Are 200 AI Trading Products Useless? A Unique Take from a Former Co-Founder of Foresight Ventures

Author: WhiteForest, Former Co-Founder of Foresight Ventures

I managed hundreds of thousands of retail traders at Bitget before I understood this matter.

Last month, an old friend invited me to dinner. He has been trading for seven years, managing only a small amount of his own money, but he hasn't outperformed BTC in the past two years. It's not because his judgment is poor; among the people I know, he is actually the one who thinks the clearest. Halfway through dinner, he asked me what I had been working on lately. I didn't explain much, just pushed my phone over to him and let him try it out directly.

He typed in the chat box: "I want to exchange half of my ETH for SOL."

The product didn't execute immediately; it first informed him of three things: First, he had made the exact same ETH→SOL exchange four times in the past 12 months, getting it wrong three times with an average loss of 11%. Second, before making this move last time, he had explicitly stated, "If SOL/ETH falls below 0.038, I will definitely exchange back," and it did fall below that, but he didn't exchange. Third, his current SOL position had already exceeded the hard limit he had set for himself by 4 percentage points.

Then it asked one question: "Are you making a decision now, or are you repeating something you've already done four times?"

He stared at the screen in silence for about 15 seconds, then looked up and said, "Damn."

In those 15 seconds, he wasn't looking at market signals; he was seeing for the first time what kind of trader he really was. The things he had done, the words he had said, the lines he had set, the line he had crossed—all laid out before him at the moment he was about to make a mistake. I kept pondering one question: Why is it that none of the 200 AI Trading products on the market can help him see himself clearly?

Is the reason so many retail traders are losing money due to lack of judgment?

To answer this question, we need to first understand why retail traders are losing money.

I managed five business lines at Bitget and reviewed the real trading behavior data of hundreds of thousands of retail traders. Before that, I thought retail traders lost money due to insufficient judgment, lack of information, inadequate analysis, and insufficient signals. This is also the problem that all products on the market are trying to solve. But the data told me otherwise.

The Bank for International Settlements studied crypto trading data from 95 countries between 2015 and 2022, concluding that 73% to 81% of retail traders lost their principal. A 2025 survey covering 1,005 crypto retail traders showed the figure at 84%. A decade has passed, tools have become more numerous, and information has become more transparent, yet this number has never truly changed.

I found the answer in the data. Those who are losing money do not have a judgment problem. They set stop-loss lines, but when the price falls below that point, they are in a meeting. When there is movement on the blockchain address, by the time they follow in, the price has already moved 20%. They exceed their own hard limit on position size, but since the market looks decent, they tell themselves to wait a bit longer. They made the same mistake three months ago and six months ago, but because no one reminded them, they completely forgot and made the mistake again. Their judgment was correct, yet the money still didn't come in.

The real problem for retail traders is not a lack of judgment; it's that at critical moments, there is no one to help them lay out their past behaviors in front of them, no one to help them recognize what kind of trader they truly are.

200 AI Trading Products Have Not Solved the Fundamental Problem

I have tested the mainstream AI Trading products on the market and watched friends get burned. In summary, there are four types:

The first type is strategy cards.

Open the app, and there is a row of cards: BTC trend 137%, SOL breakout 68% win rate, with "AI optimized" written in the upper right corner. What did the AI do? It simply placed the strategies that made money in the past 90 days at the forefront. That 137% strategy is the best one from a 90-day backtest, which means it has already consumed the alpha from this round, and the moment you buy in is precisely when it should not be bought. This is not AI strategy selection; it is an automated version of survivor bias. This has existed since 2018, back then it was called copy trading, and it didn't require $29 a month.

The second type is copy trading agents.

AI helps you track smart money, following the top 100 wallets in real-time. On-chain confirmation takes 12 seconds, the agent detects it in 5 seconds, and pushes it to your phone in 3 seconds. By the time you see the pop-up and click to confirm in 4 seconds, that "smart money" has already entered the market 24 seconds ago, and you are the last one to take over. This is still the good scenario. The bad scenario is that wallet belongs to a market maker who is flipping assets back and forth, creating signals, and waiting for thousands of copy trading agents to swarm in before dumping on everyone. Smart money is not an identity; it is a label attached to past profits. Its operations from yesterday are on-chain, but no one knows what will happen tomorrow, not even itself.

The third type is AI analysis chat boxes.

Open the app, and in the chat box, you type, "Should I buy BTC now?" It generates an 800-word analysis, including RSI, MACD, on-chain fund flow, social sentiment, and historical comparisons, which sounds very professional. The truth is one sentence: ChatGPT in a shell, using 5 free APIs. Your "on-chain data" is from Etherscan's public page, your "social sentiment" is from the free LunarCrush API, and your "historical comparisons" are stories fabricated from training data by a large model. What you pay $29 a month for is an analysis packaged from several free APIs. Institutions like Citadel spend $700 million a year on real, market intelligence that you will never see. This is not a difference in winning or losing in the same game; it is fundamentally not on the same information level. The darker side is this: every word you type in the chat box, every suggestion you click, whether you executed it, whether you made money—this data is all packaged and sold to proprietary traders. You think you are using AI, but in reality, AI is using you as training data, helping these institutions study you more thoroughly.

The fourth type is a castrated true LLM.

This one angers me the most. They have genuinely integrated Claude or GPT-4, paid for the model, and the founders truly understand AI. Then they did one thing: downgraded the LLM to a chat box, hardcoding 5 fixed strategies in the backend. The most valuable capability of the LLM, generalization, was thrown away. No matter what the user says, it only calls one of these 5 functions. It won't ask, "You've done this action 4 times already," it won't ask, "Your position exceeds your set limit," it won't ask, "Are you making a decision or repeating a habit?" It just executes. The smartest tool is used to create the dumbest product. Why? Because the founder's KPI is MAU, not the quality of your decisions. You are not a user; you are just a number in their pitch deck labeled "active user."

Four types of products, four directions, but there is one thing they all fail to do: present your own behavioral patterns to you at the moment you are about to make a mistake.

You Think You Are Using AI Trading, But In Reality, You Are Just Being Sampled by AI

For fifty years, the market has been harvesting retail traders. Retail traders survive on "instability"; they get tired, forget, skip a week, and occasionally get lucky before exiting. These human flaws have always been your armor; the market cannot achieve complete precision against an opponent with unstable behavior.

AI is now systematically patching these vulnerabilities. Always online, every hesitation is recorded. If you made the same mistake four times in 12 months, institutions like Citadel's machines will know you are about to make that mistake again 0.5 seconds before you realize it yourself. You spend money on these AI tools, and the result is not that you become stronger; it is that these institutions understand your behavior better, knowing when you will act impulsively and when you will cut losses, and then waiting precisely for you at that moment.

Behind these products, there is a term that has been thoroughly worn out.

The Term "Empowering Investors" Has Been Stolen

Every AI Trading product's landing page is stamped with the phrase "empowering investors"; every founder's pitch includes these four words, and every LP deck has it on the third page. But this term has always been packaging, not a promise.

True empowerment is allowing a person who works during the day, picks up kids in the evening, and sleeps at 3 AM to genuinely possess three things that only institutions have: cognition, judgment, and execution, truly attached to themselves. False empowerment is turning these three things into cheap replicas—a chat window, a copy trading button, a strategy card—packaged as "empowerment" and sold to you, making you feel upgraded, only to take your money with greater precision.

True empowerment is giving you the same level of information and judgment capabilities as institutions like Citadel. False empowerment is giving you a tool that looks impressive, making you think you have upgraded, while in reality, you are still that person being harvested.

If this situation is not broken, what will happen next?

What Will Happen in 18 Months

In the next 18 months, at least 180 of those 200 AI Trading shells will die, and the remaining 20 will be acquired by proprietary trading firms, with user behavior data flowing into Citadel's training pipeline. In the next five years, retail traders without agents will be like grandmas walking into brokerage offices in the crypto version, while retail traders with shells will be infantry with plastic swords, dying faster, more precisely, and cheaper. This is not a prediction; it is an inevitable outcome if we continue down this path. But this path is not the only way to go.

No One Has Done It Right, So I Am Here to Do It

The thing my friend opened was made by me; it is called Mojo.

Why did I create it? I have held three positions: at Foresight Ventures, investing in AI infrastructure; at The Block, observing the shape of the entire industry; and at Bitget, managing five business lines and reviewing the real trading behaviors of hundreds of thousands of retail traders. Three completely different perspectives told me the same thing: what this generation of AI can truly achieve, no one is doing.

What is that thing? Returning to those 15 seconds at the beginning. Mojo did not give my friend more market signals or tell him whether SOL would rise; instead, it laid out his trading records from the past 12 months in front of him—his words, the lines he set, the line he crossed—at the moment he was about to make the fifth mistake. In those 15 seconds, he finally possessed something that only institutions have: a system that truly understands his trading behavior, attached to himself.

This is what this generation of AI can truly achieve. It is not a fancier order button; it is genuinely, completely, and uncompromisingly putting the three things that only institutions have onto you. Cognition can be loaded into an agent. Impulsive judgment can be halted by a conversation. Execution can be replaced by an API that runs around the clock. This is not a PPT; it has physically become possible for the first time.

And the $1.7B of VC money is doing precisely the opposite. It is here to make VCs write beautiful investment returns, to allow Citadel to spend $50M in 36 months to buy the cleanest product among them. What they are buying is not the product; they are buying every hesitation, every impulse, and every loss cut from every retail trader over the past 1,000 days. After Citadel acquires this data, it will build the behavioral patterns of every retail trader into its trading system, knowing when you will impulsively buy in, when you will panic sell, and then precisely harvest retail traders at that moment.

If this generation of AI only makes Citadel stronger and Jane Street richer while retail traders remain the precisely harvested ones, then the so-called "progress" brought by this generation of AI is merely progress for institutions. For retail traders, this is not progress; it is a technological robbery packaged as "progress." I will not be part of that, nor will I invest in that stuff.

Once you see something, you cannot pretend you didn't see it. I see it, so I am here to do this. To return AI trading that belongs to retail traders back to them.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.