Scott Wu on first principles, what an engineer really is, and training on the loss

Welcome back to the Z Fellows newsletter! Every Monday we share 3 ideas - to help you build companies, ship products, and create your life's work.

1: Scott Wu on first principles

"The reason that young founders who are just really sharp and really determined did very well is because at the end of the day, being a good first principles thinker does beat experience." — Scott Wu, co-founder & CEO of Cognition

When Wu started Cognition, almost nobody in the industry believed an AI software engineer was buildable yet. The consensus said the models were years away. Wu looked at the same GPT-4-era models everyone else had and concluded they were already good enough to start.

He wasn't relying on a longer resume than his skeptics. He had less experience than most of them. What he had was the willingness to reason from the ground up: the world already spends trillions a year on human software engineers, the models could already write code, so product-market fit wasn't an open question — it was a foregone conclusion.

Most founders inherit the assumptions of their industry. They accept that a hard problem is "a few years out" because that's what everyone says. But "a lot of being a founder is doing something that has never existed before and coming to your own conclusions."

The takeaway: When the whole field agrees something is years away, check the assumption yourself. The gap between consensus and reality is where companies get built.

2: What an engineer really is?

"What it means to be a software engineer in my mind is basically just somebody who solves problems in the context of code. It is somebody who tells the computer what to do and makes all these decisions." — Scott Wu

"Build an AI that can do software engineering" sounds impossibly broad — until Wu defines the job down to its core. Strip away the frameworks and the job titles and you're left with one thing: a person who makes decisions and tells a computer what to do.

Once the problem is defined that precisely, it becomes tractable. Wu called building it "deeply algorithmic" — exactly the terrain his team, holding ten IOI gold medals between them, knew better than anyone. The impossible-sounding mission was really a well-shaped technical problem hiding behind a vague description.

Great founders do this reduction constantly. They take a fuzzy, intimidating goal and compress it down to the irreducible core, because that's the version of the problem you can actually attack.

The takeaway: If your mission sounds too big to start, you haven't defined it precisely enough yet. Compress it until it's a problem you can build.

3: Training on the loss

"I had an older brother named Neal... Neal started doing math competitions when he was in middle school. And so he would've been in sixth grade and I was in first grade at the time. And naturally I, as a little brother, would go and just watch what he was doing and try to learn some of the same math too." — Scott Wu

Wu's origin story isn't the perfect IOI score or the national championship. It's an eight-year-old who entered a competition two divisions above his grade, was certain his name would be called at the awards ceremony, and heard it never come. He went home furious — and trained. The next year he entered a harder division and won.

He still tells that loss two decades later. The compounding didn't start when he began winning. It started when he lost, took it personally, and went back to work. By eighth grade he was national champion; by 2014 he had a perfect score no American had ever matched.

The pattern repeats through his whole career — high school dropout, two-time college dropout, a first company that "went okay" but wasn't a smash. Each setback was raw material, not a verdict.

The takeaway:Your origin story isn't your first win. It's what you did the first time you lost. Train on the loss.

See you next Monday,

- The Z Fellows Team