How AI brought me closer to my kid

The title is tongue and cheek but here's what happened.

My 15 yo came up to me yesterday and asked me to explain the project I just finished with a client. [Adult post on that here: https://www.linkedin.com/posts/harrisonm_six-months-ago-i-parachuted-into-arrivalist-share-7443400478413139968-zL7V]

I was tired. Claude was open. I asked it to explain my work, for which it already had context, to a “very smart 15 year old who loves Minecraft and freshwater fishkeeping.”

Verdict: he know understands what I did, although "AI doesn't really understand Minecraft."

→ Claude's explanation:

What Arrivalist Travel Audiences Is

You know how in Minecraft, you can look at a map and see exactly where players have been — their path, their base, where they mine, where they fight? Now imagine you could use that map to find other players who behave exactly like the best players in the game. Not players who said they like mining, or who watched a YouTube video about diamonds — but players who actually went to the diamond layer, repeatedly, at the right time of year.

That's basically what Arrivalist built for travel advertising.

The problem it solves

Companies that sell travel stuff — airlines, hotels, ski resorts, beach towns — want to show their ads to people who are actually likely to travel. Makes sense. But most of the time, ad companies are guessing. They might say "this person googled 'ski resorts' so they're probably a skier." That's like saying someone who watched an aquarium YouTube video is definitely going to buy a saltwater tank and a clownfish. Maybe. Probably not.

Arrivalist doesn't guess. They know where 60 million people's phones actually went. If your phone was at a ski resort in Colorado last February, Arrivalist has that data. You're not a "probable skier." You're a verified skier.

How the data works

Think about your aquarium. You know the nitrogen cycle — how ammonia breaks down into nitrite, then nitrate. Each step is real chemistry, not a guess. Arrivalist's data works like that. Every data point is a real phone, in a real place, at a real time. They call it a MAID — basically a unique ID for a phone, like a username — and they know exactly which polygon (think: a drawn boundary around a ski resort, a beach, a golf course) that phone was inside.

So instead of guessing, they can say: "Here are 500,000 real people whose phones were at ski resorts last winter." That's the segment. Clean, factual, no modeling.

Standard vs. Custom segments

Arrivalist has two products. The standard taxonomy is like a pre-built Minecraft biome pack — 20+ ready-made audiences you can just grab. Beach travelers. Ski travelers. Casino visitors. Golf players. You pick one, activate it, done.

Custom segments are like building your own biome from scratch. A hotel in Colorado might say: "I want people who live in Texas or California, who skied in Colorado or Utah in the last two winters, and who stayed in hotels rather than Airbnbs." Arrivalist can build exactly that — combining origin (where they live), destination (where they went), activity (what they did), and time (when they did it).

How it gets to the actual ad

Once a segment is built, it lives on something called LiveRamp — think of it like the Minecraft Marketplace, except instead of selling skins and mods, it sells audiences. Advertisers log into their ad-buying platform (called a DSP — basically a dashboard where you buy digital ads), find the Arrivalist segment, and attach it to their campaign. Now their ski resort ad only shows to verified skiers. Not people who bought ski boots as a gift. Actual skiers.

Why it matters

Imagine you're selling rare aquarium fish — like a black ghost knifefish. You could post your ad everywhere and hope a fish person sees it. Or you could show it only to people who have actually been to aquarium stores, fish expos, or reef keeping clubs. Same budget. Way better results. That's what place-verified targeting does — it stops wasting money on people who don't care, and finds the people who already proved they do.

That's what got built. From nothing. In six months.