Agentic Lessons
Learn one move
Focused lessons for each level, in plain English for non-coders. Read one, try it, and come back when you need the next move. Each card stays short enough to use between real work sessions.
AI Is Not a Search Engine
What these tools actually do behind the scenes
Your First Productive AI Conversation
Stop testing. Start asking real questions
Building a Tab Complete Habit
From occasional use to daily instinct
Thinking in Agentic Sessions
How to structure your work for AI-assisted output
Context Engineering Fundamentals
Your context window is RAM. Treat it like it.
Scaling Context Across Projects
Managing what the model knows at scale
Compounding Engineering Patterns
How your workflow gets faster every week
How MCP Servers Work
Tools your AI can actually call
Harness Engineering 101
Building systems that make AI reliable
Running Background Agents
Async AI work that doesn't need you watching
Designing Autonomous Agent Teams
Multi-agent systems that coordinate on their own
Your First Useful Output
Get a useful, specific response from AI on your first...
Context Is Your Superpower
Write a context-rich prompt that gets a noticeably...
What AI Can't Know (And How to Check)
Identify which parts of an AI response to trust and...
Tell Claude Who You Are
Set up a personal profile so your AI tool already...
Ask Claude to Remember Things
Ask Claude to remember specific things so you don't...
Ask Like a Real Human
Write a prompt that gives Claude enough context to...
The Refinement Loop
Read a Claude response critically and do one targeted...
When to Switch Tools
Decide in under ten seconds whether to use Claude,...
Find Your Daily Use Tipping Point
You'll have identified one recurring task in your...
Set Up Your Persistent Workspace
You'll have a one-time setup in your AI tool of...
Write Stubs the Model Can Finish
Write function stubs that reliably produce useful...
Read Before Tab
You'll have a five-second scan habit that catches...
Pre-Load the Context
You'll know how to arrange open files before you...
Graduate to Chat
Identify when tab-complete is the wrong tool for your...
Share a Document with Claude
Drop any document into Claude and ask focused...
Know What Claude Sees in Your File
Tell in advance whether a file will give Claude...
Edit Without Rewriting
Ask Claude for surgical changes to an uploaded...
Tell the Model What to Look At
Use @ context references to point the model at the...
Plan Before You Code
Use plan mode to catch a model misunderstanding...
Write Your First CLAUDE.md
You'll have a working CLAUDE.md with your first five...
Do Not Stuff CLAUDE.md
Distinguish rules that belong in CLAUDE.md from...
Audit Your Token Budget
Count what's actually eating your context window and...
Prune Like a Pro
Identify context poisoning in a failing session and...
Write a CLAUDE.md That Earns Its Tokens
You'll have a CLAUDE.md that prevents your most...
Trust No Retrieved Document
Explain prompt injection through retrieved content...
Checkpoint and Clear
Use /clear and session checkpoints as a deliberate...
Build Your First RAG Pipeline
You'll have a working RAG pipeline on a real document...
Structured Output as Contract
Enforce structured output from any prompt that feeds...
Retrieve vs Fit: The Decision Rule
Apply a concrete decision rule to any document or...
Label Your Sources, Block Injection
Build a multi-source context pipeline that maintains...
The Codify-Loop
You'll run the plan-delegate-assess-codify loop as a...
Track Three Numbers
Measuring acceptance rate, iteration count, and...
Prune the Rules File
Run a structured audit of your CLAUDE.md, identify...
Session Log vs Rules File
You'll maintain two separate artifacts from every...
Wire Your First MCP
Install a real MCP server, connect it to your agent,...
Read a Tool Schema Like the Model Does
Read an MCP tool schema and predict when the model...
Package Your First Skill
Turn a repeatable 3-step workflow into a named skill...
Trust No Tool Response
Identify prompt injection through MCP tool responses...
Track Cost Per Iteration
Measure the token cost of a skill or MCP workflow,...
Measure the Design, Don't Eyeball It
Turn 'looks close enough' into a number you can defend
Build the Inner Loop
Wire a feedback loop where the agent runs tests,...
Spec-as-Test
Write the expected output before the agent runs,...
Add Structured Logging
Instrument an agent run with structured JSON logs and...
Build the Checkpoint and Replay
Add a checkpoint to an agent workflow so that a...
Define the Blast Radius
Classify any task into one of three risk tiers, add...
Your First Background Agent
Kick off one low-risk task as a background agent,...
Cross the Diff-vs-Code Threshold
Identify whether you've crossed the threshold where...
Three-Tier Setup
Describe the human/supervisor/worker architecture,...
Manage the Money
Set per-run budgets, route tasks to the right model...
Stale Context Is Not a Stale Bug
Describe the stale context coordination problem,...
Hosted or Self-Hosted
Classify any background agent task as hosted-suitable...
Hub-and-Spoke vs Peer-to-Peer
Distinguish hub-and-spoke from peer-to-peer agent...
Watch for Emergent Failures
Name the three main emergent failure modes in...
Reproducibility Is the Architecture
Explain why reproducibility is a structural...
When NOT to Deploy a Team
Apply a four-question decision test to any candidate...
Document the Frontier
Write a structured failure or success report from one...