/learn/agents-…) - 验收项目:完成 CS329Z 的 HW1(不用框架写一个 agent)和 HW2(评测一个 agent)
25
381
14
2,070
73,374
BaileyLi2021 retweeted
今晚别刷Netflix了。 去看这个SpaceX AI工程师1小时的课程。 这是我见过最清晰、端到端讲透GrokBot到底该怎么真正用起来的课。 从单agent开始,一路到loop、graph、多智能体协作、verification和agent-friendly架构。 不管你是从没认真跑过agent,还是已经天天在用GrokBot,看完都会把很多“原来如此”的点一次打通。 真正核心就这几层: 1. 99%的人只用1个agent,没用loop和graph,只发挥了1%实力 2. 真正的新stack是Chief of Staff + PM + 20+ workers全自主团队 3. 怎么用高质量skills让agent像真工程师一样干活(pstack) 4. 怎么把代码和架构改成agent-friendly(Dune、contract、CI) 5. 怎么用verification让自己真正敢把活交给它们,而不是一直盯着 大多数人花好几个月才拼出的全景图,这节课直接给你画完。 把时间空出来,这可能是你这个月最值的一堂课。
125
784
32
4,766
286,879
BaileyLi2021 retweeted
Jensen Huang has resigned as CEO of Nvidia 🔗 on.wsj.com/4ApW9kr
1,007
1,836
1,829
6,747
1,747,684
BaileyLi2021 retweeted
I created a Claude Skill that make beautiful slides on the web. The world hasn't woken up to the fact that code can create much better slides than most PPT tools. - Claude interviews you first about aesthetics, then generate a few directions to "show not tell", and you can pick your favorite - Cool transitions and animations - Interactive hover states and cursor effects - Auto-fits on any screen - Supports converting existing PPTX files to web-based slides; preserves original images and brand assets I asked Claude to make a slide show about this skill to showcase what it can do. Link to skill below
229
424
73
6,123
1,627,315
8 years of podcast lessons in one post: "Only take criticism from people doing the thing you're trying to do" - @melrobbins - Mel Robbins "Never get away from who you were before any of it worked" - @garyvee - Gary Vaynerchuk "You can find a way in most situations, as long as your first reaction isn't how terrible I am" @EstherWojcicki - Esther Wojcicki "Fear is curiosity you haven't turned around yet" - @reidhoffman - Reid Hoffman "You can't control whether it works. You can control whether you leave anything undone" - @luanalopeslara - Luana Lopes Lara "If we're going to fail, let's fail having tried everything, and let's fail honestly" - @bscholl - Blake Scholl Send this to someone who needs to hear it today!
4
38
1,313
BaileyLi2021 retweeted
这六个人,5个移民。 一个难民、二个H1B、另外两个美台双国籍(黄和苏)的父母是通过F1、F2和技术、人才移民。
280
362
41
3,577
619,113
BaileyLi2021 retweeted
Claude Code 工程师刚放出一条 28 分钟实操视频,国内一堆付费课教你怎么用claude code ,不如 Claude code 作者这28分钟。 给你讲清楚:CLAUDE.md、持久记忆、多会话并行、高阶 Prompt。 中英双语字幕
362
1,262
32
6,705
546,486
Replying to @ClorisSignal
真棒!喜欢😘
1
50
BaileyLi2021 retweeted
斯坦福大学课程 CS329A: Self-Improving AI Agents 课程 9 讲视频都已发布在 Youtube,第 1 讲 52 万观看,第 9 讲 1.5 万。趁假期坚持看完 9 讲视频,你就超过了中间的 50 多万人,和连第 1 讲都没看过的无数人 😂 开玩笑,虽然 AI Agent 已经强大到很多人觉得不需要再学习,什么问题都可以直接问,到顶级大学的基础课程,系统学习下来,还是会很有收获,全局认知的提升。 课程主页: cs329a.stanford.edu/ 视频列表: youtube.com/playlist?list=PL…
斯坦福大学把 CS329A Self-Improving AI Agents 视频都公布到 Youtube 了。 两位主讲老师做过 PaLM、Gemini、Claude。9 讲讲透一个循环: 采样 → 验证 → 筛选 → 训练 → 更强的模型 想知道为什么 coding agent 最新普及,写作却不行? 完整解读 👇
28
313
3
1,433
264,630
BaileyLi2021 retweeted
🙏
619
3,590
299
20,577
430,250
Harvard College Dean David Deming told members of the Class of 2030 that they were entering a “third era” of higher education — arguing AI would fundamentally change universities' roles in society. @SebastianC4784 and Summer E. Rose report. thecrimson.com/article/2026/…
1
17
6
67
17,392
BaileyLi2021 retweeted
MIT professor delivers a brilliant masterclass on how to effectively present your ideas with clarity, high impact and purpose (15 important lessons)
63
1
294
38,431
BaileyLi2021 retweeted
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models. images.nvidia.com/pdf/Open-W…
16,563
34,461
12,076
193,854
69,434,714
LeBron’s greatest skill may not be basketball. It may be knowing what each chapter of life requires. Miami was about learning how to win. Cleveland was about keeping a promise. Los Angeles was about building a larger legacy. The real question is never simply: Where should I go? It is: What must I not betray at this stage of my life? Education should not prepare young people to follow the safest path. It should prepare them to make wise choices when there is no obvious answer. That is how they become the CEO of their own life.@KingJames @NBA #Education #Leadership #LeBronJames #FutureReady #NapaAcademy
1
33
The world is in flux. Education can no longer prepare students for a fixed future. It must prepare them to thrive in constant change.
3
20
BaileyLi2021 retweeted
I’m sick of reading AI slop, so today I’m open-sourcing my /no-ai-slop skill that removes 20+ slop patterns from any piece of writing. 📌 Get the free skill here: github.com/petergyang/no-ai-… If you find it useful, please consider starring the repo so more people can find it. Why I built the skill: I use AI to edit my writing because it helps me fix spelling, grammar, and clarity. But even the best models keep producing the same slop that this skill removes: → Binary contrasts: “It’s not X. It’s Y.” → Throat-clearing openers: “Here’s what nobody tells you.” → Fake-profound endings: “The future isn’t coming. It’s already here.” Use this skill responsibly. I always write a first draft manually before iterating with AI on edits and I make sure to do another manual pass at the end as well. That’s in contrast to using AI to automate pumping out slop end-to-end. 📌 Read my full post for more on how I try to use AI responsibly to edit without giving into the dark side: creatoreconomy.so/p/use-my-n…
299
394
88
6,265
1,200,745