The Daily Economic News reporter: Chen Chen

This image is suspected to be AI-generated

When AI “lobsters” (OpenClaw, an open-source AI agent developed by Austrian programmer Peter Steinberg) are combined with stock trading, a brand-new AI-assisted trading model has emerged.

A reporter from The Daily Economic News (abbreviated as “Daily Economic News reporter”) noticed that, with investors’ authorization, “the lobster” has gradually moved into core stages such as monitoring the market, post-trade review, and stock selection. However, real-world returns have shown extreme polarization: some interviewees said that after a 200,000 yuan full-discretionary mandate to “the lobster,” they lost 80,000 yuan; others said they kept beating the broader market through automated trading; and still others claimed “monthly returns of 90%”……

Why does “lobster” stock trading end up with “ice and fire” outcomes? Under the many restrictions of compliance and risk, can “lobster” stock trading break through existing limitations? Can it also open up a truly profitable new channel for ordinary retail investors?

Screenshot from the OpenClaw official website

1

“Lobsters” execute full-service live stock trading

Some people beat the broader market

but didn’t even earn back Token money

OpenClaw is fierce because it is not only a “chat tool for questions and answers,” but also an “executor” that can “take over” your computer and automatically call data to complete complex tasks. As long as the user issues an instruction, it can open software on its own, organize documents, or even reply to emails.

A Daily Economic News reporter learned from interviews that many users are already using “the lobster” to monitor the market, do post-trade reviews, and select stocks, allowing it to deeply participate in each link of research, decision-making, and trade execution.

However, the real-world results vary greatly. There are astonishing track records with monthly profits of 90%, as well as real cases of losing 80,000 yuan outright.

A well-known digital economics scholar and president of the DCCI Internet Research Institute, Liu Xingliang, shared a real case from his circle. Liu Xingliang told the Daily Economic News reporter that his friend invested about 200,000 yuan, delegated it entirely to “the lobster” to trade stocks, and the trades themselves were also carried out by “the lobster.” But in recent weeks, as the broader market declined, the account at one point lost 80,000 yuan.

Meanwhile, some investors have also found logical loopholes in the AI tool in live trading. A live-trading log shared by one user on a social platform shows that on the second day after using “the lobster” for automated trading, the user’s position lost 0.9% that day, but overall it beat the broader market. On the third day, the position lost 1.57%, while the Shanghai Composite Index fell 3.63% that day, yet it still beat by more than 2 percentage points. By the fourth day, the position achieved a profit of 0.8%. Even though the data shows it could beat the index, the user also pointed out issues exposed in live trading:

First, it confirmed with “the lobster” not to chase price, but it still chased. Second, “the lobster” seems to be operating with a T+0 mindset.

Another user who actually used it also told the Daily Economic News reporter frankly that A-shares use T+1 settlement, so the significance of using OpenClaw for quant trading is limited. U.S. stocks and Hong Kong stocks can use T+0, which might be more meaningful instead. The user said they are still in a stage of trying with small capital and debugging adjustments. Although they achieved small profits, they didn’t even recoup the cost of previously buying Token (a term; the smallest information unit when AI models process natural language).

2

“Monthly profits of 90%” were actually simulated trading returns

“Lobster” ≠ AI brain

It’s just the trading execution platform

How exactly was the “monthly profit of 90%” on social platforms achieved? In response, Jin Fengchun, managing partner of Saifu Investment Fund Management, restored the real technical path behind those high returns for the Daily Economic News reporter: it was actually a simulated trading contest for AI to trade U.S. stocks, with the trading asset being U.S. stocks.

Jin Fengchun emphasized: “Need to note that 90% was the highest return reached within a month. After subsequent pullbacks, the final closing return rate was 36%.” During this period, both the Nasdaq Index and the S&P 500 Index saw clear pullbacks and the overall market conditions were not good, but the simulated trading returns still far outperformed the equity indices.

Data source: Wind

When asked about the core reason for generating excess returns, Jin Fengchun was very direct in pointing out that it was not because “the lobster” itself is wise. Rather, after the contestants manually set the basic framework, they had Kimi, DeepSeek, or MiniMax, etc., write the specific trading strategies, pick the right stocks, and then hand everything over to “the lobster” for execution. In addition, achieving 90% returns in the short term largely came from the use of leverage in simulated trading of U.S. stocks.

“During this period of high returns, ‘the lobster’ only played the role of the AI’s technical framework and execution carrier; it did not directly screen for assets, provide trading signals, or complete backtesting verification.” Jin Fengchun told the Daily Economic News reporter that the core work—such as asset screening and trading strategy formulation—is done by other AIs. Humans only define the direction and basic framework for the AI; afterward, all strategy execution steps are autonomously completed within “the lobster.”

At the execution layer, Jin Fengchun explained that this simulated trade used the strategy from a domestic artificial intelligence model and was then followed by the AI placing orders automatically. At present, many domestic brokerages support the QMT trading system (QMT and PTrade are the mainstream domestic third-party quant trading terminals). With API integration (the process of enabling communication and data exchange between different software systems through application programming interfaces), automated trading order placement can be implemented, and domestic quant trading commonly uses this model as well.

3

“Lobsters” grabbing brokerages’ functions?

Tops like Tonghuashun “refuse to accept”

Guangfa Securities begins building a security sandbox in-house

As “the lobster” spreads to the C side, investors have tried to fully integrate it with broker trading software.

An industry insider told the Daily Economic News reporter that OpenClaw can be used as the “brain” of the strategy, responsible for data analysis and signal generation, while QMT serves as the “executor,” responsible for high-speed order placement and trade execution. According to what the Daily Economic News reporter learned, to do quant trading in the first place, you need to contact the brokerage to enable trading permissions for QMT or PTrade. Brokerages usually set capital thresholds of 500,000 to 1 million yuan, and some brokerages require lower thresholds.

“The role of OpenClaw in this process is actually to write code, and it still uses the official API.” A live-trading user told the Daily Economic News reporter that OpenClaw essentially only lowers the barrier to writing Python code.

However, broker institutions remain extremely cautious about direct integration of this kind of third-party AI tool. A brokerage IT person told the Daily Economic News reporter clearly that, based on existing regulatory requirements and industry practices, trading interfaces usually need to be strictly controlled within an authorized system to prevent unauthorized programmatic integration and potential market manipulation risks. The person pointed out that currently, Tonghuashun and broker in-house trading terminals have not yet provided APIs to “the lobster.” Brokerages defend against illegal third-party access through multiple technical and institutional measures and also monitor similar trading plug-ins.

A report published by Founder Securities on February 21, 2026 shows that it previously used OpenClaw to successfully test the Tonghuashun API interface

In addition, some users have told reporters that the “lobster” stock-trading feature is quite similar to a brokerage’s own conditional orders such as “grid trading.” So what are the differences between the two?

In response, the IT person told the Daily Economic News reporter that from a functional perspective, some AI Agent tools represented by “the lobster” have certain assistive capabilities in areas like information extraction, sentiment/public-opinion analysis, and generating simple strategies. Meanwhile, functions built up over the long term in broker apps—such as conditional orders and grid trading—focus more on implementing stable and controllable trade execution within a compliance framework. The fundamental difference between the two lies in this: the former leans toward “assisted decision-making and strategy generation,” emphasizing flexibility and personalization; the latter leans toward “trade execution and risk control,” emphasizing compliance and reliability. At the current stage, these two kinds of capabilities are more complementary than simple substitutes.

Under the premise of compliance, leading brokerages have already launched explorations internally for secure and controllable usage.

A relevant person from Guangfa Securities revealed to the Daily Economic News reporter that the company has started AI Agent application and technical exploration for OpenClaw, and has set up a technical research task force focusing on business scenarios such as intelligent office work, personal assistant services, and investment consulting/investment research tools. However, Guangfa Securities also emphasized the principle of “safety first and compliance for admission,” and conducts validation through methods such as prior reporting, independent segmented security sandboxes, and least-privilege access control.

As for the boundary of AI tool capabilities, Liu Xingliang has a clear understanding. He told the Daily Economic News reporter that he keeps four “lobsters” as digital employees, responsible respectively for collecting information, handling inquiries, financial reminders, and secretarial work. The results even exceeded expectations. But regarding stock trading, he remains restrained: “At this stage, I’m not comfortable letting the lobster handle fund transfers or payments. I’m only willing for it to do information consultation and monitoring the market as assistive work; the actual trading still needs to be done by me.”

Liu Xingliang believes that “lobsters” for stock trading are still in an unformed stage, so it is not yet suitable to let them independently carry out stock-trading related operations.

4

High barriers block most retail investors

Technical dividends are hard to share with ordinary people

Although “the lobster” is hot, not everyone can easily master it. In this wave of technology applications driven by large models, ordinary investors face very high hidden technical barriers.

A brokerage IT person told the Daily Economic News reporter that the AI Agent tools appearing in the market recently essentially combine large-model capabilities with market data, strategy rules, and automated execution—lowering the barrier to constructing simple strategies. But they also warned that such tools are still in the exploration phase overall, and their strategy stability, data-source reliability, and risk-control capability still need further verification.

Jin Fengchun fully agreed with this. He told the Daily Economic News reporter that if ordinary people want to use “the lobster” to trade stocks, they must master knowledge related to AI and programming development skills—this is an extremely high skill barrier. Therefore, he does not recommend that ordinary investors who don’t understand AI and programming blindly follow the trend, because doing so involves a high investment risk.

Jin Fengchun also reminded investors: “The high returns related to ‘the lobster’ cannot be guaranteed to be replicable, and you can’t judge whether a strategy is effective based solely on short-term high returns. Investors should pay attention to long-term overall performance and drawdown conditions. Investment should pursue long-term, steady returns, instead of being lured by short-term high returns into a speculative mindset.”

Regarding market concerns about whether AI tools will replace analysts or brokerage investment consultants, Jin Fengchun gave a negative answer. He analyzed for the Daily Economic News reporter that AI tools are essentially only auxiliary tools for investment research and trading, and the industry’s core demand still exists. In addition, ordinary retail investors lack professional investment knowledge, so they can’t independently use AI tools like “the lobster” to formulate effective strategies, nor can they distinguish between market-effective information and ineffective noise.

Jin Fengchun believes: “Using AI tools has a high barrier. Most investors don’t have the operational capability—this determines that the value of professional investment consultants and analysts cannot be replaced. The industry will only replace the tool and the way it delivers services.” He disclosed that because “the lobster” has only been out for a short time, there probably aren’t many regular institutions directly using it for live trading. But last year, the phenomenon of institutions using AI to build strategies already existed, and their teams are also planning to conduct some live-trading attempts after A-share simulated trading competitions.

In response to excited ordinary investors who are eager to try, multiple professionals provided advice. Liu Xingliang suggested that “lobster” stock trading needs to be “raised” for a while first, then slowly debug it in line with its capabilities, and never rush to let it participate fully in live trading. He urged investors to try with small capital, manage risks properly, and not put in large amounts of money; and he reiterated that the core role of “the lobster” is as a reference and assistive aid for investment decisions—it cannot replace humans’ final decision-making.

The aforementioned brokerage IT person also emphasized to the Daily Economic News reporter the bottom line for risk prevention: “For individual investors, these tools can be used as information assistance and for research reference, but they should not be overly relied on, and they must not replace basic risk identification and investment judgment. During use, especially, you need to pay attention to the authenticity of data, the effectiveness of strategies, and potential risks of excessive trading, so as to avoid amplifying investment volatility due to overtrust in technical capabilities.”

Planning| Xiao Yong Du Wei

Reporter| Chen Chen

Editor| Yi Qijiang

Vision| Shuai Lingxi

Layout| Yi Qijiang

**    **

**｜The Daily Economic News  nbdnews Original article｜    **

**Reprinting, excerpting, copying, mirroring, or other uses are prohibited without permission    **

The Daily Economic News