Where I live, the national loto company, Loto-Québec, regularly run campaigns reminding players to sign their tickets or scratch cards when they play. Ostensibly, this protects the winner – in some minimal way – but the back of every ticket is also a contract with Loto Québec. That contract states that, essentially, a winner is not entitled to receive their prize unless they allow Loto Québec to run news and ads featuring them.

Please everyone meet the real person who is the winner of the grand prize in this week’s Loto 6/49, for which you could buy a ticket, wink wink.

The awesome value of lotteries and luck games in a money-first society is that they convince people to buy a participation, based on results they’re almost guaranteed not to receive. To one person, paying ten dollars to receive ten million is an absolute outsized advantage. A few hundred people will receive sub-prizes that are better than break-even. And millions of people pay the same ten dollars to get nothing.

A healthy lotto product still has to work out, financially. Its expenses and payouts have to be covered by the ticket sales. The best way to achieve this is to have very few winners, maybe one winner singular, with a ridiculous grand prize. With an astronomical potential gain, you can expect buyers to completely bypass rational analysis of the cost and benefit, which generally means lottery buyers focus on the largest possible prize. Of course, then you pay out to a winner you put in the news and in ads, fuelling a sort of Skinner Box by proxy for everyone else¹.

Now – allow me to do what will come to be known as the classic blogpost sidestep of 2025, and make this about generative LLMs.

The new, new wave of private-funded market grabbers

Let’s split these lottery characteristics I mentioned into two groups:

• Outsize gains are given to very few winners, who then act as marketing to sell further participations.
• The balance of expenses and gains must be financially healthy.

What I’d like to suggest here is that the innovation of LLM infrastructure providers as we know them today is to adopt the advantages of group one, while being able to ignore the obligations of group two.

A lot of modern VC unicorns have done some form of this scheme. The current ideal for a startup investment seems to be an incumbent in a market successfully avoiding a key responsibility that should be part of their presence in said market. They are structured to burn money in exchange for subsidized market dominance, letting them grow fast enough to stay ahead of legislation and judgements.²

But under this model, they still have to show financial health, eventually. That’s when Uber goes from cheap and luxurious to pricey and terrible, for example.

With LLM builders and infrastructure providers, they've managed to do away with IPO pressure, and go straight for deals that funnel money out of nation-level budgets into private pockets through third-party provider arrangements. These deals, of course, inflate valuations without requiring the financial transparency an IPO would demand. It's less an Instagram unicorn, and more of a Palantir: old money clumped into a papier mâché approximation of a startup.

The innovation is avoidance-stacking: first social responsibility, now financial responsibility on top of that. Legal arbitrage, sure, but now we have companies running emotional arbitrage. They gain ground so long as users and investors maintain an irrational relationship with the product and the company.

I’m of the opinion that while there is great utility to be found for LLMs, their measurable value is being obfuscated by the egregious fantasy the big players have to tell about their relevance and growth prospects, which itself sustains the egregious financial fantasy of their funding and deals, that at this point they can’t question without everything else breaking down.

Of course, the fantasies will eventually stop working, but so long as they have lottery winners to show from time to time, they should continue their fragile climb, and funnel substantial chunks of the working economy into a few private wallets.