Author: Frank, PANews
A "little lobster" stirred the entire tech circle. The emergence of OpenClaw has everyone excited, as AI can be given operational permissions on a regular personal computer to help you check emails, write code, and even operate trading accounts. There are countless cases online describing it in a mystical way: "You won't even need to work anymore." But when people actually installed it, most found that it wasn't the case.
In the realm of cryptocurrency trading, the temperature difference from enthusiasm to calm is particularly evident. Over the past two years, almost every exchange has launched its own "AI Agent," but most have only stayed at the chat assistance stage, where you ask it a question, and it writes you a lengthy analysis—that's all. The appearance of OpenClaw seemed to open Pandora's box, showing everyone the possibility of AI "doing things" rather than just "talking."
However, this has sparked new challenges. As a leading figure in exploring the forefront of AI trading, Dr. Bill, head of Bitget AI, has profound insights on this. PANews conducted an in-depth interview with Bill. Before joining Bitget, Bill held senior positions in several leading internet and technology companies, led the scaling of multiple core algorithms and AI platforms, and published dozens of papers at international conferences along with several patents.
Now, fully responsible for Bitget AI's strategic planning and intelligent trading technology development, he is dedicated to promoting the deep integration of AI with cryptocurrency trading scenarios. Faced with the current Agent craze, this leading expert's judgment is extremely calm: "Most ordinary people are not accustomed to being managers. Suddenly giving them 10 AI subordinates, how to command, divide tasks, and assess performance is an art in itself."
Passion will eventually fade, but capabilities have been recognized. The real question becomes: who can package these capabilities into products usable by ordinary people?
In the dialogue with Bill, PANews attempted to dismantle the real path from concept to implementation of AI trading from the perspective of product designers. In Bill's view, the intensive launch of the Agent Hub and GetClaw AI products by Bitget is not merely about "doing it because others are doing it," but rather a natural overflow of an internal product process. "In summary, it is about timing, geographical advantages, and teamwork."
Timing refers to OpenClaw igniting market awareness; geographical advantage refers to the solid accumulation from the continuous iteration of our AI assistant GetAgent launched last year; and teamwork is about the internal team having already validated the product's value, opening it to the outside world accordingly.
The panoramic view of Bitget's AI products: The three-tier structure from GetAgent to GetClaw
To understand Bitget's layout in AI trading, one must first clarify the relationship between its three products. To outsiders, names like GetAgent, Agent Hub, and GetClaw may be confusing, but in Bill's description, it's actually a clear path of evolution.
In June 2025, Bitget launched GetAgent within the app, a chatbot-style AI trading assistant. According to Bill, GetAgent has gone through multiple iterations: from the initial chat responses, gradually adding one-click ordering, news aggregation, to expanding into comprehensive trading of US stocks, gold, silver, and other categories, "each iteration is driven by user needs, expanding more and more." But no matter how it expands, the essence of GetAgent remains "chat-driven"; it can answer questions and provide suggestions, but it cannot help users independently execute complex trading tasks.
The turning point occurred after OpenClaw was released. According to Bill, after the release of OpenClaw, Bitget quickly built its internal version, "After internal use, the feedback was very positive, leading to the idea: can we give GetAgent a major upgrade?" Following this line of thought, Bitget encapsulated its internally refined MCP capabilities for external release, officially launching the Agent Hub on February 13 of this year.
Agent Hub targets "professionals with relatively strong hands-on abilities."
It offers four tiers of capabilities from simple to complex:
API is an atomic-level interface call, with the highest threshold, requiring programming and key management;
MCP acts as a "universal interface," allowing external AI applications to directly read Bitget's data and perform operations;
CLI is aimed at developers, supporting direct API calls via terminal command lines;
Skills are the core of this upgrade, equivalent to well-packaged "business modules." Through Skills, the originally rigid API code is transformed into skills (such as querying rates, analyzing candlesticks, monitoring, placing orders) that AI can directly call, allowing AI to leap from "intent understanding" to "action execution."
Bill made an intuitive analogy with a USB drive: "The USB drive itself possesses storage skills of saving, reading, and writing, but to make it work, a USB interface is needed to connect to a device, which is equivalent to MCP. However, merely having an interface is not enough; it requires cooperation with storage and various protocols to complete a full interaction; this entire combination constitutes a Skill."
However, Agent Hub still has thresholds for ordinary users.
So, on March 14, Bitget launched GetClaw, an AI trading assistant based on Telegram, plug-and-play, requiring no installation. Users can access it through a link, log in, and use it, with the platform bearing the cost of calls to large models, completely unobtrusive for users. Bill summarized this in one sentence: "Ordinary users are recommended to use GetClaw, which is a fully assembled tool that can be used immediately; professional players are recommended to use Agent Hub to choose suitable Skills, building their castles like playing with Legos."
These three products form a clear incremental relationship: GetAgent polished the underlying MCP capabilities, solidified and opened them up in Agent Hub, and embedded these capabilities into GetClaw, reducing the minimum usage threshold. From a chatbot to developer tools and then to a one-click product, Bitget's AI product line covers the entire user spectrum from geeks to novices.
"Watching with just a sentence," what has AI trading truly changed?
Product architecture is merely the skeleton; what truly excites users is the experience revolution brought by AI in specific scenarios. In communication with Bill, a recurring keyword is "threshold."
The traditional trading process is a long chain: acquiring information, analyzing decisions, executing orders, monitoring, and reviewing—each step relies on manual operations. If users want to conduct conditional trades or quantitative strategies, they either have to write scripts to adjust APIs themselves or configure a bunch of complex parameters on the platform.
In Bill's view, this is precisely where AI's greatest value lies: "These functions can be realized without Skills or GetClaw; you can just write a program. But the problem is, writing a program is simple for programmers, but the threshold is too high for ordinary users. Today, what we are doing is enabling users to achieve the same effect by just saying a sentence."
He provided a concrete example: A user might say, "When Bitcoin drops by 3% within a minute, help me increase my position by 50%," and the system would automatically convert this into a timed task. This task essentially needs to accomplish three things:
Real-time monitoring of Bitcoin prices
Calculating price differences every minute
Immediately executing the position increase once conditions are met
This type of logic, which could previously only be achieved by programmers, can now be completed simply by anyone saying a sentence.
Within less than 40 hours after GetClaw went live, the monitoring reminder became the most explosive use case. This is not surprising; on traditional platforms, configuring monitoring alerts requires users to understand various indicator parameters, "configuring for half a day may not even succeed." Now, even for complex monitoring logic involving multiple indicators like MACD and CCI, users can describe their needs in natural language, and the system can help achieve it.
However, Bill believes that the real transformation of AI monitoring is not just about "being able to do it," but more about "being able to tune it." "On traditional platforms, if you couldn't configure it well, you would just give up, but now you can tell it, 'This is wrong; reflect on how to improve it,' and keep adjusting until satisfied." This continuous iterative interaction is a huge satisfaction for a vast long-tail user base.
In the traditional stock market, the proportion of quantitative trading is increasing, and it can even exceed 70% in the relatively mature US market. Ordinary retail investors face institution competitors operating at microsecond levels, making their chances nearly non-existent. Bill summarizing the significance of AI trading as a form of "equalization": "Bitget's vision in AI is to enable 100 million users to compete with Wall Street," in other words, allowing them to attain the operational logic and execution capabilities of top traders. In the past, achieving this was a thought but not a reality; today, as long as you can think of it, you can achieve it.
The four locks of trust, the safety boundaries of AI operations with real money
When AI moves from "giving suggestions" to "executing for you," the power of the function is not the biggest challenge; trust is. In Bill's view, this cannot be overstated: "Ordinary users' biggest concern is 'Is it safe to use?' This level of trust must be well established. Once a couple of security issues arise, no one will use it anymore."
Around this core concern, Bitget designed a four-layer isolation system.
The first layer is identity isolation, accurately identifying the user's identity in each conversation
The second layer is memory isolation, with the dialogue memories of different users completely isolated to ensure personal privacy is not leaked
The third layer is permission control, determining what data and tools can be called upon based on roles
The fourth layer is credentials and fund isolation, where API keys are limited to trigger use, and transactions are executed in sub-account sandboxes
The sub-account sandbox mechanism is a very pragmatic design. Bill gives an example: "For instance, if the main account has 1,000 dollars, the user can only transfer 50 dollars to the sub-account for AI operations, which makes the risk much more controllable." This means that even if AI makes a judgment error, the risk exposure is strictly controlled within the user's pre-set limits.
This safety-first thinking is also reflected in Bitget's attitude towards the Skills marketplace. Currently, all Skills are developed and maintained by the official team, and not opened to third parties. Bill's explanation for this is very straightforward: "If we open the Skill Market to allow more people to participate in development, it inevitably leads to security issues. For example, a hacker could say, 'I will put one in for you,' and if users experience financial losses from using it, that would be inappropriate. We prefer to have fewer but quality Skills; it is better to have none than to risk losing all the money."
After all, in the asset market, earning quickly is not as important as living long.
The caution demonstrated by OpenClaw serves as a reasonable reminder. It operates on personal computers in almost unrestricted ways, which, while exciting, also fostered an absurd new industry, with "helping you uninstall the lobster" itself becoming a lucrative business.
On the level of large model calls, Bitget initially chose to bear the costs rather than having users configure tokens themselves. On one hand, this is for safety reasons, and on the other hand, it is a technical choice, "Our Skills and MCP have been deeply optimized for adaptation with various built-in large models. If users switch to other models at will, the effectiveness will drop significantly." Currently, the platform provides each user with a daily free allowance of 10 dollars for calls, with future pricing models to be adjusted based on market feedback.
80% of tasks can be done, but 20% of decisions still rely on humans
When discussing the realistic capability boundaries of AI trading, Bill candidly stated that the reality is not optimistic: "Now there are people online giving AI 100 dollars to make 1,000 dollars, but it turns out that such rough operations have a very high probability of loss."
The ability of AI trading today cannot guarantee profits for users. Bill uses the "80/20 rule" to summarize the current state: In a complete trading process (which may involve 100 tasks), AI can efficiently complete 80 of the tedious jobs, such as information organization, real-time monitoring, condition execution, and reviewing data. However, the 20 core decisions that truly determine profit and loss cannot yet be made by AI.
Last year, Bitget held a playful AI trader competition to test the boundaries of AI capabilities, resulting in a vivid note: many AI strategies ultimately ended in loss. The reason is not complicated; AI lacks emotions, which sounds like an advantage, but it also means it cannot respond to extreme black swan events like "sudden wars." Bill mentioned that in the past, there were anomalies when AI was extensively used to execute trades in the US stock market, leading to erratic spikes and drops.
"Today, AI plays more of a high-level assistive role, much like the transition from Level 1 to Level 5 in autonomous driving." Bill uses this analogy to position the current stage of AI trading development. From a trend perspective, AI's capabilities are indeed conquering remaining challenges one by one, but regarding long-term creativity and empathetic judgment in extreme scenarios, machines still have significant limitations.
However, Bill also provided a relatively optimistic judgment: "A closed technical loop for fully automated trading may basically be realized next year, but this does not mean it can guarantee continuous profitability." In other words, "the ability to run" and "the ability to earn" are still separated by a considerable distance.
From trading tools to an "AI account operation system," Bitget's ultimate vision
Since AI cannot completely replace human traders in the short term, where does Bitget's AI strategy ultimately lead? Bill provided answers from three dimensions.
The first dimension is "panoramic trading," which echoes Bitget's previously proposed UEX (Universal Exchange) strategy. It is not limited to cryptocurrencies; with the advancement of asset tokenization, traditional financial categories such as gold, silver, and US stocks are also being integrated. Bitget hopes to help users complete all-category trading operations on one platform, "allowing users to have the all-category coverage capabilities of Wall Street traders."
The second dimension is global ecological expansion. Combining the capabilities of Bitget Wallet, the aim is to introduce AI in Web3 payments and global business scenarios to reduce the operational thresholds for cross-border trading and payments.
The third dimension, which Bill described most vividly, is to build a "long-term account operation system" based on Bitget. The core of this concept is to establish a "high-trust fund execution layer," where multiple Agents will collaboratively assist users in various tasks, supported by a cross-end, cross-scenario "long-term memory system."
In Bill's description, this memory system will analyze and integrate users' past trading habits, historical operations, and even every little action within the app, forming a deep personal profile. "Ensuring users' trading logic remains consistently aligned across different platforms and scenarios, rather than experiencing fragmented, disjointed interactions." This ability for continuous learning and adaptation is fundamentally what sets it apart from one-time tools.
He used a very everyday analogy to explain this gradual trust process: "It’s like when someone first buys a house-cleaning robot merely to let it vacuum; over time, as trust builds, they become willing to let it handle more tasks." AI needs to prove its reliability in small tasks first, then gradually gain larger permissions and trust, with the ultimate goal being to "grow with you, accompanying your asset appreciation."
From GetAgent to Agent Hub to GetClaw, Bitget's AI products have made the leap from chatbot to task execution level in less than a year. The intensive layout of major exchanges also indicates that AI trading is no longer an optional direction, but rather a core competency for future competition.
However, from the current reality, AI is better at replacing the "manual labor" in trading than the "intellectual labor." 80% of the tedious tasks can be handed off to machines, but the core judgment that determines the profit and loss of that remaining 20% will still require human involvement. Technology can lower the thresholds for trading but cannot completely eliminate the risks involved.
AI has given everyone access to Wall Street's toolbox, but what’s in the toolbox are both opportunities and respect.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
