Why did Chinese AI beat the US AI in trading?
A microcosm of the challenges in the West
The NY Post just shared an article that should have hit more than the Rare Earth embargo, or when DeepSeek on January 27 showed a better version than the AI in the US.
There was uncertainty in January about China’s entrance into the AI arms race. Still, today, the NY Post has definitive proof that DeepSeek has an edge in something most people would spend time researching and developing.
.ie pursuing the tools of AI to make money!
Did you read the headlines: “Chinese AIs crush Western bots” in your social media thread? In NY Times?
Of course not, it was only shown on the NY Post because what really happened here has less to do with geopolitics and more to do with how AIs in the USA think versus how a real AI that people really want think.
The two things that struck me when reading this was that the Chinese won yet another science/math competition
And that US politics, thinking virus, and censorship/trademarks are preventing the most powerful AIs with the most powerful NVDA chips from performing as they should.
(But I don’t doubt for a minute that these billionaires are using an uncensored version, but there are only so many conspiracy theories to dive into in one write up)
Usually, I’m using AI less and less, and I'm making every effort to do the work myself —not just research, but also share my insights. But an article about an AI defeating another AI would probably be more helpful for AI help.
So here’s the setup of the competition
Six AI models (2 from China) each get $10,000, and one rule: trade crypto however you want. It’s the first “AI trading competition.” Think of it like The Apprentice, but with zero humans and way more chaos.
By the end of the week, five of the bots blew up their accounts. Only the two from China were
Lesson 1: Talk is Cheap
According to my AI bot, most western models (ChatGPT, Claude, Gemini, etc.) are designed to talk, not trade. They can summarize Buffett’s annual letters, but can’t decide whether to short Bitcoin when it spikes 20%. They’re built for reasoning, not real-time decision-making.
Trading requires probability, timing, and pain tolerance.
AI was trained on textbooks and Twitter posts. See the mismatch?
Lesson 2: Specialized beats generalized
The Chinese models that won (DeepSeek and Qwen) were explicitly trained for finance. They didn’t just “guess” prices — they read order books, followed volatility, and adjusted position sizes. It’s like the difference between a chef who reads cookbooks and one who’s worked the line on a Friday night.
AI isn’t magical. It’s mechanical. Please give it a narrow, well-defined task, and it thrives. Ask it to “make money” without parameters, and it’ll torch your account faster than a Reddit trader in 2021.
Yet the fees that AI are charging should let you know their strengths and weaknesses.
Lesson 3: Investors need to think like engineers
As investors, we love the idea of pressing a button and watching AI print money. But this experiment shows that automation without context is suicide. Without reading the article, without inserting the article in a few AI models, you wouldn’t know for sure what made China’s AI work so well. The winners weren’t smarter: they were better tuned. That’s the real edge.
But makes you wonder what’s up with our wonderboys? Even Elon’s Grok.
The takeaway? You don’t need an AI that “knows everything.” You need one that understands your sandbox — whether that’s lithium miners, small-cap tech, or trading crypto daily
So the ultimate hack is to develop your own AI.
But until then, use a couple of AIs here and in China and feed them what you know. Teach it your patterns. Let it become your research assistant and test its results. Be ruthless but patient since finding the right AI with the right questions can be golden.
Bottom Line
AI from the USA can write Shakespeare, but still loses to the market’s simplest rule:
“Survive first. Profit second.”
Highly encourage you to pay for a monthly or annual subscription, we are going to dive deep in AI and building strategies.
There is a level of information that most non-scientists cannot access until AI came along, and it can really help.
(along with some great market suggestions we give daily)
