The Current State of AI Trading Isn't Autopilot

One of the biggest consumer trends over the last decade has been the rise of trading. Robinhood made public markets accessible, Coinbase brought millions of people into crypto, and prediction markets have created entirely new ways to express a view on what is happening in the world. For a younger generation, trading isn't just investing anymore. It's a skill they're actively trying to develop and, for many, a way to build wealth.

At the same time, markets have become dramatically more complex. Information moves faster than ever. An earnings call, macro announcement, regulatory update, on-chain transaction, research report, or even a single social post can move a market almost instantly. The amount of information available to traders has exploded, but our ability as humans to actually process all of it hasn't. Keeping up with everything that could matter isn't just difficult—it's basically impossible.

That's what makes AI such an exciting opportunity for trading. Unfortunately, it's also why AI has become one of the most overhyped parts of the industry.

Today, a lot of AI trading products sell some version of the same story: describe your thesis, let AI do the research, backtest it, deploy an autonomous agent, and sit back while it trades for you. It's a great demo and an even better marketing pitch. It's also not where the technology is today. If someone had actually built an AI that could consistently generate alpha with no human judgment, they probably wouldn't be selling you a monthly subscription. They'd be running one of the largest investment firms in the world.

The problem isn't AI. I think AI is going to fundamentally change how people trade. The problem is pretending we've already solved autonomy when we haven't.

The best comparison is probably software engineering. Coding is one of the clearest examples we have of AI dramatically improving how people work. Great engineers use AI constantly now. They use it to explore ideas, write code, debug problems, understand unfamiliar codebases, and move significantly faster. But great engineers aren't blindly shipping everything an AI agent produces straight into production. They still review the work, test it, challenge assumptions, and understand what they're shipping. AI makes great engineers better. It doesn't eliminate the need for great engineering judgment.

Trading should work the same way, and in many ways it's even less forgiving. You can roll back bad code. You can't roll back a bad trade.

That's why I think we're asking the wrong question when it comes to AI and trading. Everyone wants to know when AI will replace the trader. I think the much more interesting question today is: how can AI make the trader dramatically better?

There is so much that AI can already do. It can help traders find information they would have otherwise missed, connect signals across markets, pressure-test a thesis, understand risk before putting on a position, backtest ideas more effectively, and build repeatable workflows that improve over time. None of that requires pretending we've built some superintelligent system that knows where markets are going. It requires building tools that take everything AI is already great at and putting those capabilities into the hands of traders.

That's ultimately where I think AI creates the most value today. Not by making every decision for you, but by helping you make better decisions yourself. A trader who can process more information, ask better questions, test more ideas, understand their risk, and react faster should have an advantage over someone who can't.

That's the philosophy behind what we're building at Button. We're not trying to build a magic black box that promises to make you money. We're building AI-native workflows that help traders process information faster, develop stronger conviction, understand their risk, and ultimately make more informed decisions.

The edge in trading has always come from making better decisions than everyone else. AI doesn't change that. What it changes is how quickly you can get to those decisions, how much information you can consider, and how effectively you can test your thinking before putting capital behind it.

Maybe one day trading really does become completely autonomous. AI is moving quickly enough that I wouldn't bet against it. But that's not the current state of AI trading, and I don't think we should pretend that it is.

For now, the best traders aren't going to be the ones who hand everything over to an AI agent. They're going to be the ones who learn how to use AI to become better at what they already do.