Exclusive Interview with Bill, Head of Bitget AI: In the Age of AI Trading, How Far Are We from “Passive Profits”?
TechFlow Selected TechFlow Selected
Exclusive Interview with Bill, Head of Bitget AI: In the Age of AI Trading, How Far Are We from “Passive Profits”?
The long-term goal is to build a sustainable account operating system that enables cross-category trading and a global ecosystem.
Navigating Web3 tides with focused insightsAuthor: Frank, PANews
A “crayfish” has shaken up the entire tech world. The sudden emergence of OpenClaw thrilled everyone—on an ordinary personal computer, AI can now be granted operational permissions to help you check emails, write code, and even manage trading accounts. Online case studies abound, describing it in almost mystical terms: “You won’t even need to work anymore.” Yet once users actually install it, most quickly realize things aren’t quite that simple.
This shift from hype to sober reflection is especially pronounced in crypto trading. Over the past two years, nearly every exchange has launched its own “AI Agent,” but most remain stuck at the level of chat-based assistance—you ask a question, and it replies with a lengthy analysis, nothing more. OpenClaw’s arrival was like opening Pandora’s box, revealing AI’s potential not just to “speak,” but to “act.”
Yet this very capability triggers new challenges. Dr. Bill, Head of Bitget AI and a pioneer leading his team to explore the frontier of AI-powered trading, offers deep insights on this point. PANews conducted an in-depth interview with Dr. Bill on the subject. Prior to joining Bitget, Dr. Bill held senior roles across multiple top-tier internet and technology firms, led large-scale deployments of core algorithms and AI platforms, and published dozens of papers at top international conferences and secured dozens of patents.
Today, as the executive fully responsible for Bitget’s AI strategy and intelligent trading R&D, he is driving deeper integration between AI and crypto asset trading scenarios. Facing the current Agent craze, this industry leader maintains remarkable composure: “Most ordinary people aren’t accustomed to being managers. Suddenly assigning them ten AI subordinates raises complex questions—how to direct, delegate, and evaluate performance. That itself is an art.”
Enthusiasm inevitably fades—but capability has already been demonstrated. The real question now becomes: Who can package this capability into products usable by ordinary people?
In conversation with Bill, PANews sought to deconstruct—from a product designer’s perspective—the genuine path from AI trading concepts to real-world implementation. In Bill’s view, Bitget’s recent launches of Agent Hub and GetClaw weren’t merely reactive moves (“Everyone else is doing it, so we should too”) but rather a natural externalization of internally developed products. “In short, it’s about timing, location, and alignment.”
Timing refers to OpenClaw’s market impact, which ignited widespread awareness; location reflects our solid foundation built over the past year through iterative enhancements to our AI assistant GetAgent, coupled with deep internal technical exploration and experimentation; alignment signifies that the team had already validated the product’s value internally—and thus naturally extended it outward.
Bitget’s AI Product Landscape: A Three-Tier Architecture from GetAgent to GetClaw
To understand Bitget’s AI trading strategy, one must first clarify how its three products relate. Externally, names like GetAgent, Agent Hub, and GetClaw may seem confusing—but according to Bill, they represent a clear evolutionary trajectory.
In June 2025, Bitget launched GetAgent within its app—a chatbot-style AI trading assistant. As Bill explains, GetAgent underwent multiple iterations: starting from basic conversational responses, then adding one-click order placement and news aggregation, and eventually expanding to cover full-category trading—including U.S. equities, gold, and silver. “Each iteration was driven by user demand—expanding incrementally, organically.” Yet no matter how far it expanded, GetAgent remained fundamentally “chat-driven”: it could answer questions and offer suggestions, but could not autonomously execute complex trading tasks on users’ behalf.
The turning point came after OpenClaw’s release. According to Bill, Bitget rapidly built its own internal version following OpenClaw’s launch. “Feedback from internal use was overwhelmingly positive—prompting the natural idea: Could we upgrade GetAgent significantly?” Following this logic, Bitget packaged and opened to the public its internally refined MCP capabilities, officially launching Agent Hub on February 13 this year.
Agent Hub targets users with relatively strong hands-on abilities—professional players.
It provides four progressively deeper capability interfaces:
API represents atomic-level interface calls—the highest barrier, requiring programming skills and API key management;
MCP serves as a “universal interface,” enabling external AI applications to directly read Bitget’s data and execute operations;
CLI caters to developers, supporting full API invocation via terminal command line;
Skills form the core of this upgrade—pre-packaged “business modules.” Through Skills, raw, rigid API code is transformed into functions directly callable by AI (e.g., fee inquiry, K-line analysis, price monitoring, order placement), enabling AI to bridge the gap from “intent understanding” to “action execution.”
Bill draws an intuitive analogy using a USB flash drive: “A USB drive inherently possesses storage capabilities—read, write, and save—but to function, it requires a USB port to connect to devices—that’s the MCP. But having just the port isn’t enough; you also need memory chips and compatible protocols to complete full interaction. This entire integrated system constitutes a Skill.”
Still, Agent Hub retains barriers for ordinary users.
Thus, on March 14, Bitget launched GetClaw—a Telegram-based AI trading assistant, ready-to-use out-of-the-box, requiring no installation. Users simply click a link, log in to their account, and begin using it immediately. Platform bears the cost of large-model invocation—users experience zero friction. Bill sums it up concisely: “Ordinary users are recommended to use GetClaw—a fully assembled tool, ready to play right away; professional players are recommended to use Agent Hub, selecting suitable Skills to build their own ‘castle,’ Lego-style.”
These three products form a clear progression: GetAgent refined foundational MCP capabilities, which were then externalized via Agent Hub, and finally embedded into GetClaw to lower usage barriers to the absolute minimum. From chatbot to developer tool to one-click product, Bitget’s AI product line spans the full spectrum—from tech-savvy geeks to total newcomers.
“Say One Sentence and Start Monitoring”—What AI Trading Truly Changes
Product architecture is merely the skeleton—the real excitement lies in AI’s transformative impact on concrete use cases. Throughout conversations with Bill, one recurring keyword is “barrier.”
Traditional trading involves a long chain: information gathering, analysis & decision-making, order execution, price monitoring, and post-trade review—each step reliant on manual input. For conditional trading or quantitative strategies, users either write programs to call APIs themselves—or configure numerous complex parameters on-platform.
In Bill’s view, this is precisely where AI delivers its greatest value: “These features can be implemented without Skills or GetClaw—just write code. But coding is easy for programmers, while prohibitively difficult for ordinary users. What we’re doing today is letting users achieve identical outcomes by speaking just one sentence.”
He cites a concrete example: When a user says, “If Bitcoin drops 3% within one minute, add 50% to my position,” the underlying system automatically converts this into a scheduled task involving three steps:
- Real-time monitoring of Bitcoin’s price
- Calculating minute-by-minute price changes
- Executing the position-add immediately upon condition fulfillment
Logic previously achievable only by programmers can now be triggered by anyone uttering a single sentence.
Within less than 40 hours of GetClaw’s launch, price-monitoring alerts became the most explosively adopted use case. This is unsurprising: configuring such alerts on traditional platforms demands understanding of various indicators and parameters—often taking half an hour or more, with no guarantee of success. Now, even multi-indicator composite monitoring logic (e.g., MACD + CCI) can be achieved simply by describing the requirement in natural language.
But Bill believes AI-powered monitoring’s true revolution goes beyond mere feasibility—to continuous optimization. “On traditional platforms, if configuration fails, users typically abandon it. Now, you can tell the AI, ‘That’s wrong—reflect and revise,’ iterating until satisfied.” This capacity for ongoing refinement offers immense satisfaction to the vast long-tail user base.
In traditional equity markets, algorithmic trading’s share continues rising—exceeding 70% even in mature markets like the U.S. Retail investors entering such environments face institutional opponents operating at microsecond speeds, rendering victory virtually impossible. Bill frames AI trading’s significance as “empowerment”: “Bitget’s vision in AI is to empower 100 million users to match Wall Street.” In other words, equip them with elite traders’ decision logic and execution capabilities. In the past, users could conceive ideas but lacked means to implement them; today, if you can imagine it—you can do it.
Four Locks of Trust: Security Boundaries When AI Handles Real Money
As AI evolves from “giving advice” to “executing on your behalf,” functional power ceases to be the biggest challenge—trust becomes paramount. Bill stresses this point emphatically: “The top concern for ordinary users is ‘Is it safe to use?’ Trust must be established firmly. One or two security incidents would be enough to lose all users.”
Centered on this core concern, Bitget designed a four-layer isolation system:
- First layer: Identity isolation—precise user identification per session
- Second layer: Memory isolation—complete separation of conversation histories between users, ensuring personal privacy remains protected
- Third layer: Permission control—access to specific data or tools governed strictly by role definitions
- Fourth layer: Credential and fund isolation—API keys restricted solely to triggering purposes; trades executed exclusively within sandboxed sub-accounts
The sub-account sandbox mechanism is a pragmatic design. Bill illustrates: “For instance, if your main account holds $1,000, you may allocate only $50 to a sub-account for AI operations—making risk far more controllable.” This ensures that even if AI makes erroneous judgments, exposure remains strictly confined within users’ pre-set limits.
This security-first mindset also informs Bitget’s approach to the Skills marketplace. Currently, all Skills are developed and maintained exclusively by Bitget—not open to third parties. Bill’s explanation is direct: “Opening a Skill Marketplace to broader participation inevitably introduces security risks. Imagine a hacker submitting malicious code—users deploying it suffer financial losses. That’s unacceptable. We adhere to the principle of ‘better none than flawed’—preferring no Skills over risking users losing money. After all, in asset markets, longevity trumps speed of profit.”
OpenClaw’s cautionary tale validates this prudence. Its near-unrestricted operation on personal computers, while thrilling, spawned an absurd new industry—“cleanly uninstalling crayfish” itself became a profitable business.
At the large-model invocation layer, Bitget initially opted to absorb platform costs rather than require users to self-configure tokens—motivated both by security concerns and technical reasons. “Our Skills and MCP have undergone deep adaptation and optimization with multiple built-in large models. Arbitrary switching to alternative models by users would severely degrade performance.” Currently, the platform provides each user with a daily $10 free invocation quota, with pricing models to be adjusted based on market feedback.
AI Can Handle 80% of Tasks—but Humans Must Make the Critical 20% of Decisions
When discussing realistic boundaries of AI trading, Bill candidly acknowledges limitations: “Right now, some online users give AI $100 and expect it to grow to $1,000—yet find such unstructured approaches carry extremely high probability of total loss.”
AI trading today cannot guarantee profits. Bill applies the “80/20 rule” to describe the current reality: Within the full trading workflow (potentially involving 100 discrete tasks), AI can efficiently handle ~80% of labor-intensive work—information aggregation, real-time monitoring, conditional execution, and data review. However, the remaining 20% of core decisions determining profitability remain beyond AI’s reach.
Last year, Bitget hosted an experimental AI trader competition to test AI’s capability boundaries—yielding vivid evidence: Many AI strategies ended in losses. The reason is straightforward—AI lacks emotions. Though seemingly advantageous, this also means AI cannot respond to black-swan events like sudden wars. Bill notes that when U.S. equity markets previously relied heavily on AI execution, abnormal phenomena like extreme intraday spikes and crashes occurred.
“Today, it primarily serves as advanced assistance—akin to the transitional phase from Level 1 to Level 5 autonomous driving,” Bill analogizes to characterize AI trading’s current stage. While AI’s capabilities are indeed steadily conquering remaining hurdles, machines still face clear bottlenecks in long-term creativity and empathetic judgment under extreme circumstances.
Nonetheless, Bill offers a relatively optimistic outlook: “The technological closed loop for fully automated trading may be largely realized next year—but that doesn’t guarantee sustained profitability.” In other words, there remains a substantial gap between ‘can run’ and ‘can profit.’”
From Trading Tool to “AI Account Operating System”: Bitget’s Ultimate Vision
Given AI’s inability to fully replace human traders in the near term, what is Bitget’s endgame for its AI strategy? Bill answers across three dimensions.
The first dimension is “comprehensive trading,” echoing Bitget’s earlier UEX (Universal Exchange) strategy. It extends beyond cryptocurrencies—driven by tokenization trends, traditional financial assets like gold, silver, and U.S. equities are increasingly integrated. Bitget aims to empower users to conduct cross-asset-class trading on a single platform—“equipping users with Wall Street traders’ full-asset-class coverage capabilities.”
The second dimension is global ecosystem expansion. Leveraging Bitget Wallet’s capabilities, AI will be introduced into Web3 payments and global commerce scenarios—lowering operational barriers for cross-border transactions and payments.
The third dimension—described by Bill with the most vivid imagery—is building a “long-term account operating system” anchored on Bitget. Its core is establishing a “high-trust fund execution layer,” where multiple Agents collaborate to handle diverse user needs—underpinned by a cross-device, cross-scenario “long-term memory system.”
As Bill describes it, this memory system analyzes and integrates users’ historical trading habits, past actions, and even granular behavioral traces within the app—forming a deep personal profile. “Ensuring users maintain consistent trading logic across different platforms and scenarios—not fragmented, disjointed experiences.” This capacity for continuous learning and adaptive tuning fundamentally distinguishes it from one-off tools.
He uses a relatable analogy to illustrate this progressive trust-building process: “Just as initially you might only trust a home robot to sweep floors—and only gradually entrust it with more responsibilities as confidence grows.” AI must first prove reliability in small tasks—gradually earning greater authority and trust—ultimately aiming to ‘grow alongside you, accompanying your asset appreciation.’”
From GetAgent to Agent Hub to GetClaw, Bitget’s AI products completed a leap—from chatbot to task-execution layer—in under a year. The intensive rollout by major exchanges confirms AI trading is no longer optional—it’s becoming a fundamental competitive capability.
Yet realistically, AI excels more at replacing the “manual labor” than the “mental labor” of trading. 80% of tedious work can be delegated to machines—but those critical 20% of decisions determining profit or loss still require human judgment. Technology can lower trading barriers—but cannot eliminate trading risk entirely.
AI has handed everyone Wall Street’s toolbox—but inside lie both opportunity and reverence.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
PANews
