CodeChella Madrid is Almost Here — Here’s What You Need to Know

A few weeks ago, I told you about the new material we’re adding to this year’s CodeChella — continuous DiD, synthetic DiD, triple differences, bounding exercises, the whole frontier. Today I want to make the ask more directly: if you’ve been on the fence, this is the post where I try to get you off it.

Come to Madrid.

CodeChella runs May 25–28 at CUNEF Universidad. Four days, 9am to 5pm, morning coffee and pastries included. Even if you barely know what a regression is, but you’re willing to learn and get your hands dirty with code, then you’re ready for this workshop. We build from the ground up.

Tickets are on Eventbrite here. Pricing:

∙ Students: $220

∙ Post-docs: $300

∙ Faculty: $500

If cost is the obstacle, email me at causalinf@mixtape.consulting and we’ll work something out. I mean it. I don’t want that to be the reason you don’t come.

---

The Claude Code Thread

I’ve been writing on this Substack for months about how Claude Code has changed the way I do empirical research. CodeChella is where you get to see it in action.

Throughout the workshop, I’ll be running my replications and demonstrations inside Claude Code environments. That means every time we work through a new estimator — event studies, Callaway-Sant’Anna, Arkhangelsky’s synthetic DiD, Rambachan-Roth bounds — you’ll also be watching me work with Claude Code in real time to build it. The diff-in-diff content and the AI-assisted workflow are woven together, not siloed.

My theory here is pretty simple: the best way to learn Claude Code is to use it for something you were already planning to do anyway. Making event study graphs. Running pre-trend tests. Building clean tables and publication-quality figures. If those are things you care about — and if you’re coming to CodeChella they probably are — then you’ll leave with both the econometrics and a working sense of how to use an AI coding agent to do applied quantitative research.

But I also want to be honest about something. Speed is not the point. The thing I want to teach — the thing I think matters most right now — is verification. How do you know what Claude Code produced is right? How do you build habits that catch mistakes before they end up in a paper? How do you structure a workflow so that the gains in speed don’t come at the cost of credibility?

That’s part of what this workshop is now. Not a demo of how fast I can run things. A serious attempt to show you how to use these tools well.

---

Madrid in Late May

The weather is perfect. The food is extraordinary. CUNEF is a great venue. And honestly, four days in Madrid with a room full of people who care about causal inference is one of my favorite things I get to do.

I’ll be back next Monday with more. But if you already know you want to come — grab your ticket here.