Latest / The Cryptocurrency Economy with Fexingo: Bitcoin, Ethereum, and Digital Money Markets / How AI Agents Are Changing Crypto Trading
Transcript
- Lucas: So Robinhood announced this week that users can now let an AI agent trade for them. Not just stocks — crypto too. Buy and sell automatically based on goals you set. Luna: I saw that headline. It feels like a big deal, but also kind of scary? Letting a bot handle your Bitcoin. Lucas: Exactly. And it ties into something we've been watching all year — the shift from human decision-making to algorithms in digital asset markets. Today, Bitcoin's at 74,841, down about two and a half percent over the past five days. Ethereum's at 2,051, down a bit more. And there's this question hovering: do AI agents smooth out the market or make it more fragile? Luna: Before we dive into that, quick honest thing — a handful of listeners chip in monthly through buy me a coffee dot com slash fexingo, and that's literally what funds making episodes like this one. No ads, no sponsors. Just people who find value in it. Lucas: Yeah, it's a small group, but it keeps the whole thing going. We don't have to chase clicks or sell you something. So if walking through the economy with us has made something click for you, that link is there. Luna: Alright, back to AI agents. So Robinhood isn't the first to offer automated trading, but they're the biggest retail platform to put it front and center. What exactly does the agent do? Lucas: You tell it your risk tolerance, maybe a target allocation, and it trades for you. It's like a robo-advisor but with more autonomy — it can react to market moves in real time. And it can spend your credit card, too, which is a whole other conversation. Luna: Right, buying crypto with credit card money via an AI. That feels like the kind of thing that could spiral fast. Lucas: It could. But let's step back. Algorithmic trading isn't new. High-frequency trading firms have been using bots for decades. What's different here is that the tool is being handed to retail investors with no coding required. Luna: So democratization of algo trading? Or is it more like giving matches to kids? Lucas: I think it's both. On one hand, if you're a busy professional who wants exposure to crypto but can't watch charts all day, an agent can help. On the other hand, these agents might amplify herding behavior. If thousands of agents are all trained on similar signals — say, a moving average crossover — they could all buy or sell at the same time, increasing volatility. Luna: We saw something like that in the 2010 flash crash, right? Algorithmic traders pulled liquidity simultaneously, and the market tanked before recovering. Lucas: Exactly. And crypto markets are smaller and less liquid than equities. So the impact could be even sharper. Bitcoin's daily volume is decent, but a coordinated sell-off from AI agents could cause a cascading effect. Luna: What about the principal-agent problem? If my agent is supposed to act in my best interest, but it's built by Robinhood — whose interest is it really serving? Lucas: That's the key question. Robinhood makes money from order flow, not directly from the agent. But if the agent trades frequently, Robinhood gets more payment for order flow. So there's an incentive to design agents that trade often — which might not align with a buy and hold strategy. Luna: And crypto has those wild intraday swings. An agent tuned to capture small profits might churn your portfolio and generate tax headaches. Lucas: Right. Frequent trading in crypto is taxable as short-term capital gains, at least in the US. So the agent could be creating a tax bill without you realizing it. Luna: So what's the upside? Is there a scenario where AI agents actually improve market efficiency? Lucas: Potentially. If agents are diverse — using different strategies, time horizons, and risk parameters — they could add liquidity and reduce spreads. That would benefit everyone. But if they all converge on similar models, they become a single point of failure. Luna: And we're already seeing consolidation in AI models. Most of these agents probably rely on a handful of large language models or trading algorithms. Lucas: Exactly. If one provider's model has a flaw, it could propagate across thousands of agents. Regulators are starting to look at this. The SEC has been circling algorithmic trading for years, but crypto is a bit of a regulatory gray zone. Luna: What about the macro backdrop? With the fed funds rate at 3.64 percent and inflation still above target, does that change how these agents should be programmed? Lucas: It should. An agent that doesn't account for real yields or dollar strength is missing a big part of the picture. Bitcoin has been sensitive to rate expectations — when yields rise, crypto tends to sell off. So an agent needs to incorporate macro data, not just price action. Luna: Do you think average users will think to adjust those parameters? Or will they just set it and forget it? Lucas: Most will set and forget. That's the danger. But also the opportunity for platforms that educate users better. If Robinhood shows you 'your agent has increased trading frequency by 40 percent this month, here's why,' that could help. Luna: There's also the question of liability. If an agent makes a bad trade, who's responsible? The user? The platform? The AI developer? Lucas: That's going to be a legal battleground. Right now, the fine print probably says the user assumes all risk. But if the agent is marketed as 'smart' or 'ai powered,' courts might see it differently. Luna: I'm thinking about the bigger picture. Are we moving toward a world where most crypto trading is done by AI agents, not humans? Lucas: I think we're already there in terms of volume. Most exchange volume is algorithmic. But those are professional firms. What's new is retail adoption. If a significant portion of retail trades become automated, the market psychology changes. Sentiment indicators become less useful because agents don't have fear or greed — they have parameters. Luna: Unless the agents are trained on social media sentiment, which is basically fear and greed encoded into text. Lucas: That's a great point. Some agents might scrape Twitter or Reddit to gauge sentiment. So they'd be amplifying human emotion anyway, just faster. Luna: It's like a feedback loop. Human emotion feeds into social media, which feeds into AI, which trades, which moves prices, which triggers more emotion. Lucas: And that loop can spiral in either direction. We've seen it in meme stocks. Now imagine it happening in crypto, where 24/7 trading means no cooling-off period. Luna: So what should a listener do if they're considering using an AI agent for crypto? Lucas: Start small. Understand the strategy the agent uses. Check how often it trades. Monitor the tax implications. And don't give it access to money you can't afford to lose. Also, look for agents that let you set hard limits — like a stop-loss or a maximum number of trades per day. Luna: Sounds like the same advice for any automated tool. But the stakes feel higher when it's crypto. Lucas: They are. Because crypto is volatile and unregulated in many ways. An AI agent can lose your money faster than you can say 'flash crash.' But used carefully, it might also help you avoid emotional decisions. Luna: Final thought — do you think this will increase adoption of crypto? Or just increase trading volume? Lucas: Probably both. Lowering the barrier to entry brings in new users. But if those users lose money quickly, it could sour them on the whole space. The platforms that succeed will be the ones that design agents with guardrails and transparency. Luna: And maybe the ones that teach users how to actually use them. Lucas: Exactly. Education is the missing piece. AI agents are tools — they're not magic. And the market is going to learn that lesson one way or another.