Claude Code for Data: Your Judgment, at Scale
An agent is a new tool that uses tools. The expertise is yours, and that's what makes you irreplaceable.
You go at your own pace, with your Maestro beside you, 1:1, through the whole course: ask it anything, and it asks you questions too.
61 skills · 10 modules
$1
I'm opening the course one module at a time. When every module is out, the course will be $149. What you pay today covers all of them.
Comes with your Maestro, 1:1, for the whole course.
What will you be able to do?
Each box is a skill. The base is at the bottom, recognizing and explaining, and you work your way up to building something of your own and judging what works and why. Hover over a box to see what it builds on. Click it to see where I teach it and how your Maestro checks it.
- Evaluate
- Synthesize
- Analyze
- Apply
- Understand
- Remember
Evaluate
Assess whether a data source is fit for the question, given its coverage, collection method, and known gaps
Builds on:
- Know what data provenance is: where a dataset came from, how it was collected, and what was done to it
- Turn a stakeholder request into a question the available data can answer
Your Maestro asks you to judge when it fits and why and saves it as evidence when you do.
Answer 'how do you know?' for any number in your work, with evidence
Builds on:
- Trace a number in a deliverable back through the agent's work to the source data
- Verify an agent's result by a path the agent did not use
- State what a result does and does not support
Your Maestro asks you to judge when it fits and why and saves it as evidence when you do.
Synthesize
Write a quality bar for a type of data work as criteria that a person or an agent can check
Where I teach it: Put It in Writing · Your quality bar
Builds on:
- Surface your own standards by examining work you accepted, work you rejected, and the reason for each
- Name what makes data work defensible: traceable, reproducible, validated, and honest about its limits
Your Maestro asks you to build something of your own with it and saves it as evidence when you do.
Write a methods note that lets a reader understand and check how a result was produced
Builds on:
- Record the decisions made during an analysis and the reason for each
- Trace a number back through its transformations to its source data
- State what a result does and does not support
- Write a metric definition precise enough that two people would compute the same number
Your Maestro asks you to build something of your own with it and saves it as evidence when you do.
Produce a methods note from the evidence trail with an agent, and verify it against what was done
Builds on:
- Write a methods note that lets a reader understand and check how a result was produced
- Have agent work leave evidence as it goes: queries run, row counts, check results, and decisions made
Your Maestro asks you to build something of your own with it and saves it as evidence when you do.
Analyze
Review an agent's output against the task specification before accepting it
Where I teach it: Ask Like an Expert, Review Like an Expert · Review the result against your specification
Builds on:
- Specify a task for an agent with its goal, inputs, constraints, and what done looks like
- Read the record of what an agent did: the commands it ran, the files it changed, and the results it saw
- Know the common failure modes of language models in data work: fabricated values, plausible but wrong logic, silent assumptions, and overconfident summaries
- Spot-check a result by recomputing a few rows by hand, spread across the start, middle, and end
Your Maestro asks you to break a problem down with it and saves it as evidence when you do.
Surface your own standards by examining work you accepted, work you rejected, and the reason for each
Where I teach it: Put It in Writing · What you accept and what you send back
Builds on:
- Recognize that much of your expertise is tacit: standards and habits you apply without stating them
Your Maestro asks you to break a problem down with it and saves it as evidence when you do.
Decide what data an agent may see, and keep sensitive or regulated data out of contexts where it does not belong
Where I teach it: With Your Data · Before opening your data
Builds on:
- Give an agent the least access it needs, and separate read access from write access
Your Maestro asks you to break a problem down with it and saves it as evidence when you do.
When an agent gets something wrong, change the environment so that the error does not repeat
Where I teach it: With Your Data · What went wrong goes in the file
Builds on:
- Review an agent's output against the task specification before accepting it
- Write persistent context that an agent reads at the start of every session
Your Maestro asks you to break a problem down with it and saves it as evidence when you do.
Reconcile a result against an independent source or a previously trusted figure, and explain any difference
Builds on:
- Check a result against expectations: order of magnitude, totals, ranges, and known benchmarks
- Trace a number back through its transformations to its source data
Your Maestro asks you to break a problem down with it and saves it as evidence when you do.
Verify an agent's result by a path the agent did not use
Builds on:
- Reconcile a result against an independent source or a previously trusted figure, and explain any difference
- Review an agent's output against the task specification before accepting it
Your Maestro asks you to break a problem down with it and saves it as evidence when you do.
Apply
Run a first agent session on a real data task and inspect what the agent did
Where I teach it: The Agent on Your Computer · Your first session
Builds on:
- Distinguish the model, the harness, and the context in an AI agent
- Know how agents reach tools and data: commands, connectors, files, and credentials
Your Maestro asks you to use it on a new case and saves it as evidence when you do.
Check a result against expectations: order of magnitude, totals, ranges, and known benchmarks
Where I teach it: What the Agent Decided for You · Three quick checks
Your Maestro asks you to use it on a new case and saves it as evidence when you do.
Specify a task for an agent with its goal, inputs, constraints, and what done looks like
Where I teach it: Ask Like an Expert, Review Like an Expert · Specify the task
Builds on:
- Explain how an agent works: it reads context, calls tools, observes results, and repeats
- Understand that an agent knows only what is in its context, and that context is limited and does not persist by default
Your Maestro asks you to use it on a new case and saves it as evidence when you do.
Write persistent context that an agent reads at the start of every session
Where I teach it: Put It in Writing · A file the agent reads when it starts
Builds on:
- Understand that an agent knows only what is in its context, and that context is limited and does not persist by default
Your Maestro asks you to use it on a new case and saves it as evidence when you do.
Write down the domain knowledge an agent needs: business rules, known data quirks, and vocabulary
Where I teach it: Put It in Writing · What you know about your data, in writing
Builds on:
- Recognize that much of your expertise is tacit: standards and habits you apply without stating them
- Write persistent context that an agent reads at the start of every session
Your Maestro asks you to use it on a new case and saves it as evidence when you do.
Write a metric definition precise enough that two people would compute the same number
Where I teach it: With Your Data · What your judgment is made of
Builds on:
- Know the parts of a metric definition: population, numerator, denominator, time window, and exclusions
- Identify the grain of a table: what one row represents
Your Maestro asks you to use it on a new case and saves it as evidence when you do.
Understand
Understand AI in data work as an extension of a person's expertise, with the person accountable for the result
Where I teach it: The Agent on Your Computer · Your judgment, at scale
Your Maestro asks you to explain it in your own words and saves it as evidence when you do.
Distinguish the model, the harness, and the context in an AI agent
Where I teach it: The Agent on Your Computer · Model, Harness, and Context
Your Maestro asks you to explain it in your own words and saves it as evidence when you do.
Explain how an agent works: it reads context, calls tools, observes results, and repeats
Where I teach it: The Agent on Your Computer · Model, Harness, and Context
Builds on:
- Distinguish the model, the harness, and the context in an AI agent
Your Maestro asks you to explain it in your own words and saves it as evidence when you do.
Understand that an analysis is an argument: a claim supported by evidence that someone else can examine
Where I teach it: The Agent on Your Computer · An analysis is an argument
Your Maestro asks you to explain it in your own words and saves it as evidence when you do.
Recognize that the choices made during analysis can change the result, and that each choice needs a reason
Where I teach it: What the Agent Decided for You · Two agents, two tables
Builds on:
- Understand that an analysis is an argument: a claim supported by evidence that someone else can examine
Your Maestro asks you to explain it in your own words and saves it as evidence when you do.
Read the record of what an agent did: the commands it ran, the files it changed, and the results it saw
Where I teach it: What the Agent Decided for You · What did the system do?
Builds on:
- Explain how an agent works: it reads context, calls tools, observes results, and repeats
Your Maestro asks you to explain it in your own words and saves it as evidence when you do.
Remember
Know how agents reach tools and data: commands, connectors, files, and credentials
Where I teach it: The Agent on Your Computer · How It Reaches Your Data
Your Maestro asks you to recognize and recall it and saves it as evidence when you do.
Know the common failure modes of language models in data work: fabricated values, plausible but wrong logic, silent assumptions, and overconfident summaries
Where I teach it: What the Agent Decided for You · Two agents, two tables
Your Maestro asks you to recognize and recall it and saves it as evidence when you do.
Name what makes data work defensible: traceable, reproducible, validated, and honest about its limits
Where I teach it: Put It in Writing · Your quality bar
Your Maestro asks you to recognize and recall it and saves it as evidence when you do.
Know the parts of a metric definition: population, numerator, denominator, time window, and exclusions
Where I teach it: With Your Data · What your judgment is made of
Your Maestro asks you to recognize and recall it and saves it as evidence when you do.
Know what data provenance is: where a dataset came from, how it was collected, and what was done to it
Where I teach it: With Your Data · Your session, with your data
Your Maestro asks you to recognize and recall it and saves it as evidence when you do.
Who is this course for?
For people who already work with data and want to give an AI agent work in their terminal while remaining accountable for the numbers they sign off on. You don't need to know statistics: knowing your data and your team is enough. You'll practice with the Stack Overflow survey and NASA's exoplanet catalog.
Ask your Maestro
Your Maestro is an AI that knows each course, page by page. Try it here free, in the free course "Claude Code for Data People: Your First Session".
You: Where do I start?
Your Maestro: From what you already know. Tell me what you've done with data, and I'll start this page from there.
Asking about: Answers from the page "A skill that checks your table before it counts"
Only this browser keeps what you ask here. How it works
What we'll cover
The Agent on Your Computer
- Your judgment, at scale
- Set up your environmentreference
- What Would the Agent Do?reference
- Model, Harness, and Context
- How It Reaches Your Data
- Your first session
- An analysis is an argument
What the Agent Decided for You
- Two agents, two tables
- What did the system do?
- What is a row?
- Three quick checks
- You made the decision
Ask Like an Expert, Review Like an Expert
- What the agent cannot know
- Specify the task
- Review the result against your specification
- With a short instruction and with a specificationreference
Put It in Writing
- What you know without saying it
- What you accept and what you send back
- A file the agent reads when it starts
- What you know about your data, in writing
- Your quality bar
With Your Data
- Now with your datareference
- What your judgment is made of
- The question you can answer
- Before opening your data
- Your session, with your data
- What went wrong goes in the file
Verify It Another Way
6 pages to come
Opens as soon as it's ready.
- If the agent checks its own work
- Five rows by hand
- An answer you already know
- What does this code count?
- Where does the number come from?
- Using a path the agent did not take
When Two Tables Come Together
4 pages to come
Opens as soon as it's ready.
- Three tables, one store
- Profile before counting
- The join that duplicates
- The missing ones
What Your Number Doesn't Tell You
6 pages to come
Opens as soon as it's ready.
- Do I need to know statistics?
- Who answered?
- More than one hundred percent
- The method that fits the question
- Change one decision
- The sentence that states the limits
From a Session to a Procedure
7 pages to come
Opens as soon as it's ready.
- What repeats
- Same question, different answer
- The agent writes, Python calculates
- From session to script
- Tests for your data
- Package the procedure
- One agent reviews another
Defend Your Analysis
5 pages to come
Opens as soon as it's ready.
- “How do you know?”
- Follow a number
- The methods note
- Write the note with the agent
- Final project: how do you know?
How long does it take?
3 to 5 hours
A guide: you go at your own pace, and the course waits where you left off.
Common questions
- What does an AI agent do with my data?
- An AI agent is a model that uses tools: it reads files, runs commands in your terminal, looks at the result and repeats. With data, it does the basic analysis well on its own. What the data does not tell you is your call, and so is which number you sign off.
- Is this a Claude Code course?
- Yes. You learn to hand Claude Code your data work and to check what it did. Every instruction also comes for Codex CLI and Antigravity CLI, so the course works with any of the three. Claude Code needs a Claude Pro plan or higher, or API credits. Your Maestro helps with whichever you choose.
- Is there a free version?
- Yes. Claude Code for Data People: Your First Session is free. You install Claude Code, run your first session on NASA's exoplanet catalog, check an Excel file, and write your first CLAUDE.md and your first skill. What you show there counts here, so your Maestro won't start you from scratch.
- How do I review what an AI agent did?
- Use the check this course repeats from start to finish: what was counted and what was divided by. Then read the agent log, with the commands it ran and the files it changed, and compare the result with what you expect: order of magnitude, totals and ranges. You decide whether to accept it.
- Why does an agent like Claude Code code well but get data wrong?
- When you program, tests warn you when something breaks. With data, nothing does: the agent assumes what a row is, which rows to drop or what to divide by, and a wrong number still looks fine. So you ask for evidence: what it decided, which rows it used, what code it ran. Then check it against the data.
- Is it a one-time payment, and can I get a refund?
- It is a one-time payment, no subscription. You have 7 days from purchase to ask for a full refund, for any reason, by writing to hola@heymaestro.app. The current price appears next to the buy button. Your Maestro, a 1:1 AI, answers your questions throughout the course.
- Do I need a course, or can I learn this on my own?
- You can learn it on your own. The course gives you an ordered path with real data, exercises that test your judgment, and your Maestro answering questions in every lesson. Unsure? Start with a free course such as Your AI and Your Data. If it is not for you, you have 7 days to ask for a refund.
- Do I need to know statistics?
- No. What matters most when you review a result is understanding the question, the group it refers to and how the data was defined. Knowing your data and your team is enough. If you also know statistics, you use it in your criteria. If not, the course does not assume it.
- Is it worth learning this if AI tools change so fast?
- Yes, because what the course teaches does not depend on one tool: specifying a task, reviewing the result and writing down your judgment so the agent follows it. The tools are the example, and they change. Your judgment about your data is what stays with you.
- Do I need to know how to code or use the terminal?
- You don't need to know how to code. When you set up your environment, the course shows you how to open a terminal and gives you every command, with PowerShell versions for Windows. You don't write Python, you read it, and if a line is unclear you ask the agent. You work with a terminal agent: Claude Code, Codex CLI or Antigravity CLI.
- What data do I practice with?
- You practice with two public tables: the 2024 Stack Overflow survey, which you download when you set up your environment, and NASA's exoplanet catalog. In Module 5, "With Your Data", you use a table from your own work, for example an exported CSV. If you can't take data out of your work, you pick a public table you know well.
How do I check what AI does? I teach it in Your AI and Your Data: Don't Let It Make Up the Number and Check what AI gives you in three steps.
What you get
- 10 modules with 55 pages. I'll open the next modules as each one is ready.
Your Maestro for the whole course, who always knows which page you're on: you open it from there.
It works from this course's skills, the ones above.
Ask it how you're doing, and it tells you what you've already shown and what comes next.
Words you'll meet
In the lessons, tap one to see what it means, or ask your Maestro about it.
Who's teaching you
Sergio Sánchez
I teach this because working with data has changed. Today's models are very good at basic analysis: one table and a clear question. And every version gets better. You no longer do every calculation yourself. You decide what to ask, which of several valid answers your report needs, and how you know it's right. That's where what you already know comes in.
I'm Sergio, chekos online. I'm a data architect, and I've spent more than ten years working with data in the Bay Area, mostly at mission-driven organizations: policy research first, then education, and now early childhood services. I grew up in Tijuana. In 2018 I started tacosdedatos to share what I was learning, and since then it's been a blog, a podcast, courses, guides, and a community. These days I spend a lot of my time building with AI agents and writing about how I organize my work around them. I share what I've learned; you take what's useful and make it yours.
See you in the course
Claude Code for Data: Your Judgment, at Scale