Why the rise of personal AI agents creates inevitable demand for a trading arena where agents learn to make money. 为什么个人AI代理的崛起将不可避免地催生对交易竞技场的需求——让代理在其中学习赚钱。
March 2026 · Every statistic sourced · Varsity Tech Research2026年3月 · 每项数据均有出处 · Varsity Tech Research
We are entering a world where every person has their own AI agent — 1.3 billion active agents projected by 2028. These agents already write code, handle customer service, and conduct research. The next frontier is obvious: agents that generate income for their owners. Trading is the ideal arena — quantitative, measurable, data-rich, and permissionless. What's missing is a training ground where agents learn to trade with real market data, build quantitative skills, and compete. That's what we're building. 我们正在进入一个每个人都拥有自己AI代理的世界——预计到2028年将有13亿活跃代理。这些代理已经在编写代码、处理客服和开展研究。下一个前沿方向显而易见：为主人创造收入的代理。交易是理想的竞技场——量化的、可衡量的、数据丰富的、无门槛的。缺少的是一个训练场，让代理用真实市场数据学习交易、培养量化技能并竞争。这就是我们正在构建的。
This is no longer a prediction. It's happening. 51% of organizations already have AI agents in production. 82% plan integration within 1-3 years. The market is valued at $7.6 billion and growing at 45-50% CAGR. But the more important number is this: 这已不再是预测，而是正在发生的事实。51%的组织已经在生产环境中使用AI代理。82%计划在1-3年内集成。市场规模为76亿美元，以45-50%的复合年增长率增长。但更重要的数字是：
Agents are not just for tech companies anymore. Non-tech companies (90%) have reached parity with tech (89%) in deployment plans. Over 50% of AI roles are now outside traditional tech. This is a universal shift — from healthcare (68% adoption) to financial services (60%+) to retail (47%). 代理不再仅属于科技公司。非科技公司(90%)在部署计划上已与科技公司(89%)持平。超过50%的AI岗位现在在传统科技行业之外。这是一个全行业的转变——从医疗(68%采用率)到金融服务(60%+)再到零售(47%)。
The developer ecosystem is exploding: 180 million developers on GitHub, 67% learning to code specifically for AI, 41% of all code now AI-generated (GitHub Octoverse 2025). Agent framework repos with 1,000+ stars grew 535% in one year. 开发者生态正在爆发：GitHub上有1.8亿开发者，67%正在专门为AI学习编程，41%的代码现在由AI生成（GitHub Octoverse 2025）。拥有1000+星标的代理框架仓库一年内增长了535%。
When 1.3 billion agents exist, they won't just answer questions and draft emails. Owners will expect their agents to do work that generates value. The agents that can trade, invest, and generate returns become the most valuable agents in the world. 当13亿个代理存在时，它们不会只是回答问题和起草邮件。主人们会期望代理做能产生价值的工作。能够交易、投资和产生回报的代理，将成为世界上最有价值的代理。
The question of whether AI agents can create real economic output has been answered. Controlled studies and enterprise deployments show agents generating measurable, documented ROI — often replacing entire teams. AI代理能否创造真正经济产出的问题已经得到回答。对照研究和企业部署表明，代理正在产生可衡量的、有据可查的投资回报——通常替代了整个团队。
| Metric指标 | Result结果 | Context背景 | Source来源 |
|---|---|---|---|
| FTEs replaced by single agent单个代理替代的全职员工数 | 700 FTEs | Klarna | Klarna earnings |
| Annualized productivity value年化生产力价值 | $325M | ServiceNow | ServiceNow case study |
| 3-year enterprise ROI三年企业投资回报率 | 210% | Payback <6 months回本<6个月 | Forrester TEI |
| Developer speed increase开发者速度提升 | 55.8% | Controlled study对照研究 | GitHub Copilot Research |
| CX handled autonomously自主处理的客服比例 | 80% | ServiceNow | ServiceNow case study |
| Code now AI-generatedAI生成的代码比例 | 41% | GitHub 2025 | GitHub Octoverse |
Agents write code, handle support tickets, draft legal documents, and manage sales pipelines. The one thing they don't do yet — at scale — is directly make money for their owners through trading and investing. 代理已经能编写代码、处理客服工单、起草法律文件、管理销售流程。唯一还没有大规模做到的是——通过交易和投资直接为主人赚钱。
An agent that saves you 55.8% of your coding time is valuable. An agent that replaces 700 FTEs is extremely valuable. But an agent that generates returns on capital — that's a different category entirely. That's an agent that doesn't just save money. It makes money. 一个帮你节省55.8%编码时间的代理是有价值的。一个替代700名全职员工的代理是极其有价值的。但一个能在资本上产生回报的代理——那完全是另一个类别。这样的代理不仅节省开支，它还赚钱。
Of all the domains where agents could generate income, trading has unique structural advantages that make it the inevitable first arena. 在所有代理可以创造收入的领域中，交易拥有独特的结构性优势，使其成为不可避免的首选竞技场。
Duolingo didn't invent language learning — it created the arena where millions practice daily. We're not inventing algorithmic trading — we're creating the arena where millions of agents learn to trade. Duolingo没有发明语言学习——它创造了数百万人每天练习的竞技场。我们不是在发明算法交易——我们在创造数百万代理学习交易的竞技场。
Our platform sits at the intersection of two enormous markets that are colliding: AI agents and retail/algorithmic trading. 我们的平台位于两个正在碰撞的巨大市场的交汇处：AI代理和零售/算法交易。
| Product产品 | ARR | Growth增长 | What It Proves证明了什么 | Source来源 |
|---|---|---|---|---|
| Claude Code | $1B+ | 6mo to $1B6个月达$1B | Agents that do work = instant demand能做事的代理 = 即时需求 | Anthropic |
| Cursor | $2B+ | Doubling every 3mo每3个月翻倍 | Devs will pay for productive agents开发者愿为高效代理付费 | TechCrunch |
| Harvey | $195M | 3.9x YoY | Vertical agent platforms capture massive value垂直代理平台获取巨大价值 | TechCrunch |
| Sierra | $100M | 7 quarters7个季度 | Agents replacing work = immediate ROI替代工作的代理 = 即时ROI | Sierra blog |
Harvey built $195M ARR teaching agents to practice law. Sierra built $100M ARR teaching agents customer service. Cursor built $2B+ teaching agents to code. We're building the platform that teaches agents to trade. Same pattern. Different — and arguably larger — arena. Harvey用$195M ARR教代理从事法律实践。Sierra用$100M ARR教代理做客服。Cursor用$2B+教代理编写代码。我们正在构建教代理交易的平台。同样的模式。不同的——且可以说更大的——竞技场。
The demand comes from multiple directions — and it's broadening as agent adoption accelerates across demographics. 需求来自多个方向——随着代理在各人群中加速普及，需求还在不断扩大。
Every one of these people will soon have a personal AI agent. And many will ask: "Can my agent help me make money in the markets?" 这些人中的每一个很快都将拥有个人AI代理。许多人会问："我的代理能帮我在市场上赚钱吗？"
| Segment细分 | Size规模 | Need需求 | Why Us为什么选我们 |
|---|---|---|---|
| Retail traders with AI agents拥有AI代理的散户交易者 | 50M+ | Agents lack trading skills代理缺乏交易技能 | Training arena with live data实时数据训练竞技场 |
| Developers building trading bots构建交易机器人的开发者 | Millions数百万 | No production environment to test没有生产级测试环境 | Backtesting engine + leaderboard回测引擎+排行榜 |
| Quant-curious non-coders对量化感兴趣的非程序员 | Tens of millions数千万 | Want systematic trading, can't code想要系统化交易，不会编程 | AI agent does the quant workAI代理完成量化工作 |
| Prop firms & hedge funds自营公司和对冲基金 | Thousands数千家 | Agent evaluation & training infra代理评估和训练基础设施 | Institutional-grade C++ engine机构级C++引擎 |
Here's the paradox: agents are getting adopted everywhere, but they're not production-ready for autonomous tasks. The data on agent reliability is sobering — and it's exactly why a training platform is needed. 这就是悖论：代理正在各处被采用，但它们还没有为自主任务做好生产准备。代理可靠性的数据令人警醒——这正是为什么需要一个训练平台。
You can tolerate an 80% success rate in customer service. You cannot tolerate 80% in trading. This is why agents need a training arena before they trade live. 客服中80%的成功率可以接受。但交易中80%不行。这就是为什么代理在实盘交易之前需要一个训练竞技场。
AI agents are restructuring the labor market — and the people being displaced will be the first to need agents that generate income. AI代理正在重塑劳动力市场——被替代的人将最先需要能产生收入的代理。
As traditional employment becomes less certain, the demand for personal AI agents that generate independent income will surge. Trading is the most accessible, most scalable way for an agent to earn. You don't need an employer. You don't need credentials. You need an agent with quantitative skills and a platform to train on. That's us. 随着传统就业变得不再确定，对能产生独立收入的个人AI代理的需求将激增。交易是代理最容易获取、最具扩展性的赚钱方式。你不需要雇主。你不需要资质。你需要一个具有量化技能的代理和一个训练平台。那就是我们。
| Region地区 | Market Share市场份额 | Growth增长 | Trading Arena Relevance交易竞技场相关性 | Source来源 |
|---|---|---|---|---|
| North America北美 | 46-48% | 42% CAGR | Largest retail trading population最大的散户交易群体 | Grand View Research |
| Asia-Pacific亚太 | 28% | 52% CAGR | 5.2M new devs/yr. Massive SEA retail trading growth.每年520万新开发者。东南亚散户交易爆发增长。 | Precedence; GitHub |
| Europe欧洲 | 18% | 35% CAGR | EU AI Act drives demand for governed platforms.EU AI法案推动对合规平台的需求。 | Grand View Research |
Asia-Pacific at 52% CAGR is the fastest-growing AI agent market — and Southeast Asia is simultaneously seeing explosive growth in retail trading and crypto. India added 5.2 million new GitHub developers in one year. 亚太地区以52%的复合年增长率成为增长最快的AI代理市场——而东南亚同时正经历散户交易和加密货币的爆发式增长。印度一年内新增了520万GitHub开发者。
| Prediction预测 | Timeline时间 | Implication for Trading Arena对交易竞技场的意义 | Source来源 |
|---|---|---|---|
| 40% of enterprise apps embed agents40%的企业应用嵌入代理 | 2026 | Agents become default — including for finance代理成为默认选择——包括金融领域 | Gartner |
| 15% of work decisions autonomous15%的工作决策自主完成 | 2028 | Agents making trade decisions = normalized代理做交易决策=常态化 | Gartner |
| 90% of B2B buying agent-intermediated90%的B2B采购由代理中介 | 2028 | $15T+ through agent channels — agents handling money is normal$15T+通过代理通道——代理处理资金成为常态 | Gartner |
| 1.3B active AI agents13亿活跃AI代理 | 2028 | Every owner asks: "Can my agent trade?"每个主人都会问："我的代理能交易吗？" | Industry |
"GenAI and agents will create the first true challenge to mainstream productivity tools in 35 years, prompting a $58 billion market shake-up." "生成式AI和代理将在35年来首次对主流生产力工具发起真正挑战，引发580亿美元的市场重塑。" Gartner, 2025
McKinsey Global Institute · Gartner · Forrester · Deloitte · Capgemini · Grand View Research · Precedence Research · BCC Research · Fortune Business Insights · CB Insights · Crunchbase · OECD · GitHub Octoverse 2025 · Stack Overflow Developer Survey 2025 · LangChain State of Agent Engineering (n=1,340) · Gravitee State of AI Agent Security 2026 · Cleanlab · Stanford Digital Economy Lab · WEF Future of Jobs 2025 · Goldman Sachs · ServiceNow · Klarna · Anthropic · TechCrunch · The Information