Wang Honwhy

👨‍💻Programmer, #Java #TypeScript #Vue #Rust #React
📜booklist,

[Wang Honwhy @honwhy19]

我们的公众号 Markdown 编辑器浏览器插件，在 Chrome 和 Edge 商店安装量都突破 1000+ 啦！一个小小的里程碑，值得纪念～🎉🎉🎉🚀
获取地址：https://t.co/jAPnlEd2t4

[Wang Honwhy @honwhy19]

@kevinzhow 好

[CodeAlpha @RealCodedAlpha]

Anthropic 官方团队把 Claude Code 从底层原理到真实项目实战，完整讲了一遍！

如果你只是会让 Claude Code 写代码，却还没搞懂它如何读取文件、调用工具、管理权限、调度 subagents，这个视频值得收藏！

我整理了重点时间线：

00:00｜Claude Code 底层如何工作
04:51｜Permissions：哪些操作允许、询问或拒绝
10:46｜如何把 Code Review、部署流程做成 Skills
15:08｜Intent-driven development 到底怎么做
22:40｜用语音 Prompt 从零构建 3D 游戏
32:55｜用 Branch、Draft PR 和 GitHub CLI 控制交付
38:15｜Auto Mode 如何减少重复授权
40:48｜用 Dynamic Workflows 并行调度多个 Subagents
46:53｜从 Demo 走向 Production 的工程原则
55:33｜Claude Code 如何用于研究、数据和媒体处理

我认为最值得看的是 Lydia Hallie 拆解了 Claude Code 的真实运行方式：模型本身没有长期状态。

文件内容、终端输出、Git 分支、工具调用结果和会话历史，都是由外层的 Agent Harness 维护，再在每一轮重新组装给模型。

这也解释了为什么真正影响 Claude Code 效果的，不只是 Prompt，还有：CLAUDE.md、Skills、Permissions、Verification、Git Workflow、Subagent 编排！

视频最后还有一个很重要的提醒：AI 能一次生成大量代码，不代表这些代码都应该被合并。

架构、验证、安全、Review 和最终质量，依然由人负责。

这可能是目前理解 Claude Code 最完整的一小时。

建议先收藏，再跟着时间线看！

[KK.aWSB @KKaWSB]

吴恩达透露，他目前几乎所有工作都交给 AI Agent 完成。

下一步，是让这些 Agent 自己发现问题、总结反馈、持续优化。

这段 视频，从零拆解了如何搭建一个具备自我改进能力的智能体系统，信息密度很高。

[Xudong Han @Xudong07452910]

现在很多人学 AI Agent，还停留在「能不能跑一个 demo」。

但真正难的可能是：这个东西上线之后，能不能长期稳定地跑。

CMU 2026 春季这门 MLiP / AI Engineering 课程，我觉得很值得关注。它讲的不是怎么训练一个模型，也不是怎么写几个 prompt，而是怎么把 ML、LLM 和 Agent 变成真正的生产系统。

里面会讲部署、测试、监控、MLOps、安全、隐私、公平性、可解释性，也包括 Agents 相关的很新的工程内容。

这个课程项目不是做一个玩具 demo，它想让学生在相对真实的生产条件下，构建、部署、评估并维护一个面向 100 万活跃用户的推荐系统。

这其实很符合接下来 AI 的发展方向。

Agent 能写代码、能调用工具、能完成任务，已经不算最稀奇的部分了。真正有门槛的是：它犯错怎么办？怎么监控？怎么回滚？怎么评估？怎么控制成本？怎么在高负载下稳定运行？

AI 工程的核心，正在从「模型能力」走向「系统可靠性」。

如果你想理解 Agent 怎么从 demo 走向真实产品，这门课可能比很多单纯追热点的 Agent 课更值得看。

https://t.co/7tbmj52Y6S

[venus @RitOnchain]

Stanford computer science professor just revealed how to master Markov Decision Processes.

83-minutes. free. By Stanford.

here's what they cover:
• search problems vs. stochastic environments
• policy evaluation & q-value recurrence math
• value iteration loop engineering
• convergence limits under a cyclic graphs

Bookmark & watch today. Then read the article below.

[📦Acbox @AcboxLiu]

Hi，我们是 Felinic 猫本原理，一个专注于 Cloud Agent 和 AI Agent Infrastructure 的团队。

我们相信，Agent 将成为下一代软件的重要交互方式，而优秀的 Agent 需要同样优秀的基础设施。我们正在寻找热爱技术、愿意一起探索未来的小伙伴加入我们。

AI Harness Engineer
希望你：对现代 Agent 的底层原理有深入理解；
熟悉 Claude Code、Codex 等 Agent Harness 的运行方式，理解 Tool Use、Context Management 等核心机制；
具备一定的模型评估（Model Evaluation）能力，对 Agent 的能力边界有自己的思考。

Infra Engineer
希望你：对虚拟化、Firecracker、Kubernetes 等基础设施技术有较深入的理解；
熟悉 Golang 或 Rust，并希望构建高性能、高可靠的 AI Infra。

Intern
如果你对 Agent、前端、设计或其他相关方向充满兴趣，并且有自己的实践项目，欢迎带着你的开源项目、作品集或其他能够展示能力的内容来找我们。

Base： 深圳 / 上海（Onsite），同时支持 Remote。

加分项
- 有自己的开源项目，或长期活跃于开源社区；
- 对 Agent Harness 有独立思考，敢于挑战现有产品形态和架构设计；
- 在 AI 时代依然保持优秀的 UI / UX Taste，关注产品体验而不仅仅是模型能力。

你将会获得
AI Tooling 全额报销（目前包含 Codex Pro 20×、Claude Max 20×）；
与一群热爱开源、热爱 Agent 的工程师一起工作，探索下一代 AI Agent Infrastructure。

如果你感兴趣，请将你的简历、GitHub Profile（或其他能够展示项目经验的作品链接）发送至 [email protected]。

期待与你一起构建下一代 AI Agent Infrastructure.

[codila @0xCodila]

Ex-Google engineer just dropped 1-hour course: loops, self-improving AI, memory systems - from scratch:

00:00 - the self-building agent
03:01 - soul.md runs everything
30:16 - RAG memory: pull 20 messages, not 2,000
31:48 - the loop that knows when to stop
35:14 - find the bug, fix the prompt
50:22 - how Claude compresses your memory

1 hour of his guide beats any paid agent course

watch & bookmark - then read Karpathy's loop method below

[Codez @0xCodez]

Andrej Karpathy just dropped a 6-hour course on how to build LLMs from scratch:

• 00:00 - Deep dive into LLMs like ChatGPT
• 03:31:23 - Building ChatGPT from scratch in live
• 05:27:43 - How to use LLMs (Karpathy method)

This course will replace a $90K Stanford LLM master’s degree.

Start watching today, then read how to become an AI engineer in article below.

[Yangyi @yangyi]

这个博主叫real屠龙的胭脂井
在复旦的这场分享里 对品牌营销的拆析鞭辟入里
一看就是经过大量减法后找到的更加底层的东西
不知道这些学生会不会在很多年后才能理解这场分享的价值

这个博主是有脑科学和认知科学背景的 我觉得这在AI时代至关重要 尤其是利用建模来辅助决策
我认为做任何事都应该这样 做营销更应该计算数据

但怪就怪在 大部分从业人员 都缺少这种最基础的认识
往往依靠一个错误的虚荣指标指导自己进行愚蠢的工作

[Akshay 🚀 @akshay_pachaar]

14x faster and 90% cheaper LLM inference.

(100% open-source, KV cache management)

your LLM does the same expensive work over and over.

every request, it re-reads the same system prompts and the same documents from scratch, even if it processed them one second ago. token prices keep falling, but agent workloads re-send so much repeated context that the bill climbs anyway.

LMCache fixes this. it's an open-source KV cache management layer that plugs into vLLM, SGLang, and TensorRT-LLM.

here's how it works:

LLMs recompute their understanding of the same content on every request. the same system prompts, the same documents, processed from scratch every time, and a single GPU throws away roughly 15 TB of this reusable cache per day.

LMCache stores that cache and serves it back on repeat requests, running as a separate process completely outside the inference engine.

the engine just asks for the cache blocks it needs. LMCache handles all the heavy data movement across GPU, CPU, disk, and remote storage in parallel, so cache work never steals compute from inference.

it also reuses cache beyond exact prefixes. their CacheBlend technique (EuroSys 2025 best paper) keeps RAG documents cached no matter what order they appear in.

on H200s with a 235B model, that adds up to 14x faster time-to-first-token and 4x faster decoding. and since reuse skips the compute entirely (the same reason providers discount cached tokens by 90%), the cost savings follow directly.

GitHub repo: https://t.co/TXlaLLu04a

(don't forget to star 🌟)

i wrote a full breakdown of KV cache management that walks through why 𝗽𝗿𝗲𝗳𝗶𝘅 𝗰𝗮𝗰𝗵𝗶𝗻𝗴 silently breaks in three common cases, the 𝗱𝗶𝘀𝗮𝗴𝗴𝗿𝗲𝗴𝗮𝘁𝗲𝗱 𝗮𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲 behind the 14x speedup, 𝗖𝗮𝗰𝗵𝗲𝗕𝗹𝗲𝗻𝗱, and how to turn every document in your knowledge base into a reusable cached asset.

the article is quoted below.

[Roan @RohOnChain]

My friend makes $850,000/year as an Anthropic AI engineer.

No MIT. No Stanford. No PhD.

I asked him how he broke into AI from scratch.

He sent me a course that was never supposed to get out. A developer teaching to build self-improving AI agents from scratch.

You will learn exactly how AI agents like Claude Code and OpenClaw are actually built.

You won't find anything better about becoming an AI engineer in 2026 than this video.

I watched it last night.

Halfway through, I realized I could land a role at a top AI lab in months, not years.

Bookmark this and read the article below.

[AI Will @FinanceYF5]

Google 近日发布了一门从零开始讲 Agentic Engineering 的 1 小时课程：

– 如何构建你的第一个 AI Agent
– 构建 Agent 记忆
– Agentic Loop，以及长时间运行的 AI Agent
– 如何构建 MCP：MCP 与 API 的区别
– 多 Agent 系统

足以替代网上 10 门付费 Agentic 课程。

[Dwarkesh Patel @dwarkesh_sp]

Adam Brown (@A_G_I_Joe) is back!

General relativity is said to be the most beautiful idea the human mind has ever produced.

Most of us will never get to fully appreciate its elegance by taking the 20-lecture graduate course Adam taught on it at Stanford.

But in the video below, Adam distills the key idea at its heart so clearly and compellingly that even I could keep up lol.

At the core of general relativity, Einstein is trying to figure out the principle behind a particular coincidence: that the mass that resists acceleration and the mass that gravity pulls on just happen to be exactly the same. Adam then leads us through the path of insight which Einstein called his “happiest thought.”

Then Adam lectures on black holes. First, by showing how even under special relativity you could create a perpetual motion machine if black holes weren't truly black. And then, by explaining why the observations of an infalling observer and a distant bystander to the black hole would be so radically different

Adam leads Blueshift, the team at Google DeepMind cracking science and reasoning.

Which gave us the opportunity to discuss at the very end how close we are to AIs that could rediscover general relativity from scratch. Stay till the close for some philosophy of science.

0:00:00 – The coincidence that led Einstein to general relativity
0:16:42 – Gravity is a consequence of curved spacetime, not a force
0:31:46 – Why black holes prevent unlimited energy extraction
0:47:12 – Black holes are the ultimate power plants
1:13:50 – What falling into a black hole would actually feel like
1:18:51 – The three ways we know black holes are real
1:24:21 – The first time we saw gravity bend light
1:29:33 – How far can AI get without experimental evidence?

Look up Dwarkesh Podcast on YouTube/Spotify to watch. Enjoy!

[Wang Honwhy @honwhy19]

教线性代数的老爷子加入了Anthropic

[h100envy @h100envy]

Ex-NVIDIA engineer who built Unsloth explained RL, kernels, reasoning, quantization, and agents in 2 hours 42 minutes - better than $5000 fine-tuning bootcamps.

pick the base model -> write triton kernels for 2x faster fine-tune -> quantize to 4-bit -> run GRPO/DPO -> ship a reasoning model on your single GPU.

That loop is why Unsloth is the default way to fine-tune Llama, Qwen, Gemma, and Phi on hardware you already own.

Unsloth + Triton kernels + 4-bit quantization + GRPO/DPO + single-GPU fine-tuning - that's the stack.

Watch and save it, then fine-tune your first model tonight.

[Roan @RohOnChain]

My friend applied to 200 tech jobs in two years. No PhD. No Stanford.

Last month Anthropic offered him $750,000.

I asked him how he broke in from zero.

He sent me a course that was never supposed to get out. A 3-hour video to build a full LLM from scratch.

A developer teaches you exactly how LLMs like ChatGPT and Claude are actually built.

I watched it last night.

Halfway through, I realized it's embarrassingly simple to break into an AI lab.

Bookmark this and read the article below.

• 00:00 - intro to LLMs
• 05:43 - LLM transformer architecture
• 40:24 - training the LLM
• 1:30:27 - modernizing the LLM
• 2:33:53 - scaling the LLM

[huangserva @servasyy_ai]

Claude Code 团队刚刚发布了一门关于使用 Fable 5 进行循环工程的免费课程：

[00:00] —— Claude Code 的底层工作原理
[05 :01] —— 智能体循环机制解析
[16 :21 ] —— 99% 开发者忽略的功能：自动模式
[19 :00] —— 为什么语音优于打字
[32: 34] —— 使用草稿合并请求进行自动代码审查
[58: 39] —— 将 Fable 5 用于非编程工作

这门免费课程可以替代付费的 Claude Code 教程

[智享 @CycleDecoded]

离谱了！搜狐直接把自家压箱底的 Redis 云平台给开源了。这波算是把后端老哥的饭碗给端了级别的自动化。

项目叫 CacheCloud，是搜狐视频内部扛过每天 800 亿次请求、18T 内存的怪物。现在直接扔在 GitHub 上，狂揽近 9000 星，完全开源（Apache-2.0 协议）。以前只有大厂才玩得起的底层架构后台，现在小团队也能直接抄作业。

傻瓜式搭建：单机/哨兵/集群一键搞定，告别手搓命令。

监控直接拉满：机器状态、命令曲线、慢查询报警一眼看穿。

弹性无感扩容：流量扛不住了直接加资源，业务稳如老狗。

白嫖党狂喜：企业级私有云平台，零成本直接本地部署。

🔗 GitHub 传送门：https://t.co/QJ0uhI1l6x

[Geek @geekbb]

Hugging Face 出的极简智能体，手把手教你从零造一个属于自己的 coding agent。学习写 agent 的入门玩具。

https://t.co/7nBeX7B2DE

[astaxie @astaxie]

各位做智能体研发的都可以深度学习一下这一篇文章《Harness Engineering for Self-Improvement》 https://t.co/Fg3fJl9RE8