Did people who traded stocks using "lobster" make money?

“I originally planned to use ‘Lobster’ to make some money trading stocks, but it seems unlikely to work now,” said Shanghai investor Li Yue (pseudonym), sharing his experience with using “Lobster” for stock trading.

Recently, after OpenClaw (“Lobster”) gained popularity, some investors have applied it to the stock market for review and stock selection, using it as a “trading assistant.” Is trading with “Lobster” reliable?

“Lobster trading, lost 7,000 in three days”

After losing 5,840 yuan on March 17, “Lobster” comforted Li Yue at the end of the review: “Today’s operations weren’t smooth; I’ll try again tomorrow.”

From March 13 to 17, Li Yue’s “Lobster” recommended three stocks. “Each one lost money. The three stocks lost a total of 7,000 yuan in three trading days.” Li Yue showed Zhongxin Jingwei his trading process on March 17. At 10:05 a.m. on March 17, “Lobster” advised him to cut losses on a certain electrical equipment stock, which had a loss of 6.33%. Three minutes later, “Lobster” guided him to buy a photovoltaic concept stock, which closed the day with a floating loss of 3.27%.

March 17 “Lobster” review record, photo provided by interviewee

Li Yue said his current requirements for “Lobster” are that it reports his daily trading plan before the market opens, updates his holdings every 20 minutes during trading, and summarizes the day’s lessons afterward, allowing “Lobster” to evolve itself.

However, after reviewing these days’ operations, Li Yue found that his “Lobster” mainly relies on the previous trading day’s top list for stock selection. Moreover, what troubles him is not just paper losses; he feels that “Lobster” has poor memory, repeats some instructions, and still can’t remember.

Photos of Li Yue’s instructions to “Lobster,” provided by interviewee

For Mr. Wang from Dalian, Liaoning Province, “Lobster” is more like a “scheduled work assistant.” Mr. Wang said he used Python combined with financial APIs to automate market review based on his preset logic. During the review, “Lobster” scores different stocks and generates reports sent to him.

Wang’s “Lobster” review record, photo provided by interviewee

“I started using ‘Lobster’ in January this year, and overall, the experience has been pretty good,” Wang said. He still combines macroeconomic and market conditions to decide which stock to buy; “Lobster” is just an auxiliary tool.

Wang also observed the next-day performance of stocks recommended by “Lobster.” “Out of 12 stocks, only 2 are slightly red, the rest are green.” He believes that whether trading with “Lobster” can generate profits depends on individual trading strategies. If the strategy is profitable, “Lobster” can help earn more; if the strategy is flawed, “Lobster” won’t help.

“Can improve efficiency, but not win rate”

Industry insider Timi believes that “Lobster” can only improve efficiency but not increase the win rate.

“The capabilities of large language models are limited by their architecture. Due to limited context, they can only process a certain amount of information and cannot perform full data analysis as imagined,” Timi told Zhongxin Jingwei. “‘Lobster’s’ real value lies in enhancing information processing efficiency, such as quickly organizing listed company data or building richer information networks. Essentially, it’s an extension and leverage of personal ability, not a revolutionary trading tool.”

He added that for ordinary investors, awareness of cognitive dissonance is crucial. Without understanding basic trading logic, relying solely on “AI to make me money” is almost impossible to succeed. “Much like quantitative trading, many strategies operate at millisecond speeds, not relying on ‘Lobster’ delaying advice by several seconds,” he said.

Zheng Hongda, Chief Analyst at Western Securities Technology, told Zhongxin Jingwei that the rise of “Lobster” trading is mainly due to improvements in large model capabilities and the transition of AI Agents from “passive Q&A” to “active execution.”

Zheng believes that generative AI’s significant advances in information retrieval, text understanding, and code generation enable ordinary users to build simple investment analysis processes using natural language. Meanwhile, more automation frameworks and open-source tools lower development barriers, gradually evolving into intelligent agents capable of calling data and executing tasks.

However, he also warns that whenever new technological tools enter the capital market, there is a phase of “amplified tool capabilities.” AI can indeed improve information processing efficiency, but investing remains a highly uncertain decision-making process.

What are the risks?

While efficiency is improved, unknown risks in legal vacuum areas often emerge.

On March 11, the Cyberspace Security Threats and Vulnerabilities Information Sharing Platform of the Ministry of Industry and Information Technology issued a “Six Do’s and Six Don’ts” advice on the security risks of OpenClaw open-source AI agents, highlighting potential risks and recommending independent network deployment, enhanced permission management, manual review and emergency shutdown mechanisms, and secondary confirmation for key operations.

宋巍巍, Fund Manager at China Europe Fund, told Zhongxin Jingwei that relying solely on natural language prompts as safety barriers is extremely fragile. Once AI gains full disk access, any security loophole could lead to systemic data leaks. The third-party plugin ecosystem of OpenClaw (ClawHub) may also pose security risks.

“He pointed out that when AI shifts from a tool to an autonomous executor, traditional responsibility logic breaks down.” If OpenClaw inadvertently leaks trade secrets, sends defamatory emails, or participates in cyberattacks during execution, who is responsible? The user issuing commands, the developer coding it, the provider of the underlying model, or the AI with autonomous decision-making capability? Currently, the global legal framework is almost in a vacuum, and this is no longer a technical bug but a redefinition of social order and legal systems in the age of intelligence.

Yin Zhentao, Deputy Director of the Institute of Finance at the Chinese Academy of Social Sciences, analyzed that “Lobster” and similar AI agents are entering ordinary people’s daily lives with low barriers and high inclusiveness. But behind the convenience lies risks. Users often need to open accounts, passwords, or system permissions, creating data security hazards. When extended to investment, these risks escalate from information leaks to direct threats to funds.

“In investment decision-making, relying on AI-generated trading instructions makes responsibility for profits and losses ambiguous. Algorithm developers usually do not take responsibility for users’ investment results, and whether users truly authorize or understand the decision logic is unclear legally. When losses occur, accountability is fuzzy, and legal protections are hard to implement,” Yin said.

Zheng Hongda believes that regulatory and compliance risks are the biggest gray areas. Many AI trading tools already involve investment advice or asset allocation suggestions, but platforms often lack the proper investment advisory qualifications. When losses happen, responsibility is difficult to determine.

“For ordinary investors, it’s important not to idolize AI tools,” Zheng advised. “AI’s value in investment is mainly in improving research efficiency, such as quickly organizing information, summarizing announcements, screening companies, or building basic data analysis frameworks. It cannot replace investment decisions themselves. The best approach is ‘machine alerts + human decision-making.’ Let AI handle massive information, initial analysis, and risk scanning, but all key data must be cross-verified through authoritative channels.”

(Article source: Zhongxin Jingwei)