References: Planning: breaking a goal into steps

Source material

Source curriculum (structural mirror, cited as further study):
• Microsoft, "Planning Design" (AI Agents for Beginners, Lesson 07)
  Author: Microsoft Cloud Advocates
  Lesson page: https://github.com/microsoft/ai-agents-for-beginners/tree/main/07-planning-design
  License: MIT
• Berkeley CS294/194-196, "LLM Reasoning" (Lecture 1)
  Speaker: Denny Zhou (Google DeepMind)
  Course page: https://rdi.berkeley.edu/llm-agents/f24
  Video: https://www.youtube.com/watch?v=QL-FS_Zcmyo
  License: YouTube standard (link-out only)
Clawdemy's lessons are original prose that follows the pedagogical arc of these
sources. We do not reproduce or transcribe them; we cite them as recommended
companions. All rights to the original materials remain with their authors.

Read or watch this next

• Planning Design (Microsoft) by Microsoft Cloud Advocates. The practitioner version of this lesson, including task decomposition, structured (JSON) plans, and iterative replanning, with runnable samples. MIT-licensed.
• LLM Reasoning (Denny Zhou, Berkeley CS294 Lecture 1). The reasoning research behind planning: why prompting a model to produce intermediate steps, rather than jump to an answer, improves how it handles complex problems. The “think step by step” idea this lesson builds planning on.

Going deeper on reasoning and decomposition

A short, durable list. Each is a primary source for a reasoning technique planning draws on.

• Chain-of-Thought Prompting Elicits Reasoning in Large Language Models (Wei et al., 2022). The paper that showed prompting for intermediate steps improves complex reasoning. The foundation under “ask for the steps, not just the answer.”
• ReAct: Synergizing Reasoning and Acting in Language Models (Yao et al., 2022). Interleaving reasoning steps with actions, the pattern behind plan-execute-observe-replan loops.

Adjacent topics

Where this leads inside this track.

• Many agents working together: multi-agent systems. The next lesson. A plan’s sub-tasks are often separate enough to hand to different agents; the next lesson covers when splitting work across agents beats one agent, and the coordination cost.
• Agents that retrieve their own information: agentic RAG. The previous lesson. Planning and retrieval combine: a planned step is often “go find out X,” which is a retrieval the agent runs mid-plan.
• How tool use turns a model into an agent. Earlier in the track. Each executed step of a plan is a tool call; planning decides which tool calls to make and in what order.