High Frequency Intelligence Shops: AI Arbitrage from the Future
What happens when a business is just an automated AI execution loop?
I’m going to let you in on a little secret.
And by little I don’t mean little.
I mean large.
Some of the most profitable businesses in the future won’t be high frequency trading shops.
No, no.
They will be high frequency intelligence (HFI) shops.
Whereas high frequency trading shops use algorithms to trade stocks based on information and latency advantages, high frequency intelligence shops will use intelligence to execute digital services 24/7 until they find arbitrage opportunities.
To understand this, you must first understand one very simple idea:
The history of technology is a history of disintermediation.
Before the internet, you had to go to a library to find a book to discover the specific knowledge you were looking for. The internet removed the need for this intermediary, in the process making information free and abundant.
Before the iPhone, you had two separate devices: an iPod to play music, and a mobile phone to text/call people. The iPhone removed the need for the iPod.
Before the dishwasher, you needed a human to wash dishes. The dishwasher removed the need for the human.
Before the lighter, you needed a matchbox to light a match. And before the matchbox you needed to rub sticks together for hours on end…
Get the picture?
Now, before artificial general intelligence (AGI), you needed humans to execute digital tasks requiring intelligence. AGI removes the need for humans.
Like all previous technological revolutions, there will be disintermediation.
When something gets disintermediated, what you tend to see is the resulting good/service becomes abundant, cheap, and more efficient.
The cost of the equivalent level of intelligence for the frontier AI models is decreasing at approximately 10x per year, whilst the models are simultaneously becoming faster and more intelligent.
This means that with just 1-2 more model iterations - perhaps GPT-6 or GPT-7 - we will approach the point where anything on the internet that can be arbitraged, will be arbitraged.
I am talking specifically about digital service arbitrage: digital marketing, outbound, design, software, copywriting, SEO, etc.
High frequency intelligence shops of the future will slurp in all public data on the internet - Twitter, Instagram, Google, Reddit, etc and in real-time do something like the following:
- Find a problem people are complaining about
- Propose a digital solution to the problem
- Automatically find and qualify leads for that solution
- Contact those leads with a sophisticated outbound funnel
- Close the deal without any human in the loop
- Funnel any capital from step 5 into a Treasury
- Optionally re-invest the Treasury capital into low-risk, liquid instruments on-chain.
- Analysis: If the strategy is profitable (based on the amount of compute spent to date), then continue steps 3-7 in a loop. If the strategy is not profitable, go back to either step 1 or step 2: either try another solution to the problem and continue the process, or find another problem and start again.
You now have a self-sustaining AI in a loop whose sole aim is to make money through digital service arbitrage.
I call this arbitrage because the public internet can be thought of as an incredibly inefficient marketplace of ideas that rarely get translated into outcomes.
For example, people post in forums all day long about problems they are having, but those problems don’t get solved without an enormous amount of friction, because solving the problem has traditionally required a human - which is extremely expensive and time intensive.
AI will make the market for solving digital problems extraordinarily efficient.
It will execute idea to outcome at the cost of compute, 24/7.
The moment someone posts something on the internet about a problem they are having, you can guarantee there will be AI Agents trying to pounce on the idea by transforming it into an outcome and automatically trying to sell it.
HFI Shops will arbitrage away the inefficiency of executing ideas to outcomes on the internet.
What I am describing is already possible today.
It’s just fairly cumbersome: the technical wiring to put this together is still reasonably difficult, the AI models are not quite there yet (but almost are), and the cost is still somewhat prohibitive.
However, this is the worst it will ever be.
If I had to guess, I’d say somewhere between 6-18 months from now the floodgates for this strategy will open.
As the supply-side of AI Agents trying to execute this arbitrage strategy increases, one factor will differentiate the best HFI Shops more than others:
Speed.
The quicker your AI can find, execute and sell digital solutions to people who need them, the more money you will make.
Which is why this market will eventually evolve into ‘high frequency’ intelligence arbitrage, not just normal frequency.
The HFI Shops who excel at this 5+ years from now will be the ones who can deploy and run intelligent AI arbitrage agents at scale, the most efficiently, to execute the loop in steps 1-8 above profitably and consistently.
When investors first asked Sam Altman how OpenAI was going to make them money, his response was ‘we will just ask the AGI’.
Well, this is essentially what the above is doing. If you turned steps 1-8 into a prompt, it would be less than 100 words.
If you’re not quite following, here’s a specific example.
Let’s say I run one of these HFI shops, and I’ve come back from the future to tell you how it works.
Here’s one strategy:
