The gap is closing.

There used to be an enormous gap between having an idea and making it real. You'd need to hire a team, raise money, wait months. That gap is collapsing. Rapidly.

Colton Hess — Haas Tech Club Show & Tell — March 12, 2026

The bottleneck moved.

There's this concept from operations: in any system, throughput is determined by the bottleneck. When you relieve one bottleneck, the constraint shifts somewhere else.

For most of history, the bottleneck was execution. The cost of building. The team you needed. And we calibrated our ambitions accordingly. We scoped down. We were "realistic."

AI is relieving those execution bottlenecks faster than anyone expected. Which means the constraint is shifting. And it's shifting to imagination. To knowing what's worth building. To taste. To understanding problems deeply.

Our ambitions were calibrated to a world that no longer exists. The dreams haven't caught up to the tools.

This is happening. Right now.

Not in ten years. Not when we're "ready." Now.

METR, an independent AI evaluation org, tracks how complex a software task AI can complete autonomously, measured by how long it would take a human engineer.

AI Task Complexity Over Time
How long the equivalent task would take a human engineer (METR data)
~30 min
~2 hrs
~6 hrs
14 hrs
2023
Early 2024
Late 2024
Today
Don't look at the point. Look at the slope.

That's not a linear improvement. The capabilities are doubling every few months. Not years. Months.

What this looks like in practice.

I'm not a developer. Zero CS training. I'm an MBA/MPH student and a former Amazon PM. But I had problems I wanted to solve. So I built the solutions.

🌙

Remi (Rhythmicly)

The sleep industry shows you data and hopes you figure it out. Nobody actually changes your environment. So I built an AI sleep assistant that talks to your wearables, dims your lights, and adjusts your thermostat. It's in the App Store right now.

🌐

Orbit

I was drowning in networking. Couldn't keep track of who I'd met, who I needed to follow up with. So I built a personal CRM that syncs my meetings automatically and sends me a briefing every morning. It nudges me when someone's about to go cold.

🧠

Command Center

Running a startup solo means drowning in ops. Content, outreach, research. So I built an AI Chief of Staff that handles most of it. I review, it executes. It runs 14 automated workflows while I sleep.

🚀

Activation Playbook

Classmates kept saying "I want to build but I don't know where to start." So I built personalized pages that give each person a starting point based on their background. You're on the site right now.

The hard part was never the code. It was knowing what to build. That's what we've been training for.

What if all 300 of us built?

I'm not up here to talk about what I built. I'm up here because of what I think we could build.

I want our class to be the first MBA cohort fully trained in agentic AI. Not because it's a resume line, but because of what it does to you internally. There's this sense of agency that hits when you realize you can just build the thing you've been thinking about. You stop waiting for permission. You stop scoping down.

We came to the school that says "question the status quo." The status quo right now is that MBAs manage and engineers build. That line is dissolving.

Think about how much better the world could be if 300 people with this training and this network just went out and built the things they can see need to exist.

Ready to build?

If you've had an idea sitting in the back of your head, a product, a tool, something you've been meaning to build, you can start right now. No coding experience required.

And if it clicks for you, help the next person get started. Let's make this contagious.

Get started with Claude Code

Already building? Check out the context engineering guide to go deeper.