The Best Workplaces list came out this week and we’re on it again, second year running. I know the version of a company that reads that sentence and high-fives. I’m apparently not running that company, because what I actually did when the news came in was sit there and ask myself how the hell we pulled it off twice. Which is a less satisfying reaction and a more useful one.

So here’s the most honest answer I have. It starts with a new hire who took a support call on her third day and sounded like she’d been here a year.

I want to be careful about how much weight that sentence can carry, because if you’ve never run a service desk it might slide right past you. So here’s the comparison. The old way — getting a new technician to where they could handle a live call on their own — took us a couple of months. Onboarding videos. Reading the documentation. Shadowing someone senior until they’d seen enough to fly. Months before solo. That wasn’t a failure. That was just how it went.

Day three is not how it went.

Day three is possible because of a thing my COO, Kemis Hancock, built. She earned the first black belt in our Claude Belt program about a month after we launched it, and the project she built to earn it was the onboarding itself. A new person now sits down with a Claude project that walks them through the actual job, carries them through the first hard days, prompts a short daily reflection they talk through with their manager, and sends the team a note at the end of each day so everyone can see the new person’s wins. The first thing the program ever produced at its highest level was the machine that makes the next person fast.

The Belt program isn’t a course. We never sat anyone down to teach them how to use Claude. People learn Claude by using Claude on real work, and they earn belts — actual karate belts, in colors — as they go. Learn by doing, do by learning.

When people ask what “going all in on AI” means at a company like ours, they expect an answer about efficiency. It isn’t that. The thing that quietly slows a company down isn’t the work — it’s the gap between the person who knows and the person who doesn’t. I don’t know. Ask Eric. Ask Cooley. Every one of those sentences is a small wait, and the waits pile up into a company where the knowledge lives in a few heads and everyone else lines up outside them. What we’re actually building is a place where a new person can reach the whole company’s knowledge on day three instead of year three — and where the work they do then goes back in, so the next person starts a little smarter than they did. The company gets smarter as a property of how it’s built, not as a function of who we happened to hire.

That last part is the answer to the question I started with.

Because Best Workplaces is a staff survey. The people at the top can’t fill it out. It measures what the team says about the place when nobody from the C-suite is in the room. One year, you can tell yourself you caught a good week. Last year we were the only Hawaii company on the list — flattering, easy to wave off as a fluke. So were we this year. Two years is harder to wave off.

And here’s the only nod I’ll make to something the people who know us already know: the team that earned this the first time and the team that earned it this year are not entirely the same team. The list doesn’t measure individuals. It measures whether the thing those individuals walk into is built to hold. Twice now, it has.

That’s what day three actually is. Not a productivity stat. It’s the proof the thing holds — that a person can walk in cold and the company can make them whole, fast, because we finally built it to. I didn’t know that for certain until the second time. Now I do.