COHORT 1 — KUALA LUMPUR, 2026 · SOUTHEAST ASIA
Stop building things that don't matter.
Y0R is three months in Kuala Lumpur to find out — working on hard problems at the frontier of AI, surrounded by people who are asking the same question and won't let you get away with a vague answer.
The most important technical problems of the next decade are in understanding AI. How do these systems actually work? What can they do that we don't understand yet? How do you build on top of them reliably? How do you evaluate whether they're doing what you think?
These questions are being worked on seriously in a handful of places. Almost none of them are in Southeast Asia.
It's not a people problem. SEA produces technically rigorous engineers. People who can build, who can think from first principles, who are capable of working on hard problems.
There just isn't a space where that capability gets pointed at the frontier.Y0R is that.
Not a course. Not an accelerator. Not a research lab. A house for 10 - 15 people in KL for three months, working on the hardest problems they can find, with practitioners who will not let them get away with shallow thinking, and peers who will push them further than they'd go alone.
The problems we care about are in AI systems, security, infrastructure, and the institutional questions underneath all of it. The work is technical. The outputs are real. The standard is high.
If you already know you want to work on something that matters and you're done waiting for the right environment — this is for you.
The program
The model
Hard problems first. No brief. No curriculum.
Residents arrive with something they want to work on — a project, a question, a problem they've been circling. The house provides the environment, the pressure, and the people. The rest is on them.
Build → Measure → Learn
Build something real. Show it Thursday. Take the hard questions. Figure out what you got wrong. Repeat.
What residents work on
Problems at the frontier of AI systems and what sits underneath them:
- Mechanistic interpretabilityHow do these models actually work — internal representations, emergent behaviour
- Evaluations & red-teamingWhether they're doing what we think — benchmarking, adversarial testing
- Adversarial conditionsWhat happens when they meet attack and defence — robustness, security
- Constrained deploymentHow to deploy them where designers never imagined — edge hardware, low-resource environments, critical infrastructure
- Governance infrastructureWhat audit and policy tooling the world needs to actually manage these systems
What residents get
- A house with eleven other people as serious as they are
- Supervision from practitioners working at the frontier — available when you're stuck, not scheduled when you're not
- Guest lectures on Monday evenings — provocations, not presentations
- Three months of unbroken time
What residents don't get
- A certificate
- A structured curriculum
- A job guarantee
- Hand-holding
A week in residence
The structure exists to protect deep work and create accountability. Nothing else.
Monday evenings are provocations — a practitioner who will say something that changes how you think about your problem. Not a lecture. Not a panel. One person, one idea, hard questions from the floor.
Tuesday through Thursday mornings are protected. No meetings, no interruptions, 9am to 1pm. The only obligation is the work.
Wednesday afternoon you write — what you're working on, where you're stuck, what you need. Your supervisor reads it. Your peers read it. You can't hide in vagueness for long.
Thursday evening is show and tell. You show what you built, not what you planned. The questions are hard.
There are no applause sessions.
Friday morning is sprint planning. Friday afternoon is yours.
Monday
Tuesday
Wednesday
Thursday
Friday
The filter
Three stages. No CV. No grades. No institution names. We select for how you think under uncertainty—not for legibility on a funding narrative or where you studied.
01
Application
Four questions. What you've built that nobody asked for. What you're obsessed with. What most people get wrong about your field — and why. We're reading for how you think, not what you've done.
- What have you built that nobody asked you to?
- What problem are you obsessed with?
- Make a case for something most people get wrong
- Why Y0R, and why now?
02
Problem set
3–4 problems. 72 hours. No domain knowledge required. First principles only. The instruction is: show your thinking. Getting it wrong the right way beats getting it right for the wrong reasons.
- One quantitative problem
- One strategic problem
- One open-ended problem
- No clean answers that can be looked up
03
Interview
30–45 minutes. We take something from your application and push on it. Can you defend a position? Can you update when you're wrong — and tell the difference between a good argument and social pressure?
- No competency questions
- No prepared answers
- Decision within 5 days
Who this is for
You're an engineer, or something close to it. You can build things. You've probably been building things with AI, on top of AI, or adjacent to AI for long enough to feel the gap between what you can do and what you actually understand.
You know you want to work on something that matters. You don't know yet exactly what that is, or you know what it is but haven't had the environment to commit to it properly.
You're 21 to 27. You're in Southeast Asia, or willing to be in Kuala Lumpur for three months. You're not looking for a credential or a job placement. You're looking for the right room.
You might be right for Y0R if:
- →You have a project or a question you'd work on even if nobody paid you
- →You can think from first principles, not just from frameworks someone taught you
- →You're capable of being told you're wrong and using it
- →You're done with work that doesn't move anything
Y0R is not for you if:
- ✕You want to be taught what to work on
- ✕You need external structure to be productive
- ✕You're looking for a network, a credential, or a launchpad
- ✕You're interested in AI as a tool rather than as a subject
The selection process will find the difference.
Frontier problems
Questions
The things people ask before they apply.
About the founder
Brendan Beh
Brendan read Physics at Cambridge. He spent his early career in deep-tech analysis before moving into venture and ecosystem development at Sunway iLabs, where he worked across investment, founder engagement, and strategic partnerships inside one of Malaysia's largest conglomerates.
He left to build AI.SEA — a builder coalition now spanning 8 countries and tens of thousands of people across Southeast Asia.
The thing he kept hearing, from serious people, was this: they wanted to come back. To Malaysia, to the region. But they couldn't point to anywhere that matched their ambition. The talent was never the problem. The infrastructure for it was.
Y0R is his answer to that. Not a startup school. Not a network. A place where the most serious people in SEA can finally operate at the level they're actually capable of.
Cohort 1.
July 2026.
Rolling review. Apply early — we read everything and respond to everyone.
Opens our secure Airtable form in a new tab. This begins the application — not a waitlist. We read everything and respond to everyone.