The futures market just got its first AI chauffeur, and you're still deciding how much gas to put in the tank.

The Summary

The Signal

Walt is opening the door to AI-powered perpetual futures trading, plugging conversational agents like Mira, ChatGPT, and Claude directly into leveraged crypto positions. You'll be able to tell an AI agent to open a 10x long on ETH or short SOL with a stop-loss, then walk away. The infrastructure treats perpetuals like any other financial primitive that agents can manipulate through natural language.

The guardrails matter more than the feature. Order execution is opt-in, not default. You fund a separate trading account specifically for AI use. The agent can't touch your main wallet. It can't go rogue and lever up your life savings while you sleep. This design choice shows Walt understands the difference between "agents can do this" and "agents should do this automatically."

"Order execution would be off by default, with AI agents restricted to a dedicated trading account that users choose to fund."

Perpetuals are already the Wild West of crypto derivatives. No expiration dates. High leverage. Liquidation cascades that can vaporize positions in minutes. Now add AI agents to the mix:

  • Agents making split-second decisions based on model outputs, not human judgment
  • Potential for increased market volatility if many users run similar agent strategies
  • Questions about liability when an AI makes a trade that goes sideways

The pitch is democratization. Sophisticated trading strategies, previously reserved for quants and prop shops, now available to anyone who can describe what they want in plain English. Ask Claude to implement a momentum strategy with risk management. Tell Mira to scalp funding rate arbitrage. The barrier to entry drops from coding skill to prompt engineering.

But democratizing access to complex instruments doesn't democratize the outcomes. Retail traders already get wrecked in perpetuals markets designed by professionals who understand order flow, liquidation mechanics, and how to extract value from information asymmetry. Giving them AI co-pilots doesn't level that playing field. It just automates the losses at higher speed.

The Implication

Walt's move puts a sharper point on the agent economy's central tension. We're building tools that make sophisticated actions trivially easy to initiate, but we're not making the consequences any less severe. An AI can open a 20x leveraged position just as easily as it can check the weather. The interface complexity collapses, but the financial physics stay the same.

If you're building in this space, the Walt model is worth studying. Segregated accounts, opt-in execution, and clear boundaries around what agents can touch. That's not just good UX. It's the difference between "AI trading assistant" and "AI trading liability." Watch how users actually interact with this. If adoption is slow, the problem isn't the tech. It's trust. And trust in AI-driven financial decisions isn't something you can prompt engineer your way into.

Sources

Crypto Briefing | The Defiant