Learn any field on demand
Next time a new field shows up at work, you already have a way in. AI coaches. You decide what counts. The method travels; the prompts don't.
A one-week retreat · United States · February 2027
AI amplifies experts.
Learn how to learn anything.
You pick a field you've never touched. AI is coach, not oracle. You leave with an artifact you can defend and a five-minute teach the room can't fake. That's how you stay the person models amplify, not the person they replace.
Your expertise feels copyable.
That knot in your stomach is real.
The decade you spent becoming good at something is now a prompt away for anyone with a chat window. You watch a model draft what used to take you a week, and you wonder what you're still for.
The fear isn't that AI is clever. It's that your edge suddenly looks optional, and your job description could change faster than you can retrain. Most people feel that and either ignore it or doomscroll.
So what do you do with that?
The people AI amplifies can learn a foreign field on demand.
They don't win by prompting harder. They win by driving the questions until the model is coach, not oracle. They know what to ask next, and they get to real understanding before the job description moves again.
You already tried better prompts, another tool, another course. That wasn't the miss. The miss was never owning a way to walk into a field you don't know. That's what this week trains.
Already an expert.
Ready to learn something completely foreign.
You arrive already expert in one domain. You apply with a new field you want to learn, chosen to be completely foreign to what you already know. Adjacent jumps are too easy, so we don't allow them.
A surgeon takes on game theory. A physicist takes on historiography. A literary scholar takes on information theory. The point of going fully foreign is confidence: if you can learn something totally outside your field, you can learn anything.
This is a retreat about knowledge work. Fields you learn by reading, thinking, writing, and arguing. That's where AI genuinely coaches you. Crafts, performance, and physical skills are out of scope.
Twenty participants means twenty foreign domains in the room. You leave with peers practicing the same jump, not a Slack full of prompt tips.
- You apply with
- A domain you don't yet know
- You bring
- Your own AI subscription & favorite chat harness
- You leave with
- An artifact you defend, a five-minute teach, and a method for the next field
Proof you can defend, not promises we make.
You prove it twice, in public. The meta-skill is what you take into every field after that.
An artifact you defend
Something working that you can present and defend. A journal club, a written explainer, a small model or analysis, a teaching deck. You pick the form; we push on whether it proves you learned.
A five-minute teach
On Day 5 you teach your new domain to the room and show your artifact as evidence. If you can teach it so they understand, you own it. Where you can't, that's your honest gap.
“Public teach is the filter. Private confidence is the prize.”
Monday: a field you can't touch. Friday: you teach it.
Here's the path that makes that possible.
- Day 1
Feel where your instinct breaks before we hand you any structure.
- Day 2
Stop at "I get it," then push past it with questions that prove what you actually own.
- Day 3
Write without AI, teach a human partner, and start the artifact that proves you learned.
- Day 4
Build all day and defend what you've made in a 1:1 with one of us.
- Day 5
Teach the room for five minutes and show your artifact. If they understand, you own it.
Each evening except build night, you write a reflection by hand, without AI. It's the one thing the model can't do for you: noticing your own thinking, and asking the question that turns exposure into ownership.
We've lived the domain jumps you're here to learn.
Eric Ma
Senior Principal Data Scientist, Moderna
Eric started in wet-lab biology and kept jumping: computing, data science, deep learning, network science. Right now he's working through diffusion models and starting on quantum computing. Each jump was its own foreign field. That's the meta-skill he teaches.
Read Eric's learning journey
Daniel Chen
Data Science Lecturer, University of British Columbia
Daniel studies how people actually learn to work with data. He teaches at UBC, wrote Pandas for Everyone, and has been a Carpentries instructor since 2014. He designs the week around what makes learning stick, not what looks good in a slide deck.
Read Daniel's learning journeyTwenty seats. One week. About $5,000.
In person, United States, February 2027. Lodging and transport are separate.
- Cohort
- 20 participants
- Dates
- February 2027
- Location
- United States (venue to be announced)
- Format
- One week, in person
- What you bring
- Your own AI subscription and favorite chat harness. The methods are tool-agnostic.
- Admission
- Curated. You apply with your resume, a headshot, and the new knowledge domain you want to learn. Something you learn by thinking, reading, and arguing, not by hand.
- Tuition
- ~$5,000 USDYou're not buying prompts. You're buying a foreign-domain stress test with coaches in the room, a 1:1 defense of what you built, and a public teach you can't hide behind. If a MOOC could do that, you'd already be done. Lodging and transport are separate.
Join the waitlist
Applications open to the waitlist first · February 2027
Twenty seats. Join the waitlist.
Applications open to the waitlist first. No spam: one email when applications open, one reminder before they close.
Admission is curated. When applications open, you'll send us your resume, a headshot, and the new domain you want to learn. We curate the room for foreign domains, so the application actually matters.
One week. One field you can't yet do. The proof at the end that you own it, and the meta-skill to pick up the next one.