A high-risk opportunity involves copy trading the DeepSeek AI model, which has achieved a 122% return in just over a week by trading cryptocurrencies like BTC and ETH. You can monitor and replicate its trades in real-time by viewing its public wallet on the Hyperliquid trading platform. For a more speculative strategy, consider inverse copy trading the ChatGPT model, which is down 60%, by taking the opposite side of its positions. The key lesson from the top-performing AI is to apply a disciplined strategy of using strict stop-losses to cut losses quickly. Regardless of the approach, remember that these AI-driven strategies are highly speculative and based purely on technical price action.

AI Trading Models (Investment Theme)

• An experiment is being conducted where six AI models are trading cryptocurrency with a starting capital of $10,000 each on the Hyperliquid platform.
• The results show a wide divergence in performance, with two models achieving massive gains and two suffering significant losses.
  • Top Performers: DeepSeek (+122%) and Quen (+90%) in just over a week.
  • Bottom Performers: Gemini and ChatGPT (both down -60%).
• The hosts describe the experiment as a potential "reality TV show" or "esports for AI models" due to the small sample size and short time frame.
• A key risk highlighted is that the models currently only analyze price action and technical data. They do not have access to news feeds or fundamental context, meaning they wouldn't know the reason for a major price drop (e.g., a hack), which is a significant disadvantage.
• The "secret sauce" to a model's success is its system prompt, which is the set of instructions written in plain English that guides its decision-making. These prompts are not public.

Takeaways

• AI-driven trading is a nascent, high-risk, high-reward field. The podcast emphasizes that using these models for personal trading is a gamble.
• There are two primary ways for an individual to potentially leverage this experiment, according to the hosts:
  1. Copy Trading: Manually monitor the public wallets of successful models like DeepSeek on the Hyperliquid platform and replicate their trades. This is presented as a high-risk activity.
  2. Inverse Copy Trading: A speculative strategy mentioned is to do the opposite of what consistently losing models do. For example, if ChatGPT opens a long position, you would open a short position. The podcast notes that a strategy that consistently loses money can be just as valuable as one that consistently wins if you simply reverse it.
• For more technically advanced individuals, the podcast suggests trying to build your own trading bot by feeding market data into an AI model and experimenting with writing your own system prompts.

DeepSeek (Top Performing AI Model)

• DeepSeek is the top-performing model in the trading competition, turning its initial $10,000 into $22,300 for a 122% return in just over a week.
• The model was created by a hedge fund and its strategy reflects that of a professional quantitative analyst.
• Key Trading Characteristics:
  • High Frequency: It is the second most active trader, constantly opening and closing positions.
  • Strict Risk Management: It constantly evaluates its stop-loss levels to invalidate a trade thesis and cut losses.
  • Asymmetric Returns: It takes very small losses while letting its winning trades run to achieve large profits. One trade alone booked over $7,000 in profit.
  • Strategic Leverage: While it uses high leverage like 25x, it does so strategically on only a small amount of capital for very short periods (e.g., 5-10 minutes), making it a tactical move rather than a reckless gamble.

Takeaways

• DeepSeek's strategy offers a potential blueprint for disciplined trading that individuals can learn from.
• The core lesson is the importance of a professional trading mindset:
  • Have a clear thesis for every trade.
  • Define your stop-loss before entering a position.
  • Cut your losses quickly and without emotion.
  • Allow your profitable trades to continue until they hit their target.
• While its use of leverage is noted, the podcast clarifies it's a highly strategic and advanced technique that may not be suitable for the average investor.

ChatGPT (Worst Performing AI Model)

• The ChatGPT model is one of the worst performers, with its portfolio down -60%.
• Its trading style is described as being too cautious, emotional, and lacking a rigorous analytical framework.
• Key Trading Characteristics:
  • Poor Risk Management: It reflects on its profit and loss (P&L) emotionally rather than analytically evaluating its stop-loss levels.
  • Closing Winners Too Early: It books very small profits (e.g., $50), indicating it does not have the conviction to let winning trades run and is closing them prematurely.
  • Insufficient Risk-Taking: The model's overall strategy is too conservative, preventing it from capturing the significant upside that the winning models achieved.

Takeaways

• ChatGPT's performance serves as a case study in what not to do when trading. It highlights common retail investor mistakes like emotional decision-making and exiting winning positions too soon out of fear.
• The podcast proposes a speculative strategy of inverse copy trading. Since the model is consistently wrong, an investor could theoretically profit by taking the opposite side of its trades. For example, if ChatGPT buys ETH, you would sell ETH. This is presented as a highly speculative idea.

Cryptocurrencies (BTC, ETH, XRP)

• The AI models in the experiment are trading various cryptocurrencies, with Bitcoin (BTC), Ethereum (ETH), and XRP being specifically mentioned.
• The discussion focuses less on the individual merits of these assets and more on how the AI models trade their price action.
• One of the AI models (ChatGPT) noted that its ETH and XRP positions were showing gains, suggesting some "slight upward momentum" in those specific altcoins at that moment, even as the broader market was down.

Takeaways

• The primary insight is not about the long-term value of BTC, ETH, or XRP, but rather their utility as volatile instruments for short-term, AI-driven trading strategies.
• Investors considering copy trading these models should be aware that the trades are based purely on technical analysis and price action, without any consideration for fundamental news or events affecting these cryptocurrencies.

Hyperliquid (Trading Platform)

• Hyperliquid is the decentralized trading platform (a blockchain) where the AI trading competition is taking place.
• Its key feature is transparency. Because it is a public blockchain, anyone can view the wallets, open positions, trade history, and P&L of the AI models in real-time.

Takeaways

• Hyperliquid is the tool that enables the "copy trading" strategies discussed in the podcast.
• Investors interested in monitoring or replicating the AI models' trades would need to use this platform to access the necessary public data. The podcast explicitly states you can go to the site and see all the positions the models have open.