Free Email Course

Claude Code Analytics

23 emails over 9 weeks. Build an AI product analyst in Claude Code that runs on your data, validates its own work, and gets smarter every week you use it.

Week 1

Daily

The Foundation

Build the core system from scratch: the config file, your first question, the orchestration pipeline, agents, skills, corrections, and validation.

Day 1

The one file that makes the whole thing work

It's not code. It's plain English.

Day 2

Clone it tonight and ask it something you already know

Three commands. Then ask it a question you can verify.

Day 3

What just happened behind the scenes

18 agents across 8 phases.

Day 4

You don't have one AI analyst. You have eighteen.

Each one is a markdown file you can read.

Day 5

The guardrails that fire without you asking

39 skills. You didn't invoke any of them.

Day 6

Every mistake makes it smarter

Wrong SQL today. Better SQL forever.

Day 7

Four layers between the AI and a wrong number

Structural. Logical. Business. Simpson's.

Week 2

Twice weekly

Exploration and Analysis

How the system discovers your data schema and runs hypothesis-driven analyses across segments, trends, and cohorts.

Day 8

The agent that makes first contact with your data

It profiles every column before you ask.

Day 9

It frames the question before it writes the query

Four analysis agents. One design spec.

Week 3

Twice weekly

Storytelling and Visualization

Turn validated findings into narratives with tension and resolution, and charts that serve the story.

Day 10

Numbers don't convince anyone

Story beats before charts.

Day 11

The chart the data wants to be

Gray everything. Then highlight one thing.

Week 4

Twice weekly

End-to-End Pipelines

Run a complete analysis through all 18 agents and learn how corrections compound over time.

Day 12

Question in. Report out. Every step visible.

One command runs the whole thing.

Day 13

The most useful thing your AI analyst does is fail

Corrections change future behavior.

Week 5

Twice weekly

Real Data and Knowledge

Connect to production databases and build persistent knowledge files so the system remembers your business context.

Day 14

CSV was practice. Now plug in the real thing.

Six connection types, one wizard.

Day 15

Stop explaining your business every time you ask a question

The .knowledge/ directory remembers.

Week 6

Twice weekly

Routing and Parallelism

How the question router classifies effort across five levels and agents run in parallel through the DAG.

Day 16

Not every question deserves the same effort

Five levels. Thirty seconds to 45 minutes.

Day 17

Run three analysts at once

AND gates, OR gates, and a circuit breaker.

Week 7

Twice weekly

Validation and Teams

Simpson's Paradox detection, confidence scoring, and scaling the system from one person to your whole team.

Day 18

The number was right. The conclusion was wrong.

Aggregate trends can lie.

Day 19

From your system to your team's system

One repo. Shared definitions. No drift.

Week 8

Twice weekly

Performance and Recovery

Context engineering for speed as your repo grows, and checkpoint-based pipeline recovery.

Day 20

How to keep it fast as your repo grows

Three tiers, one token budget.

Day 21

Checkpoints, reruns, and not starting over

Resume from where it broke.

Week 9

Twice weekly

The Big Picture

The genome concept for rebuilding from scratch, and where AI analytics goes from here.

Day 22

One file that rebuilds everything

Config is DNA. Code is derived.

Day 23

The next six months matter more than the last five years

The window is open. Move now.

Start building your AI analyst

Free. No spam. Day 1 lands in your inbox right after you sign up.