Free 75-minute live masterclass explains how LLMs become AI agents
The course unpacks the difference between LLM applications, workflows and true agentic systems, a distinction many teams still blur when choosing between building from components and adopting LangChain or CrewAI. Learners get a framework-free mental model for designing, evaluating and debugging their own agents. Anyone shipping agent features in ASEAN startups or the wider Chinese-speaking dev community can join the free session without paying for a paid course first.
A 75-minute free live masterclass at ragpack.ai/masterclass teaches how reasoning, memory, tools, planning and evaluation combine into an agentic system without relying on any agent framework.
The instructor line-up includes Jay Alammar and Maarten Grootendorst, co-authors of "Hands-On Large Language Models" and the new O'Reilly book "An Illustrated Guide to AI Agents", plus Luis Serrano, Josh Starmer of StatQuest and Chris McCormick.
The six-lecture curriculum covers what defines an agent, reasoning techniques such as Chain-of-Thought and "search against verifiers", memory and tool use, planning and reflection, agent evaluation, and what frameworks like LangChain and CrewAI actually package.
The evaluation lecture stresses assessing decision trajectories and safety, not just final answers.
到底是什么把 LLM 变成了 AI Agents ? 这门 75 分钟的免费直播大师课，不依赖任何 Agent 框架，把 "推理、记忆、工具、规划、评估" 五个组件如何组合成一个真正的 agentic 系统，讲成一个连贯的整体。 报名地址： ragpack.ai/masterclass 5 位老师堪称 AI 教育领域的“梦之队”： · Jay Alammar @JayAlammar & Maarten Grootendorst @MaartenGr — 《Hands-On Large Language Models》与 @OReillyMedia 新书《An Illustrated Guide to AI Agents》的共同作者；Grootendorst 还是 BERTopic/KeyBERT 的创建者 · Luis Serrano @SerranoAcademy — 前 Google/Apple AI 科学家，《Grokking Machine Learning》作者 · Josh Starmer @joshuastarmer — StatQuest 创始人 · Chris McCormick @ChrisJMcCormick — 以 BERT/Transformer 深度实操教程著称 # 课程结构解读：六讲的内在逻辑 课程的设计有一条清晰的递进主线，从“是什么”到“怎么想”到“怎么评”再到“用什么造": 第 1 讲: 什么才算 Agent? 从 LLM 的固有局限出发，给出核心判据：与环境交互、维护状态、自主决定下一步。并做三层区分，LLM 应用 → LLM 工作流 → 真正的 agentic 系统。这是目前行业内最常被混淆的概念，也是整门课的地基。 第 2 讲: 为什么推理重要? 讲推理型 LLM 与推理技术(Chain-of-Thought、"search against verifiers" 等)如何让模型能“想清楚问题、选对动作”。"search against verifiers"(带验证的搜索式推理)是近年推理研究的核心范式，出现在大纲里说明内容相当新。 第 3 讲: 记忆 + 工具，从“说”到“做” 记忆解决跨交互的上下文延续；工具让模型能检索信息、调 API、跑代码、真正影响环境。这两者是 Agent 区别于聊天机器人的分水岭。 第 4 讲: 规划、反思与自主性 问题分解、动作排序、根据反馈调整、复盘结果、修正计划。这是 agent 行为的“执行层”。 第 5 讲: Agent 评估 这一讲值得特别留意：它强调不能只看最终答案是否令人信服，而要评估结果、决策/动作轨迹、可靠性与安全性。“评估轨迹而非只评估输出”正是当前 Agent 工程的共识性难点，把它列入入门课大纲，反映了课程的专业度。 第 6 讲: Agent Harness 讲清 LangChain/CrewAI 这类框架到底打包了什么、替你做了什么、代价是什么。整体立场是“从第一性原理理解，而非依赖框架”，这一讲是那个立场的落点。 Jay Alammar @JayAlammar Join the us for a free masterclass on AI Agents next week! Together with @MaartenGr @joshuastarmer @ChrisJMcCormick @SerranoAcademy we have been cooking a gentle introduction to AI agents. 📅 Sept 29, 2026 ragpack.ai/masterclass R 🔗 View Quoted Tweet📊 1 ⚡
This report is categorized under AITOP's "How-to" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include AI agents, Jay Alammar, LangChain, CrewAI.
Source attribution: curated from shao__meng (@shao__meng).
