一只为 AI 时代设计的科技对冲基金：用人的前瞻抓住历史上没出现过的新物种，用量化纪律驾驭 AI 的强周期，用智能体投研团队跟上每周都在变化的产业。 A technology hedge fund built for the AI era: human foresight to catch what has never existed before, quantitative discipline to ride AI's violent cycles, and an AI-agent research team to keep pace with an industry that changes every week.
五个支点，共同构成 xResearch 的投资方式。Five pillars define how xResearch invests.
AI 周期的巨大摆幅，加上 AI 加速科技进步带来的新物种。The outsized swings of the AI cycle, and the new species born from AI-accelerated discovery.
没发生过的事，模型学不出来；强周期，靠人扛不住。两者缺一不可。Models can't learn the unprecedented; humans alone can't ride brutal cycles. We need both.
智能体负责覆盖、执行和复核；人只做四件事：给直觉、挑错、拍板、判决。Agents handle coverage, execution and review. Humans do four things: supply intuition, find flaws, decide, adjudicate.
看增长引擎、不看公司；按阶段定买点和卖点，解决“买早了”。Analyze growth engines, not companies; set buy and sell points by stage, so we don't buy too early.
仓位规则事前写下、AI 监督执行；资金托管、独立核算、年度审计。Sizing rules written in advance and AI-supervised; segregated custody, independent NAV, annual audit.
技术进步是指数级的，人的认知是线性的——认知差不断放大，市场在过度乐观和过度悲观之间大幅摆动。Technology advances exponentially; human perception improves linearly. The gap keeps widening, and the market swings hard between over-optimism and over-pessimism.
蒸汽机放大体力，电力放大能量，互联网放大信息传输——而 AI 复制的是研究员本身。工具让人做得更快；复制科学家，让“发现”本身变快。The steam engine amplified muscle, electricity amplified energy, the internet amplified information. AI replicates the researcher. Tools make people faster; replicating scientists makes discovery itself faster.
统计：费城半导体指数（SOX）2000-01-04 至 2026-08-26 收盘价，双向 20% 划分；平均跌幅不含进行中的一轮。历史统计只来自有限样本，不代表未来表现。Source: Philadelphia Semiconductor Index (SOX) closes, 2000-01-04 to 2026-08-26, ±20% threshold; average drawdown excludes the round in progress. Historical statistics come from a limited sample and do not predict future results.
这三道难题，靠多雇几个分析师解决不了——我们用规则加智能体来解。Hiring a few more analysts won't solve them. We solve them with rules plus AI agents.
上万家公司，一个人看不过来。Tens of thousands of companies, more than any person can watch.
新机会，往往在你的认知之外。New opportunities usually sit outside what you already know.
概念出来就买，结果不一定好。Buying the moment a concept appears often ends badly.
每一种走法都提前写好预案：再回撤，按预设档位分批加仓；信号累加，有仓位不踏空。Every path has a playbook written in advance: on renewed drawdowns, add in tranches at preset bands; as signals accumulate, stay positioned.
核心命题按三组证据季度打分，分数决定基础仓位档——信念有多强，看证据说话。Core theses are scored quarterly on three groups of evidence; the score sets the base position. Conviction must show receipts.
顶部／底部信号序列＋市场状态，在基础档上加减——资产分数相同，周期位置不同，仓位不同。Top/bottom signal sequences and market state adjust the base. Same score, different cycle position, different exposure.
市场对好消息反应变钝＝哨兵亮灯，仓位上限自动压低——防止在别人不买账时独自重仓。When good news stops working, the sentinel lights and the cap drops: no heavy positioning while the market isn't buying.
极端防御状态下净敞口可降至零附近。方向性策略，非市场中性：净敞口就是表达周期观点的仪表。Net exposure can fall to near zero in extreme defensive regimes. Directional, not market-neutral: net exposure is the dial that expresses our cycle view.
所有让人觉得繁琐的合规流程，背后都是对资金安全的保护。Every compliance step that feels tedious is part of what keeps the money safe.
基金业绩、条款与完整介绍材料，仅向经确认的专业／合格投资者提供。欢迎来信。Performance, terms and full fund materials are available only to verified professional / accredited investors. We welcome your inquiry.