AI-Proof Businesses: Community Moats Explained

Picture a Midwestern auto dealer in loafers and baggy jeans, the kind of guy who clears five million dollars a year and still has, as Sam Parr puts it, "hillbilly energy" (00:42). Once a year, three times a year actually, he flies to a hotel conference room with twenty-three other dealers he will never compete against. He opens his laptop and reads his company's full profit and loss statement out loud, line by line. The others do the same. Someone tells him he is leaving money on the table. Someone else copies his idea.

This is NADA 20 Group, run by the National Automotive Dealers Association, and it has apparently been doing this since the 1970s (01:40). Membership runs six to eight thousand dollars a year, groups are capped at twenty four non-competing dealers, and Parr guesses the whole operation pulls in thirteen to fifteen million dollars annually just from dues (02:20). The website, everyone on the podcast agrees, looks terrible. It does not matter. The business has survived a century of marketing fashions by selling something that cannot be faked: other dealers' real numbers.

That contrast, a plain website sitting on top of an unbreakable product, is the quiet engine of this episode of My First Million, hosted by Sam Parr and Shaan Puri. Their argument is that artificial intelligence is busy eating the moats that used to protect software companies, the ones built on clever code or a head start. What AI cannot easily eat is a room full of people who trust each other enough to show their books.

"Previously technology was a moat. That's not a moat anymore. And I think communities, networks, and marketplaces are still a really good moat." — Sam Parr [10:47]

The second example is BNI, a referral network that Parr says was just acquired for a figure rumored to top one billion dollars (05:11). It has roughly 800,000 members worldwide (06:18), organized into local chapters where a realtor, a garage door installer, and a lawyer meet weekly and trade leads instead of competing. The dues run into the low thousands per year. Nothing about the format is exotic. It is a room, a list of names, and a shared understanding that everyone benefits from everyone else's business staying open.

A newer version of the same idea is Moda Communities, a set of roughly twenty to twenty five Facebook groups built around buying and selling watches, cars, gold, and jewelry (08:44). A founder in Sioux Falls, South Dakota started one group in 2018 after struggling to find a trustworthy place to sell a watch. It has grown to about 600,000 members, with tiers for watches under ten thousand dollars, watches above that, and a group reserved for collectors spending six figures (09:10). Dealers pay between one thousand and ten thousand dollars a year to list, after a vetting process that includes a human interview and a second member vouching for them. Parr estimates the business does five to ten million dollars a year (10:03). Technology did not build this. A Facebook group, trust, and a fee did.

The Whale Economy

Underneath every one of these communities sits a less polite truth: most of the money comes from a small number of people who spend enormous amounts. Puri, who previously worked at the livestreaming platform Twitch, says this concentration shows up everywhere he has looked, from collector marketplaces to mobile games (11:22).

"Whales drive just a shocking amount of the economy. 80, 90% of the revenue will come from a just a tiny sliver of those players." — Shaan Puri [11:22]

The puzzle game Candy Crush has 82 million monthly active players, and all of its revenue comes from just 4% of them (13:23). The other 96%, Puri says, are essentially the cost of finding the paying 4%. Twitch learned to serve this group directly when it launched a feature called Gifting Subs, which let a viewer buy up to two hundred subscriptions at once and hand them out in a chatroom, a move that functions like buying bottle service at a club and that noticeably increased revenue per channel once whales had a flashier way to spend (12:16).

Puri's description of the engineers who design these systems is blunt.

"It's literally like talking to a slot machine operator when you talk to them." — Shaan Puri [14:17]

He explains that game studios track a player who has paid before but gone quiet for about thirty days, then engineer a run of easy wins followed by a frustrating wall timed to push that player back toward the payment button (15:09). Unlike a Las Vegas slot machine, which regulators require to pay out at a set rate, a mobile game can set its odds to anything it wants.

The same mechanic, stripped of the screen, shows up in physical collectibles. A new wave of apps lets users buy a virtual pack of trading cards, open it, and if they are unhappy, sell it straight back for slightly less than they paid, say forty three dollars on a fifty dollar pack. That near-refund feels like free money, so the player buys again. One such app has done a billion dollars in revenue in a year (17:17), and a related one, Collector Crypt, grew from fifty thousand dollars a month to twenty five million dollars a month using the same buyback trick (18:30). Its founder previously worked as a quantitative trader at the investment firms Citadel and Bridgewater. The underlying format, called gacha in Japan, became so addictive that the country eventually restricted it by law (19:34). A toy industry contact told Puri that switching identical products from an open shelf to a mystery box increased revenue by 7x, even though buyers technically got less choice (21:27).

A Postscript on Trust

Not every AI-resistant business is glamorous. Private Photo Vault, an app for hiding photos behind a calculator icon, has over a million reviews and has quietly generated significant revenue for years, run by one American founder with a small team (23:16). It survives for the same reason the communities do: people need somewhere private to put something they do not want floating around, and no algorithm negotiates that trust away.

The episode closes on a story that inverts the whole theme. Binance, a cryptocurrency exchange, was allegedly hacked for five billion dollars after an operative linked to North Korea reportedly used a romantic relationship to gain physical access to an engineer's laptop (27:57). Binance had built a two billion dollar reserve, called the SAFU fund, from 10% of trading fees collected since 2018 (27:32), and said it would cover the rest itself. The attackers immediately moved the stolen funds into a decentralized finance (DeFi) protocol, a kind of permissionless bank that let them borrow stablecoins against the stolen crypto as collateral, locking in cash even if the original funds were later frozen (29:54).

Put side by side, the two halves of the episode make an odd kind of sense. A room of dealers trusts each other with a profit and loss statement. A chatroom of collectors trusts a vetting process. A blockchain trusts nothing and nobody, which is exactly why it needed a five billion dollar insurance fund in the first place. AI can write the code for all three. It cannot manufacture the twenty four strangers willing to open their books to each other, and that, this episode argues, is where the money will keep hiding.