Two summers ago, three people I cared about lost a meaningful amount of money trading crypto in roughly the same way, in the same six-week window. None of them were stupid. One ran a real engineering team. One had been a finance major. One is one of the calmest people I know in any other context.
They didn’t lose because they bought the wrong tokens. As it turned out, they were mostly right about which tokens to buy.
They lost because they were asleep when their stops should have triggered. Because they moved their stops at 2am after a glass of wine. Because they revenge-traded a losing position into a worse one at 7am before they were really awake. Because, like every retail trader who has ever existed, they couldn’t be in the seat the whole time the market was open — and the market was always open.
Meanwhile, the people on the other side of their trades had algorithms. The same kind of execution discipline a hedge fund’s junior PM gets from infrastructure, the retail trader is expected to manufacture out of caffeine and willpower.
That’s the gap I started AutoCoin to close.
The asymmetric system
Here is what I think the consumer crypto market actually looks like, with the marketing stripped out:
- 93% of retail traders lose money over any meaningful timeframe.
- They do not lose because they’re irrational. They lose because they are biological humans in a market that is open 168 hours a week, and the people on the other side of their trades are not.
- The trading-bot category that supposedly fixes this problem has been dominated, for years, by platforms that ask the user to deposit funds on a centralized service in a jurisdiction the user has never heard of.
- Several of those platforms have already collapsed in spectacular ways. More will.
If you back up and squint, the consumer crypto trading category is one of the strangest mismatches in any consumer technology: a category where the customers are losing money 93% of the time, and the products designed to help them have track records that range from “respectable” to “we are unable to locate the funds.”
I don’t think the answer to that is more aggressive marketing of better-sounding strategies. I think the answer is to build a product that takes the two things the institutions have — execution discipline and regulatory standing — and ports them down to the people who actually need them, without making them take on risks they didn’t sign up for.
That’s the entire pitch for AutoCoin in one paragraph. The rest is implementation.
Two unpopular decisions
I made two early decisions that any reasonable observer of the crypto space would have said were unforced errors. I’d make them both again.
The first was to get registered. AutoCoin is a registered Money Services Business with the U.S. Financial Crimes Enforcement Network — FinCEN — with AML and KYC compliance built in from day one. Almost none of our competitors are. Setting this up cost time, money, and several months of moving slower than I wanted to. It also means I sleep better, and it means that when an institutional partner or a serious investor asks me the compliance question, the answer fits on one slide.
The cheap reading of the regulatory environment in crypto is: “the rules are confused, just stay nimble.” The expensive reading — the one I think is right — is that every financial category, without exception, ends up regulated, and the only question is whether you helped write the rules from the inside or got caught on the outside when they landed. I want to be inside that conversation.
The second was to never hold user funds. AutoCoin connects to a user’s existing exchange via API with trading permissions only. No withdrawal access. No deposits to us. The user’s money never leaves the exchange they already chose to trust.
This is not a small architectural choice. It costs us optionality on certain product features. It means we can’t do certain kinds of yield products that a custodial competitor could. It also means that the worst-case version of an AutoCoin failure is “the bot stopped trading” — not “the bot took your money to the Bahamas.”
After what we all watched happen to FTX, I’m not interested in re-running that experiment from the other side.
What “AI” actually does here
I have to use the word “AI” because that is the language of the category. But I want to be specific about what AI does in our product, because the term has become a free pass to oversell.
Here is what AutoCoin’s AI does:
- Picks among a known set of trading strategies, given the user’s risk tolerance and the current volatility regime.
- Executes those strategies consistently — sizes positions, enforces stop-losses, refuses to chase, holds the line on rules a human under stress wouldn’t.
- Runs all of this 24/7, on a small number of well-understood exchanges, with full audit trails.
Here is what AutoCoin’s AI doesn’t do:
- It does not predict which token will go up next week. No model does that, despite a lot of marketing to the contrary.
- It does not promise a return. It is, in fact, illegal to do so, and also dishonest.
- It does not replace the user’s judgment about how much of their money they are willing to risk.
The honest sales pitch for AI in retail trading is not “we’ll find you the best trades.” It’s “we’ll execute the boring discipline you would do yourself if you were superhuman and never slept.” That’s the version we’re selling, and we’re going to keep selling it that way. I’ve written separately about what AI is and isn’t doing for retail traders, in more detail than this essay has room for.
Why I’m writing this
If you’re a retail trader and you’ve been burned by either the market or by a platform that lied to you about what their bot could do, I want you to know there is at least one operator in this category trying to do this seriously. I’d like the chance to earn your trust. (autocoin.ai)
If you’re an investor, a partner, or a journalist, the short version is: AutoCoin is fully operational, registered, growing slowly and honestly, and I think the next five years of this category belong to the small handful of companies that chose compliance and safety architecture before the market forced them to. I’d like to be one of them. That argument has its own essay: The Regulated Crypto Decade.
If you’re another founder building in or near this space — especially the under-funded, over-prepared kind, the ones doing the unglamorous parts first — let me know. There are more of us than the timeline suggests. We should know each other.
Either way: thanks for reading. The product is at autocoin.ai. I’ll keep writing here about what we learn building it.