Why AI Makes "Fast Eaters" the New 10x

I once read about a Japanese company that started in a three-tatami warehouse and grew into a global monster. Their hiring test? Eat lunch as fast as you can, speak loudly, and clean the toilet properly. The founder’s logic was simple: people who eat fast work fast, people who speak loudly carry energy, and people who clean toilets without being told actually give a damn. It sounded insane, almost cartoonish.

Until I started hiring in the AI era and realized we need exactly those people again.

AI didn’t just speed things up. It quietly rewired the division of labor.

What used to feel like 70% planning and 30% execution has flipped. Today, the first 10% of “thinking work” gets drafted by Claude in 8 seconds. The remaining 90% is execution, verification, iteration, and aggressive ownership. The old default — “hire the meticulous analyst who thinks deeply before moving” — is now partially commoditized. The new scarce resource is the person who takes the AI output and smashes it into reality at full speed.

I’ve been running exactly this experiment.

I started selecting candidates who had ridiculous GitHub commit velocity, who could understand a sentence once and react in 30 seconds, who would open the editor on the spot when I said “code this quick,” and who radiated proactivity and hunger. I also looked for something more subtle: the aggressive team player — the person who moves fast, but pulls others in instead of going rogue.

These are the ones who, the moment something is unclear, immediately walk over or DM you: “Hey, 5 minutes? I don’t fully get this part.”

They treat blocking issues like personal enemies.

And crucially — they know that if they blindly ship AI-generated work without truly understanding it or verifying it properly, it’s on them. They will get called out, and they accept that as part of the job. They internalize that accountability before anyone says it.

The result? Insane momentum.

A seasoned manager throws an AI-generated goal and rough outline. These people don’t look for reasons it won’t work. They look for ways to make it work — usually by feeding more context back into the model, running quick tests themselves, and coming back with “I tried X, it broke here, here’s the fix I’m shipping.” When the manager only needs to set the guardrails and key invariants upfront, the team moves at rocket speed.

Of course, they are not perfect. They sometimes gloss over documentation or long-term elegance. That’s why you still need the senior who sets the frame and the “verify, don’t trust” culture. In my experience, though, 75% good at 8× speed beats 98% perfect at 0.2× speed in most markets that actually matter — especially when you can iterate daily and correct course long before the perfectionist would have shipped.

The system has changed.

AI commoditized a big chunk of the thinking and drafting. What it rewards now is execution athletes with radical ownership: high energy, bias-to-action, aggressive communication, and the self-awareness that “if I don’t understand this output, I will be the one who looks stupid.”

We spent decades optimizing for the careful thinker.

Maybe it’s time to optimize again — for the fast eater who asks loud questions, ships messy but ships today, and knows the difference between “AI wrote it” and “I made it mine.”

Strong opinion, held weakly.

If your context is different — regulated industry, safety-critical systems, places where failure kills people — the meticulous default still wins.

But for most of us building software and products in the real world in 2026?

The fast eaters are eating everyone’s lunch.