③ A semester worth multiple years
I remember last semester when the 4th cohort students walked in. Not one of them could write a line of code. For business school undergraduates, this shouldn't be surprising. But the project they were about to take on required building a system where AI generates market forecasts, integrates them into a Black-Litterman model, and optimizes a portfolio.
None of them believed that they could finish what they have actually accomplished by the end of the semester.
To explain how that happens, I need to walk through what a semester in LABA actually looks like from the inside. And in following that process, I think the reason this structure produces such fundamentally different outcomes becomes clear.
The first few weeks began with learning how to use Claude. This sounds mundane (e.g. what exactly is there to learn when you just type in stuffs to LLM?), but it is quite important. Most students don't really know how to work with AI; not because the tools are complicated, but because they don't know what to ask. Formulating a good question requires knowing something about what you don't know, and that's a skill almost no one teaches. So it starts with a demonstration. I show them how to talk to AI: how to frame questions, how to build on the answers, how to use the conversation to construct understanding layer by layer. The same process I described in the first essay, when I taught myself robotics. You give the AI a destination, let it draw a map, and learn to walk it.
Once this clicks, the pace of learning changes visibly. Need to learn Python? Ask AI. Unfamiliar concept? Ask AI. Code won't run? Have AI debug it. Not long ago, I would have had to teach all of this personally: syntax, libraries, data processing, model implementation. A single semester would have barely covered the fundamentals. Yet, AI is compressing this entire journey.
But this is only half of the equation. No matter how efficiently AI delivers knowledge, it cannot tell a student that they're heading in the wrong direction. AI answers questions well, but it does not tell you when you're asking the wrong question. A student building confidently on a flawed assumption needs a human being to stop them and say, "Hold on. Is that assumption actually valid?" Or: "If you keep going this way, you'll have nothing to show at the end of the semester. Step back." Or simply: "Why did you choose this approach? Walk me through the reasoning." I have these conversations with my students every week. Often every day. To be honest, this is all that I do at LABA.
The crucial difference this creates comes down to one thing: the speed of feedback. In a seventy-person lecture, a student discovers what they don't understand exactly twice per semester: at the midterm and the final. By the time they get the exam back and realize they had a misconception, the course has moved on by several weeks (assuming that the professor finished grading on time, which often does not happen for myself).
Learning, at its core, is a cycle: try something, find out whether it's right, fix what isn't, try again. In a large lecture hall, this cycle turns two or three times in an entire semester. In LABA, the cycle turns several times a day. A student attempts something, AI shows the result immediately, I correct the direction, the student adjusts and tries again. This happens daily. The feedback loop that takes a semester in a lecture hall completes in hours in my lab. The difference in growth is exponential.
Through this process, the fourth cohort completed a working Black-Litterman portfolio optimization system by the end of the semester. Students who couldn't code four months earlier. And this wasn't unique to them. The first cohort built an ML model predicting IPO overpricing. The second automated relative valuation with ML and large language models. The third simulated FOMC committee members' reasoning using LLM agents to forecast interest rates decisions.
Currently, the fifth cohort, undergraduate business students, is building a bipedal robot that walks, grasps objects, and converses in natural language! (can you believe this?!) Every semester begins with "Can we really do this?" and ends with "We did."
When I tell people this, the first response is usually, "Those students must have been exceptional." Well, yes, they were strong. But all my students sit in every seventy-person lecture hall are strong. What creates the difference is not the talent in the room. It's the structure of the room. AI transforming the speed of knowledge access, a professor providing directional correction, and a feedback loop that runs daily instead of biannually; that combination is what's different.
None of this is easy. LABA students pull all-nighters. They hit walls and struggle. It's a completely different species of difficulty from sitting in a lecture hall copying slides. But here is the part I find most striking: they don't resent it. They seem, against all logic, to enjoy it.
One student told me that "This is the life I've always wanted to live." He said this during a week when she'd been up past midnight almost every night working on the project. He wasn't performing gratitude. He meant it. What he was describing was the joy of wrestling with something real: of failing, adjusting, trying again, and eventually building something that works. He told me he had never felt anything like it in years of sitting in classrooms. And I think this connects directly to what I observed in the second essay: the reason students throw themselves into extracurriculars but not lectures isn't about the subject matter or the difficulty. It's about the difference between doing something and having something done to you.
That feeling, the joy of active, self-directed struggle, does not emerge in a seventy-person lecture hall where the primary activity is copying slides to fulfill credit requirements.
This leads to the obvious question. If this works so dramatically well, why doesn't every classroom operate this way? Well, it is not because universities don't want to, but rather because they can't. The next essay is about why.