BlackRock's Goldstein Says AI Scale Moats Protect Core Tech From SaaS Shift

贝莱德Goldstein：AI规模护城河保护核心技术免受SaaS冲击 像Aladdin这样的企业级软件工具在AI代码扩张下将会得到强化而非瓦解，因为监管、合规和权限控制层无法被轻量级模型轻易复制。 贝莱德首席运营官Rob Goldstein剖析了塑造现代资产管理格局的关键大趋势，重点探讨了自公司创立以来技术和并行计算如何推动贝莱德的发展。他阐述了公司在受监管环境中部署非确定性AI时所遵循的严格“初稿原则”，并解释了AI智能体如何拓展Aladdin等核心平台的能力。针对有关软件颠覆、token消耗成本以及公开与私募市场界限日益模糊的担忧，Goldstein进行了回应，并强调未来的投资优势将依赖于全投资组合解决方案、创造性的提示词工程以及一线直接的人类智慧。 贝莱德从小规模公司跃升为全球最大资产管理机构的巨大增长，反映了现代金融的关键大趋势：买方崛起、技术融合、私募市场扩张以及幂律分布。 (0:00) 贝莱德最初是建立在信息处理理论之上的，当时通过联网的Sun Microsystems工作站替代昂贵的卖方超级计算机，以进行固定收益风险建模。 (5:00) 与传统的确定性代码不同，现代生成式AI是非确定性的，运作机制更贴近人类思维，这促使贝莱德在组织管控中引入严格的“初稿原则”。 (10:00) 生成式AI通过加速软件工程以及利用自然语言智能体释放平台潜能，在Aladdin等复杂系统中创造企业价值。 (15:00) “SaaS大灾难”（SaaSpocalypse）的论调夸大了核心企业系统面临的风险，因为受监管的零容忍环境需要深度的权限分配、控制层和复杂工作流，这是轻量级AI工具无法轻易复制的。 (25:00)

BlackRock Chief Operating Officer Rob Goldstein argues that scale and technology are converging to reshape finance, positioning comprehensive enterprise platforms and holistic asset managers to outlast lightweight software disruption. Technology as BlackRock's Core Operating Thesis (5:00) BlackRock's expansion from $19 billion in assets under management and roughly 80 employees in 1994 into the world's largest asset manager was built on treating asset management as an information processing business. In the late 1980s, the firm's founders addressed an analytical asymmetry between sell-side banks using supercomputers and buy-side firms receiving faxed yield tables by chaining together ten $10,000 Sun Microsystems workstations to run complex risk models at a fraction of standard costs. Enterprise AI Adoption and the First Draft Principle (10:00) Goldstein contrasts traditional deterministic software with generative AI models, which operate non-deterministically by translating raw compute and energy into language-based outputs. Because institutional finance maintains near-zero tolerance for automated errors, BlackRock enforces a firmwide 'First Draft Principle,' mandating that AI-generated presentations, code, and documents undergo human verification—sometimes through dozens of internal reviewers—before formal implementation. Aladdin's Moat Against the 'SaaSpocalypse' Amid market concerns that generative coding will make traditional SaaS platforms obsolete, Goldstein argues that enterprise infrastructure like BlackRock's Aladdin system will become more entrenched rather than displaced. Aladdin's moat relies on regulatory governance, proprietary data, and deeply embedded permissioning frameworks that lightweight, bespoke code cannot easily replicate. While software layers that merely summarize public data face severe displacement, core institutional platforms benefit as clients use natural language agents and APIs to interact programmatically within secured control planes. Private Asset Transparency and Holistic Balance Sheets (40:00) Private markets are undergoing a multi-year transition toward transparency and standardization, paralleling the historical digitization of public fixed-income securities. Goldstein reframes the private asset illiquidity premium as an 'effort premium' that diminishes as tech tools standardize workflows. Consequently, the historical boundaries separating asset classes are blurring, shifting competitive advantage toward multi-asset managers capable of running an institution's entire balance sheet across both public and private exposures under unified risk models. Shifting Investment Edge Toward Qualitative Intelligence (45:00) As AI models accelerate software development cycles from months to days and rapidly assimilate public text, traditional quantitative coding offers less differentiation. Goldstein highlights the emerging value of articulate communicators and English majors who can effectively steer AI tools, alongside qualitative human networks that gather localized, non-digitized intelligence on the ground. Co-hosts Joe Weisenthal and Tracy Alloway observe that the market pendulum is swinging back toward off-model research and investigative field work that algorithms cannot scrape. Enterprise-wide AI adoption across finance remains in its opening phase, constrained by surging compute expenses, unoptimized token consumption, and the friction of organizational redesign. Whether technological standardization will fully eliminate private market yield premiums without eroding asset-class distinctions remains an open question.

贝莱德首席运营官罗伯·戈尔茨坦指出，资产管理本质上是信息处理业务，生成式AI的技术爆发不仅没有削弱以Aladdin为代表的底层金融基础设施，反而强化了合规控制平面与深度工作流壁垒。随着自动化编程门槛降低以及公开信息被大模型迅速消化，未来的核心投资优势正转向跨资产的全组合管理能力，以及算法尚无法触及的线下实地洞察。 创办初期的计算基因与信息处理本质 (5:00) 戈尔茨坦回顾了贝莱德在20世纪80年代末及1990年代的创立历程。贝莱德从1994年的80名员工与190亿美元资产管理规模起步，如今已成长为全球最大的资产管理机构。创立初期，创始人发现卖方银行利用昂贵的超级计算机构建复杂的抵押贷款支持证券，却仅通过低效的传真收益率表向买方推介。贝莱德通过将十台单价1万美元的Sun微系统工作站串联，以低成本复制了数百万美元级超级计算机的计算能力，自建风险模型向资产所有者提供前所未有的透明度。 戈尔茨坦在1994年以初级数据分析师身份入职时，数据分析远非流行领域，团队甚至需要共用计算机。这一经历确立了贝莱德的核心认知：资产管理根本上是一个信息处理行业。当下个人通过连接多台Mac Mini在本地独立运行大语言模型的做法，正映衬了当年利用分布式硬件突破算力瓶颈的技术周期循环。 生成式AI的非确定性与初稿原则 (10:00) 戈尔茨坦将当前的人工智能周期与以往技术变革对比，指出AI将原始能源转化为智能，属于高资本密集型趋势。不同于输出结果始终一致的传统确定性软件，前沿大语言模型具有非确定性，行为模式更贴近能够自行编写代码并排查漏洞的人类心智，然而社会对计算机出错的容忍度远低于对人类错误的容忍。在受到严格监管的金融服务领域，必须设立严密的组织流程以控制此类不可预测性风险。 为此，贝莱德在全公司推行“初稿原则”：允许并鼓励由AI生成客户演示文稿、内部文档或招股说明书初稿，但明确要求每一份初稿必须经过多道严格的人工审核关卡。戈尔茨坦认为，尽管个人AI效率工具层出不穷，但整个金融行业在深度业务流程再造与组织架构重组方面甚至尚未进入“第一局”，全面落地仍需要漫长的重构周期。 核心工作流护城河与轻度SaaS危机 针对AI自动化编程或将摧毁传统软件即服务（SaaS）的担忧，戈尔茨坦给出了不同判断。贝莱德科技投资经理托尼·金预计，若以当前全球存量代码规模为100行基准，到2030年代码总量将呈指数级扩张至100万行。然而，代码的普及非但不会颠覆像Aladdin或彭博终端这样的核心平台，反而会强化其地位。这类平台深嵌于零容忍的合规环境中，掌握着高度敏感的专有数据，其核心价值在于复杂的权限管理与治理控制平面，能够精确管理成千上万员工对投资组合查看、交易执行与核对的精细授权。 与此相对，那些仅仅抓取公开数据进行清洗排版的“便利型SaaS”将面临严重的淘汰危机。现代大语言模型能够直接从公开来源检索并输出定制答案，缺乏专有数据壁垒与深度工作流整合的轻量级工具将难以维系原有商业模式，唯有能够对接核心底层API并承担严密合规审查的平台才能在代码爆炸的生态中持续受益。 软件工程提速与代币算力治理挑战 在具体企业应用中，生成式AI显著提升了软件工程效率。在贝莱德推进私募与公募市场统一透明度的一项跨部门项目中，团队通过端到端结合AI工具，在规格研讨会后自动将录音转录为规范文档，随后输入内部AI编程工具自动生成可用代码并完成调试。原本需要耗时数月的复杂功能原型开发被压缩至数天完成，且生成的是具备深度查询能力的实际运行系统。 然而研发提速也带来了严峻的算力成本约束。戈尔茨坦引用早期合伙人查理·哈莱克的观点指出，若缺乏约束，模型师与工程师对算力的渴求足以让任何公司破产。目前贝莱德内部代币消耗量正呈数倍增长。在缺乏物理机房采购周期自然约束的云端时代，金融业亟需类似“代币版Aladdin”的效率管理工具来防止算力失控。此外，贝莱德AI实验室负责人斯蒂芬·博伊德提醒，大模型高度流畅的语言容易让人产生盲目信任，这也对输出审计提出了更高要求。 投资优势重构与线下未数字化洞察 (45:00) 戈尔茨坦与主持团队探讨了AI普及后未来投资管理人的核心竞争优势。戈尔茨坦认为，过去将固定收益、股票与私募信贷割裂在不同孤岛的模式正在被打破，未来的超额优势属于能够提供跨资产“全投资组合解决方案”的管理者。同时，私募市场的“非流动性溢价”本质上是“工作量溢价”，正如早年公共债券从手工录入招股书条款走向标准化一样，私募市场也在被技术逐步标准化并融入统一组合视角。 在技术落地摩擦大幅降低的背景下，戈尔茨坦提出看重“英语专业”人才，因其具备清晰表达能力与想象力，能更精准地引导和调用AI系统。主持人乔·韦森塔尔与特蕾西·阿拉韦在总结中进一步指出，当公开数字化文本被算法瞬时消化后，纯粹的量化与编程技能壁垒正在降低，真正的超额收益正重新向线下实地调研、人际网络与未数字化的现场信息回归，传统的人际交往与实地勘察能力再度展现出不可替代的价值。 戈尔茨坦指出，尽管技术正在模糊公募与私募界限并重构企业的日常运营，但企业级AI的成熟仍面临算力开支失控、模型非确定性以及组织再造迟缓等实际挑战。正如他在谈及与贝莱德首席执行官拉里·芬克的分歧时所言，远期技术愿景往往看似近在眼前，但真正完成底层系统搭建与合规流程改造仍需数年时间的稳步推进；在大模型平摊了代码编写与公共信息处理成本的未来，金融业的核心壁垒最终仍将落在深层工作流、权限治理控制体系以及线下未被模型消化的真实商业洞察之上。

## Background 本期节目来自 Bloomberg 旗下知名播客 Odd Lots，发布于 2026 年 4 月 30 日。节目简介显示，对话围绕数十年来重塑金融行业的几大超级趋势展开，重点聚焦贝莱德（BlackRock）首席运营官 Rob Goldstein 的深度访谈，探讨贝莱德早期的技术基因、核心风控系统 Aladdin 的演进历程、AI 时代下的 SaaS 危机、算力与 Token 消耗瓶颈，以及私募市场的未来走向。 ## TL;DR 贝莱德 COO Rob Goldstein 在节目中回顾了贝莱德从 10 台 Sun 工作站起步、将资产管理视为“信息处理业务”的创业历程，并剖析了核心风控平台 Aladdin 在受监管环境下的生态护城河。他指出，随着大语言模型和 AI Agent 的崛起，缺乏私有数据壁垒的“便利层”SaaS 面临存亡危机，而企业软件的核心价值将转变为由 Agent 驱动并受 API 控制平面监管的深度工作流。此外，访谈深入探讨了私募市场透明度不可逆的提升趋势以及公募与私募界限的光谱化演进，并断言在数字信息被模型高度定价的未来，能够提供“全资产组合解决方案”、具备丰富想象力以及拥有深厚真实世界本土人际调研网络的从业者，将掌握最稀缺的超额收益（Edge）。 ## Atmosphere & style 整场对话在轻松风趣又不失专业严谨的氛围中推进，主持人 Tracy Alloway 和 Joe Weisenthal 与受访嘉宾谈笑风生，穿插着错叫“黑石”以及关于招聘英语系毕业生的行业趣闻。Goldstein 的讲述兼具资深华尔街老将的沉稳务实与硅谷科技布道者的敏锐前瞻，既有对“外星科技”式 AI 浪潮的惊叹，又对落地过程中的合规摩擦与组织再造保持极度清醒的克制。听感上如同一场顶尖金融大脑与硬核技术思想的碰撞，让人深刻感受到技术正在以不可逆的加速度重构整个全球资本运转的底层逻辑。 ## Bulletin - `30:00` 贝莱德（BlackRock）COO Rob Goldstein 指出，AI 的底层逻辑是将能源转化为智能；在企业软件领域，仅靠整理公开信息的“便利层”SaaS 正面临灭顶之灾，但拥有专有数据、深嵌零容错合规业务流的中心化平台（如 Aladdin）护城河反而更加坚固。 - `12:55` 贝莱德内部确立了“初稿原则”（First Draft Principle），鼓励员工使用 AI 起草备忘录与招募说明书第一版，并依托金融机构严密的 16 道人工复核机制把控合规；Goldstein 认为当前企业级 AI 部署仍处于“棒球赛奏国歌”的极早期阶段。 - `37:00` 针对全行业面临的计算瓶颈，已故联合创始人查理·哈拉克的历史警示再次显现：不加约束的算力胃口足以拖垮企业；贝莱德自身的 Token 消耗同比成倍翻番，但全行业尚未真正启动对 Token 消耗与推理成本的精细化优化治理。 - `45:00` 当线上数字化信息被大模型全盘吸收并极致定价后，未来投资机构的核心超额收益（Edge）将回归至三大支柱：全资产组合方案、想象力与精准表达能力，以及深扎物理世界的本土人际网络和现场第一手实地调研。 ## Key takeaways - `04:00` 金融巨头不断壮大的底层催化剂并非单纯的资本积累，而是技术驱动的高效价值主张；贝莱德早在创办初期就将资产管理定义为“信息处理业务”。 - `05:15` 贝莱德的技术底色始于用 10 台单价 1 万美元的 Sun 工作站串联网络，以颠覆性成本打破卖方投行对数百万美元超级计算机的垄断，为最终资产所有者提供前所未有的风险透明度。 - `10:00` 与传统软件的确定性不同，生成式 AI 具有非确定性与自主探索纠错的类人思维特征，其本质是重资本与重能源驱动的“外星科技”。 - `12:55` 金融行业的严格合规流程是吸收不可预测 AI 的独特优势，贝莱德通过“初稿原则”让 AI 负责生成草稿、人类团队负责多重审查，实现兼顾风控与个人生产力释放。 - `15:20` 传统 AI 核心在于处理结构化数字，现代 AI 则攻克了人类语言交互；AI 辅助编程正让贝莱德庞大的工程师团队以爆发式的速度交付 Aladdin 系统长期积压的功能需求。 - `17:40` AI 智能体（Agent）将成为企业软件的“超级操作员”，通过自然语言直接调度后台模块，彻底唤醒企业级软件中常年被用户忽视和沉睡的大量高阶功能。 - `24:50` 尽管未来代码总量将以十倍级爆炸增长，但处于生态中心、受严格合规监管且拥有私密资产数据的底层平台不会被“氛围编程”轻易颠覆。 - `27:15` “Open Aladdin”通过在封闭生态内开放 API，构建了一套在毫秒级别精细穿透交易权限、组合可见性的控制平面（Control Plane），构成了高度不可替代的技术壁垒。 - `30:00` 只作为公开数据搬运与界面包装的“便利层”SaaS 工具正面临灭顶之灾，但由于企业系统退役与迁移工程极其繁重，旧系统的淘汰将伴随较长的长尾效应。 - `32:00` 生成式 AI 将软件开发周期进行了不可思议的压缩：数小时跨职能头脑风暴会议录音可直接被转化为功能规格文档，并由 AI 编程工具迅速生成扎实、可运行的代码原型。 - `40:00` AI 技术在企业内的落地遵循三阶段演进：从追求基础模型智能，到探索企业级具体用例，最终必将进入对调用成本与计算吞吐的极致效率优化阶段。 - `43:20` 私募市场的“流动性溢价”本质上很大程度是处理非标准化信息的“辛劳溢价”（Effort Premium），随着技术推进，私募资产走向标准化与高透明度的趋势不可逆转。 - `45:00` 随着代码编写门槛趋近于零，工程人才的溢价正从抽象数学与编码技巧转向具备构想概念的想象力，以及能用精准语言向 AI 传达指令的表达素养。 - `51:30` 在算力能够自底向上精确建模单项资产的支撑下，公募与私募之间僵化的界限正在消失，转变为涵盖流动性、披露频率和托管形态的连续“光谱”。 - `01:00:00` 当线上一切数字化文本都能被 AI 迅速消化并体现在价格中时，离开屏幕走入物理世界的非标准化现场调研（Channel Checks）重获极高的超额收益（Alpha）溢价。 ## Detailed walkthrough ### `00:00` 金融界的四大超级趋势与 AI 浪潮的席卷 节目开场，主持人 Tracy Allaway 和 Joe Weisenthal 探讨了近几十年来深刻重塑金融格局的数个“超级趋势”（megatrends）。Tracy 首先提出了前两个核心动力：其一是**买方的崛起**（the rise of the buy side）——在传统格局中，华尔街投行与商业银行等卖方机构占据绝对核心与公众注意力，但随着时间推移，资产管理行业的规模呈爆发式扩张；其二是**技术的大规模应用**，涵盖电子交易、量化模型以及模型驱动的风险管理系统的全面普及；她随后补充了第三个关键趋势——**私募市场的蓬勃发展**。Joe 则提出了第四个同样重要的趋势：**幂律分布与超级巨头的赢家通吃**（winner-take-all / winner-take-most），即极少数巨头凭借规模优势变得越来越庞大。 这些长期趋势正与当下席卷全行业的 AI 浪潮剧烈交汇。眼下专业金融机构的每一个人都渴望以某种方式搭上 AI 的快车，甚至催生出某种近乎荒诞的紧迫感。Tracy 分享了一个近期见闻：一位在大型银行工作的朋友提到，公司管理层正执着地要求员工把 20 年前做过的所有历史工作翻出来填进巨大的 Excel 数据库中，理由是“马上就要把这些数据全部喂进 AI 模型”。Joe 调侃道，大家都在盲目地把散落在各处的杂乱数据强行塞进所谓的数据湖（data lake），以确保模型能够读取。这种行业现状引出了一个关键疑问：这究竟是管理层为了追逐风口而制造的虚假繁荣，还是 AI 真的能转化为金融行业中实打实提高生产力的有效工具？为了解答这些问题，本期节目请来了贝莱德（BlackRock）首席运营官 Rob Goldstein，他不仅亲历了上述所有巨变的发生，如今也正掌舵着全球最大资产管理机构的技术与运营体系。 --- ### `04:00` 规模扩张的本质：以技术驱动的价值主张 面对主持人提出的四大趋势以及“大机构越来越大”的幂律现象，Rob Goldstein 提出了一个微妙且关键的修正视角。他礼貌地指出，金融行业之所以出现巨头愈发庞大的现象，其底层催化剂并非单纯的资本积累，而是**那些能够提供更优质价值主张（value proposition）的机构在不断壮大**。虽然 Joe 随即指出这种更强的价值主张往往也是规模的副产品——因为只有足够庞大的机构才有财力提供全套的完整服务菜单——但 Rob 强调，如果对所有超级主题进行排序，其核心归宿其实全在于“技术”。正是技术赋能了前所未有的服务能力，使许多过去根本无法实现的价值主张成为现实。 在正式切入技术底色之前，Joe 开玩笑地提到了几周前与 Rob 进行预沟通时的尴尬趣事：Joe 当时口误将“贝莱德”（BlackRock）错叫成了“黑石”（Blackstone），电话那头随即陷入了一阵微妙的沉默，Joe 只好自嘲“我在这行干了太久了，请原谅我”。Rob 则幽默地回应，自己在贝莱德待了这么多年，早就对这种口误习以为常。 --- ### `05:15` 贝莱德的创办原点：十台 Sun 工作站与风险透明度 谈及贝莱德早期的技术基因，Joe 询问了流传已久的创业故事：贝莱德的底层文化是否与 Sun 工作站以及创始人之一 Ben Golub 紧密相连。Rob 证实了这一说法，并表示 Ben Golub 至今仍是他亲密的朋友与导师。他指出，贝莱德的诞生与发展路径，极其生动地映射了过去 30 到 35 年间整个金融商业世界最成功企业的发展轨迹。 贝莱德的几位创始合伙人最早是结构化产品与抵押贷款市场演进的先驱。在那个年代，华尔街卖方银行利用昂贵的大型超级计算机来进行极其复杂的证券结构化设计，但其向下游客户销售这批精密产品的方式却极度原始——“基本上就是把打印出的收益率表（yield tables）用传真机发往全世界”。当时的大多数从业人员桌上甚至没有个人电脑，往往是一个部门共用一台机器。面对这种信息与工具的严重不对称，贝莱德创办时的核心假说（founding thesis）是：**买方完全可以自己建立量化模型，为资产的最终所有者（end asset owner）提供风险透明度**。 ```mermaid flowchart TD accTitle: BlackRock Tech Inception accDescr: Comparison between Wall Street sell-side supercomputer model and BlackRock's distributed Sun workstation model. subgraph SellSide ["传统华尔街卖方模式"] A["数百万美元超级计算机"] --> B["设计复杂结构化产品"] B --> C["向全球传真收益率表<br/>(信息不透明/工具不对称)"] end subgraph BlackRockFounding ["贝莱德早期技术突破"] D["10 台 Sun 工作站<br/>(每台仅 10,000 美元)"] --> E["网络互联并行计算<br/>(匹敌昂贵超算算力)"] E --> F["构建内部风险评估模型"] F --> G["为最终资产所有者<br/>提供前所未有的风险透明度"] end ``` 这种雄心在当时受制于超算的惊人成本，但 Sun 工作站的横空出世带来了破局点。创始团队敏锐地发现，只要有足够的巧思，“你花每台 1 万美元买 10 台 Sun 工作站并将它们串联在一起，实际上就能做到过去必须斥资数百万美元的超级计算机才能做到的事情”。正是这种以技术为支点打破算力与信息垄断的想法，构成了贝莱德作为资产管理机构的立足之本。 - **1994 年的贝莱德规模**：全公司仅约 80 名员工，资产管理规模（AUM）约为 190 亿美元。 - **冷板凳上的“极客”**：Rob 于 1994 年作为数据分析师加入贝莱德的数据与分析团队。他回忆道，如今数据、技术和数据分析是人人都想挤进去的“酷孩子阵地”，但在当时，这完全是无人问津的冷门领域。 - **超前的业务本质**：贝莱德在三十多年前就敏锐地意识到“资产管理业务在核心本质上就是一项信息处理业务（information processing business）”。这在今天看似理所当然，但在当时是一个极度超前且新颖的洞察。 --- ### `07:30` 计算范式的循环：从工作站集群到本地 LLM 的思考 对于当年“串联 10 台 Sun 工作站构建超级计算能力”的创举，Joe 感慨科技史仿佛正在经历一场“重回未来”（Back to the Future）的轮回周期。Joe 分享了身边的例子：他的一个技术好友最近正打算帮他在家里用四五台 Mac mini 串联搭建一个本地集群，用来在个人电脑上私有化运行大语言模型（LLM）。这样不仅能彻底免去对外部科技巨头数据中心的依赖，还能获得无限量的 Token 调用额度。 Rob 对这一技术轮回的类比深感共鸣。他认为这精准切中了当下 AI 发展中一个最核心的未解问题：计算架构究竟会走向何方？当前业界正面临着两种力量的剧烈博弈——一种正是 Joe 所描述的算力小型化、分布式化与本地化回归；而另一种则是算力极端集中于耗资百亿美元、消耗海量电力的大型集中式数据中心。金融机构究竟该如何在这两种范式之间权衡计算瓶颈与数据归属，构成了当今技术布局的最前沿议题。 ### `10:00` 能源换智能、非确定性与“外星科技”隐喻 谈及当下人工智能的本质，罗伯·高盛（Rob Goldstein）给出了一个极为底层的技术与经济学视角：AI 的核心运作逻辑在于“将能源转化为智能”（convert energy to intelligence），在能源上投入的资金越多，获得的智能水平就越高。与过去几十年里轻资产模式的技术浪潮不同，本轮 AI 浪潮极度依赖重资本开支与巨额基建投入。不过他也指出，业界目前仍存在激烈的路线分歧：这究竟会是一场专属于科技巨头的昂贵资本博弈，还是最终会演变成“五个人窝在车库里、靠几台 Mac mini 就能创造奇迹”的轻量化范式？当下尚无定论。 主持人乔·维森塔尔（Joe Weisenthal）随即抛出了金融领域的核心痛点：传统软件具有完全的确定性，一段无 bug 的代码运行一千次都会输出完全一致的结果；而现代 AI 模型不仅具有非确定性（non-deterministic）——输入相同的问题可能得到不同的答案，且常常无法清晰“解释自己的计算步骤”（show your work）。对于必须具备高度合规性、可解释性与问责机制的金融行业而言，这种不可预测与不可知性是否构成了巨大的焦虑来源？ 高盛对此直截了当地回应：“毫无疑问，绝对如此。”他打了一个生动的比方：多年来一直有人预测完全自动驾驶汽车即将普及，然而人类社会对“计算机犯错”的容忍度，与对“人类犯错”的容忍度存在天壤之别。在他看来，当前的生成式 AI 就像“掉落在地球上的外星科技，人类正在摸索如何去使用它”。例如在编写代码时，AI 会写出带有 bug 的代码，随后自己去发现并修复这些 bug。这彻底颠覆了人类过去几十年对计算机“非 0 即 1”的二元确定性认知——AI 展现出来的不是机械的指令执行，而是一种更接近人类思维方式的探索与试错过程。 - **资本与能源转换**：现代前沿大模型将算力与能源直接映射为认知智能，形成重资本驱动特征。 - **容错心理差异**：公众与监管对自动化算法错误的容忍阈值远低于人类操作失误。 - **思维模式重塑**：从传统的确定性规则系统，转向类似人类思考、具备自主试错修复能力的类人认知系统。 --- ### `12:55` 受监管行业的合规优势与“初稿原则” 前沿大模型提供商在接触金融等垂直行业时，总会挂在嘴边一个词——“受监管行业”（regulated industries）。高盛指出，虽然金融机构在流程、控制和合规上存在极多摩擦力，但从另一个角度来看，严苛的合规体系反而是金融机构应对不可预测 AI 的独特竞争优势，因为“流程控制原本就是金融企业肌理的一部分”。 为了在确保安全的前提下拥抱新技术，贝莱德内部确立了一项核心管理规范——“初稿原则”（first draft principle）。无论是客户路演 PPT、内部备忘录还是复杂的基金招募说明书（prospectus），贝莱德鼓励员工一律交由 AI 撰写第一版。之所以敢于大步推行这一原则，关键在于受监管流程本身的设计：后续会有 16 个人对初稿进行层层审核、校验与把关。这种在成熟风控框架内使用 AI 的做法，不仅大幅释放了个人层面的生产力，更成为了组织熟悉并消化新技术的强大催化剂。 然而，谈到企业级落地时，高盛给出了相对克制的评估。他用棒球赛作比喻：很多人以为大模型已经引发了产业海啸，但在企业级部署的漫长赛程中，“我们甚至连第一局都没开始，球场上大概还只是在奏国歌的阶段”。模型技术本身的突破，与真正将其嵌入复杂的企业组织架构之间存在巨大的滞后壁垒。企业级落地需要进行组织架构重组（organizational design）和业务流程再造（business process reengineering），这一过程极其艰难且漫长，真正的战役才刚刚拉开序幕。 ```mermaid flowchart LR accTitle: 贝莱德的初稿原则工作流 accDescr: 展示从AI生成第一稿到16道严格人工审核再到合规产出的受监管业务流程。 subgraph S1["内容生成阶段"] A["业务需求\n(路演文稿/备忘录/招募说明书)"] --> B["AI 辅助生成\n(初稿原则 First Draft)"] end subgraph S2["严格风控与合规流水线"] B --> C["16道人工复核网\n(合规/风控/法务/业务主管)"] C --> D["双人复核原则\n(Four-Eyes Principle)"] end subgraph S3["合规产出"] D --> E["企业级正式交付物"] end ``` --- ### `15:20` 拆解“黑箱”标签：从数字 AI 到语言 AI 与编程生产力爆发 针对主持人追问贝莱德核心资产与风险管理系统 Aladdin 长期被外界戏称为“黑箱”（black box）、以及引入不可预测的大模型是否会让系统更加模糊的问题，高盛幽默地反驳：“首先澄清一点，我们自己可从来没把它叫作黑箱。”他强调，AI 绝不是凭空出现的新鲜事物：麻省理工学院（MIT）的 AI 实验室早在 20 世纪 50 年代便已成立，而贝莱德自身的 AI 实验室（AI Lab）也在 2018 年就已设立并深度运转。 高盛用极简的视角厘清了 AI 演进的本质： 1. **传统 AI（Old AI）**：核心是处理**数字与结构化数据**（numbers），做定量建模与风险预测。 2. **现代 AI（New AI）**：核心是处理**人类语言与语义**（language）。 正因为人类绝大部分的协作与沟通都是依靠语言而非数字完成的，语言能力的突破带来了极其惊人的技术蔓延，但也伴生了诸多非预期的复杂结果。高盛更以现场录音的彭博（Bloomberg）终端打趣，指出无论是 Aladdin 还是彭博终端，此类顶级企业平台都面临着“干得越好，任务越多”（the reward for good work is more work）的甜蜜负担。 在贝莱德，技术团队规模庞大，常年维持着约 5,000 名工程师、数据分析师和量化建模师的规模。尽管技术投入持续加码，但每年 Aladdin 系统积压的待办开发清单（to-do list）不仅没有缩减，反而越来越长。高盛强调，当前 AI 在企业级层面最成熟、产生回报最直接的场景正是辅助编程（coding）。AI 编程工具的介入，让整个团队消化和交付这些庞大待办功能的迭代速度（velocity）呈现出爆发式的增长。 --- ### `17:40` 释放企业软件沉睡潜能：让 AI Agent 成为 Aladdin 的超级操作员 除了提升研发交付速度，高盛指出了企业级专家系统（enterprise expert systems）面临的另一大终极痛点：功能沉睡与认知断层。他分享了自己频繁经历的尴尬场景：在与机构客户开会时，客户常常抱怨“为什么 Aladdin 不能支持某某功能？”高盛往往暗自思忖这功能似乎早已有之，但为了稳妥起见，他会在会后致电贝莱德内部的技术专家核实。而专家的答复往往是：“Aladdin 七年前就已经支持这项功能了。” 高盛坦言，企业软件功能越来越庞大，指望用户时刻跟踪每一次版本迭代与功能更新是不现实的。“绝大多数人使用技术的唯一目的，就是借助它完成手头的工作，然后准时下班回家。”这就导致大量先进的企业级功能常年处于沉睡状态，用户并未真正榨取出软件的全部价值。 而大模型与智能体（AI Agent）的融合将彻底终结这一困境： - **交互范式革命**：用户未来无需掌握繁琐复杂的专业菜单与操作路径，只需用自然语言输入业务诉求。 - **Agent 充当超级终端操作员**：AI Agent 将作为 Aladdin 的“实时超级用户”，在后台自动调度系统底层的功能模块去执行复杂操作。 - **治理与红线保留**：传统金融严密的操作治理框架（如双人复核原则 / four-eyes principle）将全程保留，Agent 的每一步操作依然受控于严密的权限与审计追踪。 高盛总结道，借助 AI Agent 将企业专有系统未被发掘的潜能完全激活后，以 Aladdin 为代表的企业级核心软件未来能够为客户创造的实际商业价值，将远远超越以往任何历史时期。这不仅消除了“客户抱怨系统缺乏他们早已拥有的功能”这一最大痛点，更是企业软件在 AI 时代最具想象空间的演进路径。 ### `20:00` AI 时代的 UX 悖论与网站重构 当探讨软件交互的演进时，主持人提出了一个长期关注的问题：如果未来使用像 Aladdin 这样的软件系统的主体变成了 AI Agent（智能体）而非人类，这是否会彻底改变用户体验（UX）的设计思路？BlackRock（贝莱德）首席运营官 Rob Goldstein 对此表示，公司内部对此展开了大量辩论，虽然 Agent 崛起是必然趋势，但他坚信“依然会有人想要亲自动手操作”。正如在移动互联网时代一切都转移到了手机端，企业依然必须保留供用户访问的网页端一样，人机界面的直观性依然具有不可替代的价值。 随后 Goldstein 分享了他前一天亲眼目睹的一场令人震撼的技术演示：一家 AI 公司的工具直接抓取了贝莱德的一个公开网站，在无需贝莱德内部参与的情况下，仅仅经过计算机约十分钟的处理，AI 就对该网站进行了彻底的重构，生成了一个在可用性与人机友好度上“最优化”的全新版本。这一幕展现出了一种奇妙的逆向逻辑——AI 并没有淘汰人类界面，反而被用来消除复杂性、让系统对人类更加友好。Goldstein 感慨道，在未来的几个月或一到两年内，这类工具将重塑地球上的每一个网站，让它们变得更加易用，他还风趣地补上一句：“希望这带来的是对我们人类的便利，而不只是为了 Agent。” --- ### `24:50` 代码爆炸、SaaSpocalypse 与 Aladdin 的生态护城河 面对当前市场上关于“SaaSpocalypse”（SaaS 软件浩劫）的激烈争论——即在“氛围编程”（Vibe Coding）普及的时代，所有人似乎都能凭借 AI 工具随手写出一套属于自己的投资组合与风险管理系统，Aladdin 究竟如何守住自己的护城河？Goldstein 先从一段宏观维度的对话切入。一两年前，他曾询问贝莱德的一位资深科技投资组合经理 Tony Kim：“如果今天世界上有 100 行代码，到 2030 年会有多少？”Tony 的回答是：“一百万行。”面对 Goldstein 的难以置信，Tony 严肃地解释这是经过深思熟虑的复合增长测算（每年以 10 倍速度递增）。Goldstein 指出，无论每年的乘数是 3 倍、7 倍还是 14 倍，AI 工具必将推动全球代码总量呈爆炸式级数增长。 ```mermaid flowchart TD accTitle: Aladdin 生态与控制平面架构 accDescr: 展示外部用户与智能体如何通过 Aladdin API 穿透权限控制层访问核心高壁垒系统的流程图 subgraph Access["访问与接入层"] A["个人开发者 / 氛围编程"] B["AI 智能体 / 自定义脚本"] end subgraph ControlPlane["治理与控制平面 (Control Plane)"] C["Aladdin API 端点"] D["企业级权限校验<br/>(组合可见性 / 交易权限 / 审核隔离)"] end subgraph CoreEngine["底层高壁垒核心"] E["专有敏感数据与模型"] F["零容错风控与交易执行系统"] end A --> C B --> C C --> D D --> E D --> F ``` 然而，代码激增并不意味着企业级核心平台的瓦解。Goldstein 强调，像 Aladdin 和 Bloomberg（彭博）这样的平台，位于整个金融生态系统的最中央，具有极高的壁垒： - **严格监管与零容错：** 该生态系统处于严苛的合规监管之下，对任何操作失误或系统故障的容忍度为零（Zero tolerance for fault or error）。 - **极度敏感的专有数据：** 平台处理的不仅是海量专有数据，更是客户最核心、最私密的财务资产数据，这些数据根本不会也不可能向公共模型开放。 - **人、流程与技术的深度交织：** 真实的资产管理工作流高度特异化（Idiosyncratic），是人员经验、业务流程与底层技术的复合体。 因此，Goldstein 认为展望未来十年，此类中心化平台承担的任务不仅不会减少，反而会变得更多。未来被释放的生产力，是让普通用户借助个人编程工具与 Agent 去访问这些平台，但这些平台依然是不可或缺的底层枢纽。 --- ### `27:15` Open Aladdin 与 API 控制平面（Control Plane）的核心价值 针对此前外界关于系统是“黑盒子”的讨论，Goldstein 坦言，一二十年前许多企业级软件确实被设计成了封闭系统，早期的 Aladdin 也是如此。但早在本轮生成式 AI 浪潮爆发的十年前，贝莱德就前瞻性地启动了“Open Aladdin”（开放阿拉丁）战略变革。当时管理层就预见到，未来的技术交互方式必将分化：一部分人继续使用键盘与鼠标，而更多的新一代从业者——哪怕大学主修的是英语专业，往往也上过编程课——则会倾向于通过编写代码与系统交互。 但 Goldstein 明确澄清，“Open Aladdin”并不是指开源代码，而是指“在封闭生态系统内部实现开放”（Open within a closed ecosystem），即对外开放大量的 API 端点。这种模式在实践中展现出了一个巨大且极具价值的副产品：**API 级别的权限治理控制平面（Control Plane）**。 在一家拥有数千名员工的资产管理机构中，权限管理的复杂程度令人望而生畏： - 不同的经理只能查看各自被授权的特定资产组合； - 有些人员拥有下单交易权限，而另一些人员被严格限制为只能进行交易确认或对账审核。 当用户通过代码调用 Aladdin 的 API 时，系统内嵌的精细权限机制会毫秒不差地随行穿透，精确判定该请求“能看什么、能做什么”。Goldstein 总结道，整个行业越是热衷于自己写代码、用更技术化的手段与底层系统交互，这个能够确保数据安全、合规与风控边界的控制平面，其商业价值和不可替代性就越发凸显。 ### `30:00` SaaS 的“便利层”危机与软件长尾淘汰 如果说沉淀了核心专有数据、深嵌在受监管工作流中的底层基础设施（如 Aladdin）其价值在 AI 时代会越发凸显，那么硬币的反面又会如何？那些缺乏壁垒的软件是否会彻底归零？罗布·戈德斯坦（Rob Goldstein）直言不讳地给出了肯定回答。不过他也指出，软件的淘汰往往伴随着极长的长尾效应——在企业实际运转中，退役和更换任何一套既有系统的工程量都极其繁重，因此这类工具并不会在一夜之间凭空蒸发。 然而，真正的生存危机正笼罩在特定类型的 SaaS 产品头上。市场上存在大量本质上只是“收集整理公开信息并提供便捷访问界面”的软件。戈德斯坦强调，现代 AI 模型正是搜罗全网公开数据、并以用户最习惯的方式呈现信息的“终极神谕（ultimate oracles）”。这意味着仅仅作为“便利层（convenience layer）”存在的工具正面临灭顶之灾：它们既没有核心专有数据，又未深度融入核心业务流，仅仅提供了一层交互上的方便。在 AI 原生工具的冲击下，这些便利型科技公司正陷入极大的被动，不得不拼命寻找并重塑自身的价值主张。 - **基础设施 vs. 便利层**：核心受监管系统拥有高粘性专有数据；便利层 SaaS 仅搬运公共数据，最易被大语言模型直接取代。 - **系统的长尾惯性**：即便工具失去不可替代性，企业端退役旧系统的复杂流程仍会提供一定缓冲期。 - **价值重塑压力**：缺乏私有工作流阵地的软件公司若不能找到新的价值立足点，将面临被市场彻底清算的风险。 --- ### `32:00` 真实生产力跃迁：从会议录音到可用原型的周期压缩 许多企业高管在谈及 AI 带来的“生产力提升”时往往停留在泛泛的宏观概念，而贝莱德兼具 AI 提供商与使用者的双重身份。被问及内部真实的生产力跃迁究竟是何模样时，戈德斯坦分享了就在上周五发生的一个真实案例。贝莱德当前的一大战略重点，是通过增强 Aladdin 的透明度，让私募市场（Private Markets）能够像公开市场一样清晰透明，以推进其“全投资组合（Whole Portfolio）”理念。为此，内部多个团队长期推进着庞大的开发计划。 在周五下午的产品演示（Demo）中，团队展示了一款令人惊叹的新工具。但真正震撼戈德斯坦的是其背后的诞生过程：包括投资组合经理（PM）、风险控制专家、软件工程师和产品经理在内的一群核心人员坐在会议室里，进行了数小时密集的业务逻辑头脑风暴。这套系统端到端地将该会议全程录音，并由 AI 自动提炼生成功能规格文档（Functional Document）；经过人工轻微校验微调后，该文档被直接投喂给内部的 AI 编程工具；编程工具迅速生成了可运行的代码原型，并经历了现场即时调试。在资产管理行业摸爬滚打 32 年的戈德斯坦见过无数原型，深知大部分所谓的原型不过是一层“稍微用力按压就会碎掉的薄壳（thin shell）”，但眼前这个工具却具备实打实的完整深度。 ```mermaid flowchart TD accTitle: AI-Driven Software Prototyping Pipeline accDescr: End-to-end workflow showing how a BlackRock product demo was created from a multi-hour cross-functional meeting transcript to a working prototype. subgraph Ideation["需求与设计阶段"] A["多职能团队会议<br/>(PM/风控/工程师/产品经理)"] --> B["全程会议录音"] B --> C["AI 生成功能规格文档"] C --> D["人工轻微校验修正"] end subgraph Development["代码实现与交付"] D --> E["输入 AI 编程工具"] E --> F["AI 快速生成代码原型"] F --> G["即时调试与修复"] G --> H["可交互的扎实软件原型"] end ``` 以往需要跨部门反复沟通、撰写文档、排期编码数月之久的开发循环，被不可思议地压缩成了短短几天。虽然新功能在真正面向生产环境发布前，仍必须走完贝莱德严格的软件开发生命周期（SDLC）与各项严苛测试，但这印证了行业代码总量以十倍速（10xing）爆炸的趋势，工程实现的门槛正在发生结构性坍塌。 --- ### `35:20` 提示词权重猜想与白板多模态输入 面对这种“开会即生成软件”的新范式，主持人半开玩笑地提出一个敏锐的技术问题：当 AI 编程工具解析这类多角色会议纪要时，是否会根据参会者的头衔或职级（例如董事总经理 vs 普通员工）来加权分配意见？戈德斯坦笑着回应称，虽然理想情况下模型应当识别决策权重，但他直觉上认为目前的 AI 并没有层级偏见，反而呈现出一种朴素的机制——“谁说话最多，谁的想法被体现得最充分”。 随后，主持人分享了业界另一个正在落地的多模态实施范例：一家协助客户实施部署软件的第三方咨询公司，在与客户进行 Zoom 视频会议时，不仅录制语音，还会专门架设一台摄像机对准参会人员正在推演的实体白板。因为白板是“软件工程领域的通用语言（lingua franca）”，AI 能够同时处理音轨对话与白板涂鸦的视频流，完整消化团队讨论架构时的草图、箭头和逻辑框架。这种将现实工作流无缝转化为数字工程输入的进化速度，让与会者感叹其疯狂的演变步伐。 --- ### `37:00` 算力无底洞：从查理·哈拉克的警告到 Token 消耗激增 讨论随即转向全行业普遍面临的“Token 账单刺痛（sticker shock）”与推理算力瓶颈。针对贝莱德当前的 Token 消耗以及物理算力（GPU）限制是否构成制约的问题，戈德斯坦回忆起了已故的贝莱德联合创始人之一查理·哈拉克（Charlie Hallac）曾对他传授的深刻经验教训：“如果你招揽了最顶尖的模型构建师和工程师，却在算力上不对他们加以任何约束，他们永无止境的算力胃口足以让整家公司破产。” 在过去的自建物理数据中心时代，硬件订购周期与机房物理容量构成了天然的制约屏障；当员工抱怨机房不够用时，管理层往往会选择“等一年看看实际情况”。然而在云原生与弹性伸缩的今天，这种约束已荡然无存。戈德斯坦坦言，虽然手头没有贝莱德今年具体的 Token 消耗统计，但其规模相比去年已经实现了“数倍（multiples）”的激增。 更关键的是，戈德斯坦认为这场算力革命的大幕甚至根本还没有真正拉开。不论是贝莱德还是整个金融行业，不仅无法预判未来的终极 Token 消耗总量，而且更重要的是，“目前还没有任何人真正开始着手优化自己的 Token 消耗”。当主持人打趣称行业迫切需要“一款专门优化 Token 效率的 Aladdin”时，戈德斯坦断言这类优化工具必然会出现。他提到贝莱德 AI 实验室负责人、斯坦福大学工程学教授斯蒂芬·博伊德（Stephen Boyd）曾向他分享过早期洞察，强调人们忽视了两件事：其一，“精准流畅的语言（articulate language）极具欺骗性的说服力”——在学术界，一篇论文如果文笔晦涩，审稿人会逐字核查；但如果写得行云流水，人们往往会下意识假定其结论正确；其二，包括斯坦福在内的全球高校里，无数优秀的年轻学者正将精力投入到底层算力与算法的极速优化之中。 - **查理·哈拉克法则**：不对顶级工程师和量化建模师的算力消耗设限，其胃口足以拖垮企业现金流。 - **Token 爆炸前夜**：贝莱德的 Token 消耗同比成倍翻番，但行业层面的消耗优化尚未真正启动。 - **语言流畅度的认知盲区**：斯蒂芬·博伊德指出，AI 高超的语言表达能力容易让人放松对其严密性的审视，产生“表达得好即代表内容正确”的心理错觉。 ### `40:00` AI演进的三部曲：从追求智能、落地企业场景到极致效率 在谈及当下人工智能的推进节奏时，罗伯·戈尔茨坦（Rob Goldstein）指出，贝莱德当前正在处理这项技术中“最枯燥但最关键的环节”——如何大幅提升这些大模型的运行效率。在他看来，整个行业的技术演进有着非常清晰的节奏：最初大家都在盲目开展“对模型基础智能的探索”（quest for intelligence）；而当下，这一焦点正在逐步转向“对企业级落地用例的探索”（quest for enterprise use cases）；紧接着，行业很快就会全面转向“对资源调用与访问效率的极致追求”（quest for efficiency of how you're accessing things）。尽管目前全行业尚未完全抵达最后一个阶段，但计算成本与吞吐效率注定会成为下一个决定胜负的战场。 ```mermaid flowchart LR accTitle: AI技术演进三阶段 accDescr: 展示企业AI发展从智能探索、落地用例到追求效率的演进路径。 subgraph S1["第一阶段"] A["智能探索<br/>基础模型能力突破"] end subgraph S2["第二阶段"] B["用例探索<br/>企业级落地与业务整合"] end subgraph S3["第三阶段"] C["效率探索<br/>计算成本与调用优化"] end A --> B B --> C ``` --- ### `43:20` 私募市场的“流动性溢价”还是“辛劳溢价”：透明度不可逆转的浪潮 主持人将话题引回私募市场（Private Markets），探讨当通过类似阿拉丁（Aladdin）这样的技术系统，投资者能够像管理公开市场资产一样管理私募资产、并享有更高的定价透明度时，私募资产赖以维系的核心卖点——“流动性溢价”（Illiquidity Premium）是否会随之消解。戈尔茨坦给出了独到的定性：他更愿意将其称为“辛劳溢价”（Effort Premium）。在私募领域，除了流动性锁定本身，投资者还付出了大量繁复繁重的基础工作，多做了工作自然应该获得超额补偿。 为了说明技术对透明度的必然改造，戈尔茨坦自嘲“虽然只有52岁，但说起往事觉得自己像个老古董”。他回忆道，早年间彭博终端还处于雏形阶段，分析师若想了解一只公募债券的条款，必须亲自逐字翻阅招股说明书（prospectus），然后一个键一个键地将到期日（maturity）、偿债基金计划（sinking fund schedule）和赎回计划（call schedule）手动敲进电脑系统里。如今要是把这套工作方式告诉入行二十年以内的年轻从业人员，他们会觉得你简直不可理喻。在他看来，“技术潮流是无法阻挡的”，未来十年私募市场的透明度毫无疑问只会大幅提升。 - **辛劳补偿的逻辑**：过去私募市场的高收益相当一部分来自处理非标准化信息所需的重度人工付出，即“辛劳溢价”。 - **不可逆的技术同化**：正如公募债券与公募股票市场经历过的从手工录入到全面标准化、数字化的历史进程，私募市场的部分组件也将逐步走向标准化与高透明度。 --- ### `45:00` 未来投资者的超额收益（Edge）何在：全资产组合、想象力与全球本土网络 伴随整个物理世界都在走向高度数字化与透明化——就像现代人手腕上普遍佩戴着实时监测生理指标的智能设备一样，终端资产所有者（Asset Owner）在私募敞口不断扩大的背景下，绝不会容忍透明度的倒退。主持人紧接着提问：当早年依靠搭建几台微机系统（Micro-systems）或特定工作站就能形成的早期技术壁垒荡然无存时，未来投资经理的核心“超额优势”（Edge）究竟来自哪里？ 戈尔茨坦将全行业的竞争比作一条“每个人都必须越跑越快的跑步机”，并将未来的竞争壁垒拆解为三大核心支柱： ```mermaid flowchart TD accTitle: 未来资产管理的三大核心超额优势 accDescr: 展示贝莱德COO提出的三大未来投资优势支柱。 Edge["未来投资者的核心优势 (Edge)"] Edge --> P1["全资产组合方案<br/>Whole Portfolio Solutions"] Edge --> P2["想象力与工具落地<br/>Imagination & Tooling"] Edge --> P3["复杂地缘与本土连接<br/>Geopolitics & Ground Networks"] P1 --- D1["打破传统细分割裂<br/>围绕客户资产配置重构组织"] P2 --- D2["编码摩擦归零<br/>想法与概念表达成为最高门槛"] P3 --- D3["跨越地缘与科技冲突<br/>依赖真实深度的本土人际圈层"] ``` 第一重优势在于**全资产组合（Whole Portfolio）的服务能力**。回顾资产管理行业的历史发展，各大机构以往的组织架构是与客户构建投资组合的真实逻辑相割裂的：业内被生硬切分成纯固收机构、纯股票机构、被动指数管理机构、主动选股机构、量化系统化机构，以及公募和私募壁垒，最终却把拼装所有拼图的难题推给了客户。未来的核心优势将是打破这些藩篱，帮助客户从顶层设计并管理一揽子的综合全资产组合。 第二重优势在于**使用先进工具的想象力与表达能力**。虽然生成式AI让编程门槛大幅降低，构建软件的摩擦成本趋近于零，但构建技术本身的能力反而变得愈加重要。戈尔茨坦分享了一则趣闻：他曾在一场峰会上被问及“从大学里最想招聘什么专业的毕业生”，他脱口而出回答“英语系毕业生”（English majors），结果导致成千上万封来自家长推荐自己孩子的邮件瞬间涌入他的邮箱。但他极其认真地坚持这个观点：在代码实现周期从过去的数年缩短为短短几天的时代，最珍贵的品质是拥有卓越的想象力，并能用清晰精准的语言将其构想表述出来。 第三重优势在于**应对世界剧变的本土化物理网络**。当今世界正在经历国家安全、尖端科技与全球资本之间前所未有的剧烈碰撞，地缘政治走向碎片化，人口结构加速改变，加上以AI为代表的“仿佛降临地球的外星科技”（alien technology），投资管理所要求的全球互联深度远超以往。戈尔茨坦强调，无论阅读多少彭博资讯或传统报章，都无法替代真实的本土网络——正如他原本计划亲自到访、后改为线上举行的海湾地区（Gulf region）客户交流中，当地一线投资人所反映的宏观与市场图景，与公开媒体报道的内容截然不同。这种深扎一线的人际网络与本土洞察，将构成算法无法替代的长期壁垒。 ### `50:00` AI 时代的招聘要求与底层业务重构 谈到前线实地调研（on the ground）对于获取超额收益（edge）无法被取代的价值后，主持人抛出了一个招聘层面的问题：作为贝莱德（BlackRock）的高管，在招募具体岗位的员工时，AI 时代的面试流程和所提问的问题，与四年前相比是否发生了显著变化？Rob Goldstein 坦言，他希望整个公司的大规模招聘都在发生转变，而就他个人的面试而言，提出的问题确实已经彻底改变。他表示自己对贝莱德未来的机遇无比兴奋，但同时也极度清醒地意识到当前时代的刚性要求——“一切都需要被重新想象（everything needs to be reimagined）”。 Goldstein 指出，商业领袖们眼下面临的核心挑战在于：如何重构你正在做的事、如何重构业务增长模式，以及在新的现实下如何重新定义与客户的互动方式。他认为这种重塑往往不是自上而下强推的，讽刺的是，许多真正的创新将由基层员工自下而上地驱动。目前各家机构仍处于摸索与脑力激荡的试验阶段，根本的变革尚未全面展开。但他坚信，如果将时间推移至 2030 年，几乎所有顶尖企业都将变得与今天截然不同；这不会在一夜之间发生，也不会拖延到五年以上，真正的蜕变将发生在两者之间的时间窗口内。 - Goldstein 强调领袖面临的挑战是重塑做事方式、增长逻辑和客户交互界面。 - 当前行业大部分行动还停留在小规模实验和概念验证阶段，尚未发生系统性替代。 - 预计到 2030 年，主流卓越企业的组织形态和运作流程将与现今产生根本性断层。 --- ### `51:30` 公募与私募界限的模糊化：流动性与披露的光谱 随着对话延伸到代币化（Tokenization）与另类资产，主持人提出了一个关于资本市场结构的终极问题：现在很多热门的头部初创公司（如 Anthropic 等），其股权已经被包装成 SPV，甚至在周末通过 Hyperliquid 等链上平台以代币形式交易；而传统上市公司的披露负担可能在未来逐步放宽，或许只需半年、一年甚至更久披露一次。这是否意味着公募（Public）与私募（Private）之间的那道明确界限正在消失，最终演变成一种关于流动性与信息披露程度的连续“光谱（spectrum）”？在回答前，Goldstein 特别插话指出，AI 和数字资产（Digital Assets）表面看似不同，实则是极度紧密相关的两个领域。 Goldstein 明确赞同“光谱化”的观点。回顾过去数十年的金融业发展，几乎所有领域的界限都在不断模糊，而这背后的核心推动力正是技术。在过去的投资管理中，之所以存在大盘股、中盘股等僵化的资产分类“盒子”，是因为当时的算力无法做到自底向上的精准建模，只能将某种资产类别作为一个乐高积木块来粗略拼装；而当今的技术让投资者能够细致建模到底层单个标的，并直接置于整个投资组合中考量。尽管前几年市场的主流叙事是“在私募市场停留更久（private for longer）”，导致当下公开市场的上市公司数量竟然比 Goldstein 刚入行时还要少，但像 OpenAI、Anthropic、SpaceX 这样的巨头最终依然展现出奔向公开市场的紧迫感，因为公开市场的资本容量与估值功能仍是私募市场无法替代的。未来的市场不会非黑即白，而是呈现出高度个性化的连续光谱：有的机构偏好数字钱包托管，有的则坚守传统托管账户。 ```mermaid flowchart TD accTitle: Public to Private Asset Spectrum Evolution accDescr: 展现从过去的机械分类盒子向基于现代技术与流动性连续光谱转变的架构图 subgraph Legacy["传统划分模式（基于技术限制）"] L1["固定分类盒子（Box Approach）"] L2["资产类别机械拼装（如中盘股乐高块）"] L3["公募与私募间泾渭分明"] L1 --> L2 --> L3 end subgraph Modern["现代技术驱动形态（光谱化）"] M1["自底向上精确建模底层单资产"] M2["多维度连续光谱（流动性 / 披露频率）"] M3["灵活账户形态（数字钱包 vs 传统托管）"] M1 --> M2 --> M3 end Legacy -.->|"算力与数据整合升级"| Modern ``` --- ### `55:00` Goldstein 与 Larry Fink 的分歧及对行业护城河的复盘 在节目的尾声，主持人抛出了一个趣味八卦问题：作为长期共事的亲密搭档，Rob Goldstein 和贝莱德 CEO 兼董事长 Larry Fink 之间最大的分歧点通常是什么？Goldstein 笑着回应，其实两人在长远目标和终点愿景上的判断高度一致，真正的分歧往往在于推进的节奏与时间线：Larry Fink 往往是一个“明天就要做成（tomorrow person）”的激进派，而 Goldstein 自己则更倾向于提醒“这里还有大量脏活累活要做，落地至少需要好几年”。 访谈结束后，两位主持人 Joe Weisenthal 与 Tracy Alloway 展开了复盘讨论。他们分析了贝莱德这类金融巨头的护城河（Moat）： 1. **规模与合规壁垒**：当前许多程序员流行使用轻量模型进行所谓的“氛围编程（vibe coding）”，虽然小巧灵活，但绝对不会有大型金融机构敢将敏感数据和关键业务架构托付给自制的玩具应用。金融领域极高的监管合规门槛，在 AI 时代反而构成了坚不可摧的护城河。 2. **全资产组合可视性（All-Portfolio Visibility）**：传统的金融服务往往按资产类别垂直分割。但大客户需要的是如何将私募信贷、指数化被动投资等各种乐高积木拼在一起的整体解决方案。只有贝莱德这样掌握全谱系资产数据与系统能力的超级巨头，才能真正提供全景式的投资组合构建。 3. **真实世界调研（Channel Checks）重获超额溢价**：主持人打趣道，像去商场数 Gap 货架上有几件毛衣这种草根调研（channel check），可能变得比以往更具价值。在有效市场假说（EMH）的框架下，一旦信息变成数字化文本被传上网（甚至被做成播客或彭博新闻），它就会立刻被市场定价并被喂给大模型做训练集；因此，那些“尚未被转化为模型训练数据”的离线第一手信息，将成为市场上争夺超额回报最珍贵的源泉。 ### `01:00:00` 公募与私募市场的融合与实地调研价值的回归 主持人围绕着公募市场与私募市场界限的日趋模糊展开了深入探讨。如果各类私募资产在未来被全盘代币化（Tokenized），并且在全组合视角下能够与公募资产以完全相同的方式被追踪、管理与呈现，那么传统意义上公私募之间的楚河汉界将彻底消失。对话随即碰撞出一个更深维度的洞见：当所有公开和已被数字化的信息都被大语言模型和量化算法彻底吞噬并极致提炼后，未来市场中最稀缺、最高价值的阿尔法（Alpha）会流向何方？ 答案很可能是那些能够亲自“走向真实世界”、搜集尚未被数字化的物理世界信息的人。对话中特别提到了近期市场对 Citrini Research 派分析师亲自前往霍尔木兹海峡实地考察这一事件的热烈反响——这种走出数字终端、奔赴现场获取第一手非标准化情报的做法，正成为最受瞩目的超额认知来源。过去二十年里，社会将极高的溢价赋予了那些精通代码、擅长抽象数学与定量计算的人才；而如今，随着模型逐步接管了纯粹的数字分析与代码撰写，天平正在迅速摆向另一种截然不同的核心能力：在真实世界中穿梭走访、建立人际信任、敏锐捕捉现场现实的人际交往与社会技能。 - **市场边界消融**：代币化技术与统一资产组合管理平台正消除公募资产与私募资产在流动性和可见度上的鸿沟。 - **信息差的新阵地**：当模型能够消化一切线上数字化数据时，未经数字化录入的现实世界“现场情报”成了唯一无法被算法预先掌握的资源。 - **人才溢价的钟摆摆动**：金融行业的核心价值正在从过去二十年极度崇尚代码与数字逻辑的技术极客，重新转向具备实地调查、社交连接与人际沟通能力的现场调研者。 ### `01:01:10` 播客结语与制作致谢 在对金融与科技未来趋势的探讨告一段落后，主持人 Tracy Alloway 与 Joe Weisenthal 正式结束了本期对谈，并进行了节目的收尾致谢与推广介绍。 - **主持人社媒信息**：Tracy Alloway（账号：`@TracyAlloway`）与 Joe Weisenthal（账号：`@TheStalwart`）。 - **制作团队致谢**：感谢制作人 Carmen Rodriguez（账号：`@CarmenArmand`）、Dashiell Bennett（账号：`@Dashbot`）、Kale Brooks（账号：`@KaleBrooks`）以及 Kevin Lozano（账号：`@KevinLloydLozano`）。 - **社群与内容获取渠道**：听众可访问彭博社官网的专属页面（`bloomberg.com/odd-lots`）阅读每日通讯与回溯过往全期音频，亦可通过 Discord 讨论社区（`discord.gg/oddlots`）进行全天候交流。彭博订阅用户更可通过 Apple Podcasts 上的 Bloomberg 专属频道享受全程无广告收听体验。 ## Notable quotes > `05:15` 基本上就是把打印出的收益率表（yield tables）用传真机发往全世界 — Rob Goldstein > `05:15` 你花每台 1 万美元买 10 台 Sun 工作站并将它们串联在一起，实际上就能做到过去必须斥资数百万美元的超级计算机才能做到的事情 — Rob Goldstein > `05:15` 资产管理业务在核心本质上就是一项信息处理业务（information processing business） — Rob Goldstein > `10:00` 掉落在地球上的外星科技，人类正在摸索如何去使用它 — Rob Goldstein > `12:55` 我们甚至连第一局都没开始，球场上大概还只是在奏国歌的阶段 — Rob Goldstein > `15:20` 干得越好，任务越多（the reward for good work is more work） — Rob Goldstein > `17:40` 绝大多数人使用技术的唯一目的，就是借助它完成手头的工作，然后准时下班回家 — Rob Goldstein > `37:00` 如果你招揽了最顶尖的模型构建师和工程师，却在算力上不对他们加以任何约束，他们永无止境的算力胃口足以让整家公司破产 — Rob Goldstein 引述 Charlie Hallac > `37:00` 目前还没有任何人真正开始着手优化自己的 Token 消耗 — Rob Goldstein > `50:00` 一切都需要被重新想象（everything needs to be reimagined） — Rob Goldstein ## Mentioned BlackRock（贝莱德）, Rob Goldstein（罗伯·高盛 / 罗伯·戈尔茨坦）, Tracy Allaway, Joe Weisenthal, Ben Golub, Charlie Hallac（查理·哈拉克）, Tony Kim, Stephen Boyd（斯蒂芬·博伊德）, Larry Fink, Aladdin（阿拉丁）, Sun Microsystems（Sun 工作站）, Bloomberg（彭博）, OpenAI, Anthropic, SpaceX, Hyperliquid, Citrini Research, MIT AI Lab, Stanford University（斯坦福大学）, Apple Podcasts ## Video description <p>The last few decades have been marked by a number of megatrends in finance including the extraordinary growth of asset managers, the rising importance of technology, and the ascent of private markets. BlackRock, the world's biggest asset manager, is emblematic of all these developments. On this episode, we talk to BlackRock COO Rob Goldstein about the company's early technological history, the development of its famous risk management technology Aladdin, and how BlackRock is navigating being both a user and major provider of AI. We discuss his view of the 'SaaSpocalypse,' how BlackRock is thinking about token consumption and compute constraint, as well as the future of private markets.</p> <p><a href="http://bloomberg.com/subscriptions/oddlots" data-link-cookie="1" data-destination="http://bloomberg.com/subscriptions/oddlots">Subscribe to the Odd Lots Newsletter</a><br>Join the conversation: <a href="https://discord.gg/oddlots" data-destination="https://discord.gg/oddlots">discord.gg/oddlots</a></p><p>See <a href="https://omnystudio.com/listener">omnystudio.com/listener</a> for privacy information.</p> [Source](https://podtrac.com/pts/redirect.mp3/tracking.swap.fm/track/UVBrz8bN8aM2Xe47PEPu/traffic.omny.fm/d/clips/e73c998e-6e60-432f-8610-ae210140c5b1/8a94442e-5a74-4fa2-8b8d-ae27003a8d6b/4cec22d8-717e-46c6-ab87-b43b01512ecb/audio.mp3?in_playlist=982f5071-765c-403d-969d-ae27003a8d83)

[5:15] Rob Goldstein: 基本上就是把打印出的收益率表（yield tables）用传真机发往全世界 [5:15] Rob Goldstein: 你花每台 1 万美元买 10 台 Sun 工作站并将它们串联在一起，实际上就能做到过去必须斥资数百万美元的超级计算机才能做到的事情 [5:15] Rob Goldstein: 资产管理业务在核心本质上就是一项信息处理业务（information processing business） [10:00] Rob Goldstein: 掉落在地球上的外星科技，人类正在摸索如何去使用它 [12:55] Rob Goldstein: 我们甚至连第一局都没开始，球场上大概还只是在奏国歌的阶段 [15:20] Rob Goldstein: 干得越好，任务越多（the reward for good work is more work） [17:40] Rob Goldstein: 绝大多数人使用技术的唯一目的，就是借助它完成手头的工作，然后准时下班回家 [37:00] Rob Goldstein 引述 Charlie Hallac: 如果你招揽了最顶尖的模型构建师和工程师，却在算力上不对他们加以任何约束，他们永无止境的算力胃口足以让整家公司破产 [37:00] Rob Goldstein: 目前还没有任何人真正开始着手优化自己的 Token 消耗 [50:00] Rob Goldstein: 一切都需要被重新想象（everything needs to be reimagined）

[5:00] Rob Goldstein: The founding thesis of BlackRock was really about how do we bring those technology capabilities, which were not really available on the buy side... use them as the core of building an asset manager. [10:00] Rob Goldstein: Why can't we have a first draft of everything we produce be created through AI, whether it be a client presentation, an internal document, a prospectus?... The reason why you're very deliberate about saying first draft is because we have 16 people who check the first draft. [45:00] Rob Goldstein: We're living through a time where those who could have imagination and articulate it, the ability to implement that has never been as fast... literally if it used to be years, now it's like days. [50:00] Rob Goldstein: Larry is more of a tomorrow person, and I'm more of a 'well, there's work to do here, this is going to take a few years' person. [60:00] Joe Weisenthal: Field trips out to the world to collect information that has not been digitized yet is going to be where all the action is.

[0:00] Whatever your goal, trade show giveaways, client gifts, or team gear, 4imprint has the promo products to match. With thousands of options, from apparel and drinkware to tech and totes, it's easy to find the right fit for your brand and budget, with standout choices at every price point. And with their 360-degree guarantee, you can be 4imprint certain your order will show up just right, right on time. Explore more at 4imprint.com. 4imprint. 4certain. The thing about AI for business? It may not automatically fit the way your business works. At IBM, we've seen this firsthand. But by embedding AI across HR, IT, and procurement processes, we've reduced costs by millions, slash repetitive tasks, and freed thousands of hours for strategic work. Now we're helping companies get smarter by putting AI where it actually pays off, deep in the work that moves the business. Let's create smarter business, IBM. We be right back on Google with no unwelcome surprises Plus most transfers happen in under 20 seconds which means your money arrives in less time than you been listening to me It simple and free to sign up when you download the Wise app. Be smart, get wise. T's and C's apply. Bloomberg Audio Studios. Podcasts, radio, news. Hello and welcome to another episode of the Odd Lots podcast. I'm Tracy Allaway. And I'm Joe Weisenthal. Joe, when I think about the world of finance. Yeah. And what I would describe as mega trends of recent years and decades. There are definitely two that spring to mind. Maybe a third trend, although I don't know if it's mega. It probably is. The first is definitely the rise of the buy side. So this idea that, you know, it very much used to be all about the banks and those were the ones that we kind of obsessed over. And then you have this extraordinary growth in the asset management industry. The second megatrend has to be technology, right? Think about the rise of electronic trading electronic risk management model driven risk management I like this And then the third semi megatrend I don know rise of private markets right Yeah Oh I have another one Okay That I think is legit The sort of power law domination of a few mega companies and sort of whether it the big get bigger, the winner take allness or the winner take mostness of the industry. I would add that as a mega trend. Okay, that's great. And I would say that connects to this episode. Absolutely. So So we have four big slash mega trends and we have the perfect guest to talk about all of those four things. And it comes at a time when obviously we're in this new sort of technological wave with AI. It very much feels like everyone in professional finance wants to figure out a way of being involved in AI in one way or another. It's funny. I was meeting up with someone who works at a very large bank the other day and they were talking about how work that they've done 20 years ago, their managers are now adamant. it has to be put into a big Excel database because they're about to shove all of that into an AI model. So you can see there's this urgency. And there's a perennial question over how much of it is real. Random data that exists somewhere, they need to have it as part of a data lake or whatever so that the AI model knows about it. Exactly. And I think the question for this is how much of it is managers who are latching on to the AI trend versus how much of this is actually going to become productive useful technology for finance So we can definitely talk a little bit about that but we do have the perfect guest. Let's do it. All right. So we're going to be speaking with Rob Goldstein. He is, of course, the COO of BlackRock, someone who has literally lived through basically all of these megatrends that we just described. Yes. And the question is, how many hours do we have? Needless to say, those megatrends we could talk about for quite some time. Yeah. We could do a series. We could do a series. A five-part series. And one thing I would just politely identify, I think as people talk about the big getting bigger, I think there's an underlying catalyst towards those who could provide a better value proposition are getting bigger. And I think that's the key theme that's happening, particularly with regard to the buy side. I know you were saying, what was that, a polite, you didn't push back, but a polite nuance or something like that. But one could say that the ability to provide a better value proposition is itself a function of size in many instances, because the larger have the full suite, the whole menu, right? Of services. So even there, there's like a nuance upon nuance. [5:00] A hundred percent. And what's interesting is I think if you ordered the mega themes, you can make a very strong argument that it all comes down to technology. And technology is enabling things and value propositions to be achieved that traditionally just couldn't be done. Great. Let's just talk about that. Is it true that like some of BlackRock's foundational, by the way, The chances that I say Blackstone in this conversation. Yeah, that's not good. I apologize in advance. We had a pre-call and I said that I've been in this business a long time. We had a pre-call with Rob like several weeks ago. I said Blackstone. And then like I could tell there was like a silence. I said the wrong thing. He's heard it all before. I hope so. So is it true that like the foundational culture of BlackRock is very much tied to technology? because the story that I always used to hear was about a Sun workstation and Ben Golub. Yeah, and Ben is still a close friend and mentor, Ben being one of the founding partners of BlackRock. I think if you zoom out a little bit, because I think the history of BlackRock is very reflective of what the past 30 35 years have been in terms of the companies that have been most successful And if you look at the founders of BlackRock they were a group of people who were pioneers with regard to structured products and the evolution of the mortgage market. And what they realized is that banks at the time, the sell side, as you guys sort of laid the groundwork. Banks at the time were using supercomputers and they were very expensive computers to structure things. And then the way they were selling those products was literally by faxing yield tables all over the world. And I know when I say this to 20-something-year-old, 30-something-year-olds, including my own children who are when they're early 20s, a lot of the people back then didn't even have computers. Computers were like there was one for a group of people as opposed to everyone had one on their desk. So the thesis behind forming BlackRock was that we could actually build models that would help provide risk transparency for those type of instruments and help the end asset owner We could build those models And through the Sun workstation is the innovation If you were reasonably clever you didn need to be a genius but if you were reasonably clever, you could buy 10 Sun workstations for $10,000 each and link them together and effectively do what previously only supercomputers that cost millions of dollars could do. So the founding thesis of BlackRock was really about how do we bring those technology capabilities, which were not really available on the buy side, how do we use them as the core of building an asset manager? That was the founding thesis. And I think one of the real success factors, and I think that when you look today, what I'm about to say seems like very odd, but I guess this is odd loss. So it's perfect. But when I started at BlackRock in 1994, when we had roughly 80 people, $19 billion in assets under management, I was in the data and analytics team I was effectively a data analyst And like today data technology analytics are where the cool kids are Back then it was not where the cool kids are trust me And the whole concept of recognizing very early on that the asset management business at its core is an information processing business today is so obvious. But if you rewind back 10 20 30 years ago that was a very unique novel concept we really do need like five hours i know and this is kind of a tangent you mentioned the idea of like you could string 10 sun sun workstations together to make a supercomputer we are actually going back to the future or the future back a little bit these currently in computing i have a good friend who has promised to help me later this year i'm going to buy like five mac minis because he says you can host your own llm now from home if you just have like four or five Mac mini strong together and then you don't have to do then you don't have to depend on any other company's data center for unlimited token usage so this is going to kind of come back well it's it's interesting though I think that's a real question in terms of where we are right now because I don't know the answer to this but I could make two good arguments one is along the lines of what you're saying yeah the other is we are living in an [10:00] If you really just think about what's happening with AI, you could convert energy to intelligence. And the more money you spend on energy, the more intelligence you have. You could argue different than many technology trends that we've had over the past couple of decades. This is a technology trend that requires capital and it requires spending a lot of money. Yeah. And I think there's a real question about whether or not this is a very expensive technology or if this will ultimately wind up being five guys at home in a garage with a handful of Mac minis can accomplish miracles. I don't know if that's certain yet which one is going to prove true. I want to get in more on your history and Aladdin and the technology that you've been involved in building over these years. but maybe a big picture question is like one of the things about AI that really strikes me as potentially interesting with how it's going to affect finance is AI is non-deterministic. You put in a query and you don't really know what you never know if you're going to get the same output twice or whatever And I curious like for one you don know how it arrived often and models can explain themselves And this is an issue for finance which is they often not explicable why the output came out But then there this other element of you're not going to get the same thing twice. And I'm curious, like, when you think about how that fits into the history of technology, whether that is a source of anxiety for finance, you need to be able to show your work often, in many cases, or you need to be able to like, traditional software you write a line of code and as long as there's no bug it'll produce the same result a thousand times in a row is this new is this something that is going to be a difficult tension to work around no question absolutely no question and just to give an analogy okay i am sure there have been many many many people who sat in this seat through the years and said by x x being years ago there will be self-driving cars you won't need driver's licenses and so on and so forth. I think the tolerance people have for computers to make mistakes is very different than the tolerance people have for humans to make mistakes. So that's just a societal starting point For all intents and purposes I think you could make a cohesive argument I love Marvel movies as a family That one of our things This is like alien technology has been found on the planet Earth and now we figuring out how to use it. And one of the remarkable things about the technology, even if, and I know from listening to you, you've played with a lot of the coding tools. If you look at the coding tools, they write code and then there are bugs in the code and then they find the bugs and they fix the bugs. So like the way we've been trained to think about a computer is how could that happen? Wouldn't it be smart enough to write the code without the bugs? Right. Like why do I have to prompt it to fix itself? Because it's much more like a person. It's much more about thinking than this binary zero and one structure that we've become used to for computers. And I think one of the remarkable, if you spend time with any of the big technology firms at the big AI companies, the frontier model providers, they use this term regulated industries. And needless to say regulated industries like financial services we need to be certain that we have the appropriate processes and controls in place So through one lens that a big friction Through another lens I think it actually a competitive advantage to the industry, because if you think about it, we have so many controls. We have so many controls as a natural part of the process. So for example, one of the first things we did within BlackRock, when the technology became available, we created this rule that we call this principle that we call the first draft principle. Why can't we have a first draft of everything we produce be created through AI, whether it be a client presentation, an internal document, a prospectus. And the reason why you're very deliberate about saying first draft is because we have 16 people who checked the first draft. And that ability as a starting point is a very sort of strong catalyst towards leveraging and getting to know the technology. But I would actually argue that today, the technology has provided a lot of people like individual productivity. But at an enterprise level, if you look at the case study, [15:00] It's not clear to me we've entered the first inning of the actual enterprise implementation. I still think the national anthem is sort of being played. And I think that the actual overhang between what the models can do and the fact that this is technology that needs to be implemented, it requires organizational design, business process reengineering. Like implementing technology is hard and takes time. And we haven't even started that enterprise implementation yet. Can you talk a little bit more about this in the context of Aladdin? Because I think part of the concern here that Joe was getting at is that you have these models that are getting more complex and more difficult to predict. We don't know what they're going to spit out. They're non-deterministic, as Joe said. And meanwhile, you have this risk management technology that has already for years been described as a black box. And I'm sure you have opinions on that particular label. But if it gets more sophisticated, are people going to fully understand what it's actually doing? Well, let me start out by saying we haven't described it as a black box No not yet So we come back to that in a minute But importantly AI as a technology is not new I wish I knew the exact year but like the AI lab at MIT was created in the 1950s We started our AI lab in 2018. So these methods have been used for a long time. At a very sort of simple level that I'm sure would offend a lot of people, you could think about old AI was about numbers. New AI is about language. And the language element of it creates all sorts of humans communicate much more through language than numbers. So it creates a whole host of other unintended consequences. But when you look at a platform like Aladdin as an enterprise platform, and by the way, I would make a cohesive argument. Everything I'm saying, you could make the same case with regard to the Bloomberg terminal. I knew this was going to come up in this conversation. But I waited the first couple of minutes. Fair point. So if you think about these technologies, first, the reward for good work is more work. So the to list for these technologies is infinite like genuinely infinite Every year BlackRock winds up having more engineers We have roughly 5 engineers data analysts modelers Every year we wind up having more engineers. And every year, we wind up having a bigger to-do list of enhancements we could put within Aladdin. So the first element of the AI capability, and I would argue the most mature use case that exists at an enterprise level, is coding. So the ability to go through that to-do list, the velocity of that, is off the charts. The second component, and this is one of the challenges of these enterprise expert systems is that the number of times I've been in a meeting with a client where they say, you know, why doesn't Aladdin do this? And I'm like, hmm, I think Aladdin does that. But let me let me I don't want to like blurt it out. Let me sort of follow up and then I'll leave the meeting. I'll call the people smarter than me and they'll be like Aladdin done that for seven years And you like okay the ability for people to keep current in technology is very hard And you know as much as we like the technology most people their goal with technology is to just interface with it to do their jobs and then like go home. So the ability to take an expert system that today requires a lot of knowledge and keeping up with it and instead just type in what you want it to do and an agent will be the ultimate real-time user of aladdin that will then do those activities all the same controls will exist the four eyes principle all of those controls will exist but that ability to have users to have clients access all the untapped capabilities that today they don't know about, I think the value enterprise technology is going to provide going forward. Aladdin and other enterprise technology is actually going to be much greater than at any point previously. It's actually, it's extremely exciting to me because there's nothing more frustrating than being in a meeting where someone is complaining you don't do something when you actually do it. This must be a thing for a bunch of enterprise software, right? Because no one... [20:00] talk to other software people and no one uses all the specs and no one uses all the features, no one knows all the features, et cetera. But you said something that is something that's kind of one of my hobby horses. If it's the agent that's using Aladdin rather than the sort of inhuman, does that change how you think about UX? It's a great question. We debate this a lot. and i'll tell you about something i saw yesterday but i think it has to but at the same time i think there will be people who will want to do it themselves okay i think there will be people who want to do it themselves you know it's interesting even in this age of everything being on the phone yeah you still need a website that people could access so i think there still will be people who want to do it themselves. I saw a demo yesterday of a tool from one of these AI companies that I hope I could articulate it well enough, but it will look at a website and it showed it looked at one of our websites and redesigned it to be more like optimal user friendliness So it was the opposite of it was using AI to almost do the opposite of what you said make it easier for humans And it was one of these things where, you know, in the demo, because websites are public, so they were able to do things with our own stuff that we didn't know about. And as they're showing what they would do, and when I say they, what a computer would do after 10 minutes of processing with their own website, you're like, hmm. In X months or maybe a year or two, tools like this will re-engineer every website on the planet Earth and they will all be more user-friendly. For us, hopefully, and not the agents. Whether you're planning a big tech event, launching a new campaign, or just stocking up on team gear, finding the right promotional products makes all the difference. 4imprint offers thousands of options from on apparel and premium drinkware to tech totes and giveaways so you can find the right fit for any audience purpose or budget You can customize it all your logo your message your look and many items come with no setup charge to help you save. And if you're really watching the bottom line, you'll find standout choices at every price point, so you can make a real impact while staying on budget. Plus, you'll get expert help, fast turnaround times, and their 360-degree guarantee, so you can be four-imprint certain your order will arrive on time and look exactly right. Whatever your goal, 4imprint makes it easy to find your perfect promo match. Explore the possibilities today at 4imprint.com. 4imprint, for certain. Support for this show comes from public.com. If you're actively involved in your portfolio, you probably catch yourself repeating the same actions. Buying the dip, manually sweeping idle cash, putting on a hedge. On public, you can now create AI agents that handle all these tasks on your behalf. Just describe what you want to do in plain English, like if the VIX hits 25, buy a put option on the S&P 500. Or if my cash balance goes above $20,000, move the excess into my direct index. You approve the workflow and your agent handles the rest Monitoring the market watching for your conditions and executing your strategies exactly as defined An investing platform driven by your intent not just your clicks You can also get full read and write access to your account via the public API Go to public.com slash market and fund your account in five minutes or less. That's public.com slash market. Paid for by public investing. Brokered services by open to the public investing, Inc member FINRA and SIPC advisory services by public advisors, LLC, SEC Registered Advisor. Complete disclosures available at public.com slash disclosures. Get the news you need in just 15 minutes. Start your day with Bloomberg Daybreak, the podcast with a global view on the stories that matter. I'm Nathan Hager. And I'm Karen Moscow. Join us each morning for curated stories on current events, politics, business, and foreign relations. Plus one conversation on the day's biggest developments, all in just 15 minutes. Subscribe to Bloomberg Daybreak for a precise, thoughtful take on the stories that matter. Listen to Bloomberg Daybreak each morning on Apple, Spotify or anywhere you listen. Can you talk a little bit more about the moat around Aladdin? Sure. Because this is the other big talking point of the moment, which is the Sasspocalypse idea. And in the age of vibe coding, everyone is just going to go out and design their own portfolio. [25:00] Risk management system. Let me start out. Let me provide a little context broadly about what we think is going to happen or what a group of us think are going to happen. And then let me go into Aladdin because I think that they're somewhat related. So one of our portfolio managers, one of our technology portfolio managers is a gentleman, Tony Kim. and a year ago two years ago i said tony if today there are a hundred lines of code in the world in 2030 how many are there and he said a million and i was like you're out of your mind and he said no no no like i was quite thoughtful about that i didn't make up a number 10 times. We've all seen how much Joe is coding. Well, but this is an important dimension because 10 times 10 times 10 times 10 times. So if you believe which, you know, it's going to be a multiple, you could argue, is it three? Is it seven? Is it 10? Is it 14? But it's going to go up every year by a multiple given these tools So the amount of code in the world is going to go up dramatically And I think that when you look at a platform like Aladdin and in many regards I think it would be like things with Bloomberg. They're at centers of an ecosystem. The ecosystem is highly regulated. The ecosystem has zero tolerance for fault or error. the processes you do not only leverage tremendous amounts of proprietary data, so it's not within these models or accessible by the models, but importantly, clients are putting their most sensitive data into these platforms. And then what you're doing in terms of workflow is highly, highly, highly idiosyncratic and requires this combination of people, process, and technology. So it's a hard time in this world, in my opinion, to predict what 2050 looks like. But when you look forward 10 years in our industry, these platforms, if anything, are going to do more not less And I think what really is going to be unlocked is that users like Joe are going to access these platforms through their own coding tools but these coding tools are going to be central to those platforms. I didn't forget your black box comment. Going back to... You'll never forget. Going back... No, I think it's important. other people no i i think it's important and i think that technology 10 or 20 years ago certain technologies were designed to be closed systems okay you know where i'm going certain technologies were designed to be closed systems aladdin was one of those technologies and roughly 10 years ago we started this open aladdin campaign well before anything to do with this round of AI we're in, which was all about the future of technology is going to be some people are going to want to interact with it through a keyboard and a mouse. And many people are going to want to interact with it through code. Because even if you're graduating with an English major, if you're graduating at most schools at this point, you taken a coding class So just the amount of technical expertise that you come in is a whole different level so this is like opening up more like api endpoints and things like that as opposed to say like open source like that open in that yeah open within a closed ecosystem and that is a critical element because open within a closed ecosystem one of the things and it's like obvious after the fact but one of the things that we realize like oh my god this is amazing and it's so valuable is that when you call our APIs, for example, in Aladdin, your permissions go through. So if you think about the complexity of managing permissions in an asset manager with thousands of people, you could see some portfolios, you could see some portfolios, they're different portfolios. You could trade, you're not allowed to trade, you could only confirm trades. The complexity of those permissions. So the fact that when you call an API, it knows what you can and can't access. That whole control plane and layer is an incredible value proposition. And the more you have people coding and the more you have people interacting with systems in more technical ways, the more valuable those control planes are. [30:00] actually wind up being. What happens to the flip side? Okay, you've described some core pieces of infrastructure that are regulated, there's proprietary data, etc. Clients are putting their most sensitive info in the world. And your view is that the value of those platforms will grow. Is there a software that's not that? Are those zeros? Absolutely. No, absolutely. Absolutely. Well, first of all, these things have long tails. So actually going through the process of retiring a system is a lot of work, no matter what the system is. So there are long tails. But there are certain technologies that are really, and we all use them, there are certain technologies that are about collating public information and making it easy for you to access. And I think it fair to say that the AI tools that exist are the ultimate oracles in being able to do that in being able to scour all public sources and give you back information in the way that you most comfortable with So particularly that segment of SaaS, which is that convenience layer, where they don't really have proprietary data, they're not really in the workflow. They're a convenience technology. I think those convenience technologies are in trouble. They're certainly going to be looking for ways of reimagining their value proposition. Well, one of the reasons we wanted to have you on the podcast is because you are both a provider of AI, viz Aladdin, and also a user of AI at your own company. And we've all seen various executives and managers talk about AI as a productivity enhancing tool, and they tend to talk about it in very general terms. So I would be curious to hear from your perspective exactly what a productivity enhancement at BlackRock actually looks like. I will give you a productivity enhancement from Friday of last week And by the way what I describing I think is the future everywhere i hope blackrock gets to that future faster than others but i believe it the future so we have been doing a lot of work enhancing a lot in many ways and one of the big themes that we have is how do we provide more transparency in the private markets to be as close as possible to the public markets in pursuit of this whole portfolio. So we have a large program of work that's been going on for quite some time. So I try to spend hours every Friday getting demos of things we're working on. So my Friday afternoon demo along the theme that I just described was a demo of a tool that was awesome. But let me go through how it was created. And this was the first time end to end that at least I've been shown this. So So a group of people that included portfolio managers risk professionals engineers product managers a group of people sat in a room for multiple hours talking about how this capability should work That discussion was recorded That discussion, through the recording of it, created a functional document. That functional document was lightly tweaked. That functional document was then put in some of the AI coding tools that we use. That document through the AI coding tools led to a prototype. There was a debugging process that you lived through that we just described. And on Friday, I've seen a lot of prototypes before in my sort of 32 years can imagine so i i know the questions to ask where you see the prototype is like a thin shell yeah but if you press enough the shell cracks like this wasn't like a prototype this was like the real deal and when you look at that cycle we effectively collapsed what would have taken the unit of measurement would have been months. Now the unit of measurement... [35:00] was days. Now, it will still go through our software development lifecycle. It will be tested, all of those things. But when you look at that as a productivity tool, this goes back to the 10xing, the amount of lines of code in the world every year. There's just going to be an explosion in the ability to engineer things. Now, I'm wondering if the coding tool, when it sees a transcript of a meeting like that, Do you think it weights the participants by like title and level of importance? No, this is a real question. It probably does. It would be weird if it didn't, right? It's interesting. My intuition is it doesn't. My intuition is this would be a great exercise in those who talk most are probably most reflected. I heard about a company that's doing third party consulting for helping companies implement software. And one of the things that they're doing is they get on a Zoom with the client and they have a camera trained on the whiteboard that they're working on. And that video file also is part of so in addition to the audio because whiteboards are the lingua franca of software development so that also gets uploaded to it And so that it like sees that whole workflow and stuff Right Pretty It pretty wild stuff I curious Tokens There been a lot of headlines lately about token sticker shock. We are spending a lot on inference. I'm curious if you could tell us anything about token consumption this year versus last year at blackrock but also like one of the things that we're also seeing and it's very it's related to this is like the compute constraints that have long been talked about as theoretical are starting to actually bite people who are sort of like anthropic users or like whatever like can you talk a little bit about token consumption and is the compute constraint real from your perspective if you could snap your fingers and get 100,000 more plugged in GPUs, would that be a big help right now? Well, I think that you have to sort of go back in time a little bit. So one of the key, key, key lessons learned that I've experienced, and I remember one of the near founding partners, a gentleman, Charlie Halleck, who unfortunately passed away, what he instilled in me, among many other things was if you have the right modelers and engineers and you leave them unconstrained they will bankrupt the company in terms of their insatiable appetite for compute In the old days, you had a physical data center that was the constraint. You would have to order hardware. At some point, people would say, we're outgrowing our data center. You'd say, let's wait a year and see what happens. So in today's world, the elasticity is obviously very different. But as a starting point, there are a group of people within BlackRock, and I think this is true in any sort of great financial services company, that if you leave them unconstrained from a compute power perspective, 20 years ago, they would have bankrupted the company, 10 years ago, and today they they will so part of it is how do you think about where to invest i don't know offhand the number of tokens we're consuming today relative to a year but it's multiples it's multiples higher and i think that we're still at a point again as a company but also at a broad industry level where i don really think the game has started yet so i don think anyone really knows what the token consumption is going to be And equally as important I don think anyone has really started optimizing their token consumption. You know what someone needs to do is build an Aladdin for token efficiency. It is a certainty that not only will that exist, but one of the gentleman who the person who actually leads our AI lab, a gentleman, Stephen Boyd. I had called him in the early days of this happening. And I'm like, what are we missing? Like, Stephen, help me understand what are we missing? He said two things. And he is a professor of engineering at Stanford. He said, you're missing two things. One is that articulate language is very, very powerful. And I'm like, okay, like, what does that mean? And he's like, think about it. I'm a professor, if I'm reading a paper and it's not written that well, you check everything. And if you're reading a paper that's written really well, you just assume it's right. So that was the first point. The second point he had, which relates to this topic, he said, just remember, there's a bunch of graduate students here and everywhere else [40:00] Right now we're working on like the most boring aspects of this too, including how to have these models be more efficient. So right now it's all about the quest for intelligence. I think we're going to see it pivot slightly to the quest for enterprise use cases. And then I think it's going to pivot very quickly to the quest for efficiency of how you're accessing things. We're not there yet. We'll see you next time. Thank you You find standout choices at every price point so you can make a real impact while staying on budget Plus you get expert help fast turnaround times and their 360 guarantee So you can be 4imprint certain your order will arrive on time and look exactly right. Whatever your goal, 4imprint makes it easy to find your perfect promo match. Explore the possibilities today at 4imprint.com. 4imprint. 4certain. Support for the show comes from public.com. If you're actively involved in your portfolio, you probably catch yourself repeating the same actions. Buying the dip, manually sweeping idle cash, putting on a hedge. On public, you can now create AI agents that handle all these tasks on your behalf. Just describe what you want to do in plain English. Like, if the VIX hits 25, buy a put option on the S&P 500. Or, if my cash balance goes above $20,000, move the excess into my direct index. You approve the workflow and your agent handles the rest. Monitoring the market, watching for your conditions, and executing your strategies exactly as defined. An investing platform driven by your intent, not just your clicks. You can also get full read and write access to your account via the public API. Go to public slash market and fund your account in five minutes or less That public slash market Paid for by Public Investing Brokered services by Open to the Public Investing Member FINRA and SIPC. Advisory services by Public Advisors, LLC, SEC Registered Advisor. Complete disclosures available at public.com slash disclosures. Got a business problem? There's a TED Talk for that. Stay updated on everything business on TED Business, a podcast hosted by Columbia Business School professor Modupe Akinola. Every week, she'll introduce you to leaders with unique insights on work, answering questions like, how do four-day work weeks work? Do will a machine ever take my job? Get some surprising answers on TED Business wherever you listen to podcasts. I want to go back to something you said about private markets quickly, which is this whole portfolio management idea, this idea that via a system like Aladdin, you can manage your private assets the same way you would manage your public assets and you get more transparency around pricing and things like that. So part of the sales pitch with private markets has been this idea of an liquidity premium and the idea that you earn a little bit more because these things are not treated like public market assets Once you start to integrate them via new technology into your system once you start to be able to manage them much more similarly to a more liquid asset does some of that sales pitch start to go away? I think of it as an effort premium. There's a liquidity element to it. And then there's like an effort premium. If you have to do more work, presumably you need to be compensated for that. But I also believe that, and I feel so old as I talk like this. I'm only 52, but I feel so. As crazy as it sounds to many people, when I started, the way you would get information about public bonds, because even Bloomberg back then was a bit nascent. Yeah. You would read a prospectus and you would type into a computer is the maturity. This is the sinking schedule. This is the call schedule. Now, today you say that to people who entered the industry in the past 20 years, and they think you're nuts. So I think there's no fighting technology. That's my own opinion. So I think it is certain that if you say in 10 years are the private markets more or less transparent, they're certainly more transparent. And I think [45:00] If you think about your life, almost everything in your life is becoming more transparent. Like, it's funny, I don't think any of us, I try not to affix technology on my body, but it's very rare that three people would be together where someone doesn't have some device that's monitoring like their blood pressure in real time. So like everything is pointing towards more transparency. I have a hard time believing that as private markets, exposures in portfolios grow, the end asset owner is going to wind up with less transparency. So I think this is just the direction of travel. I do think over time, certain components of it will start to become more standardized, very similar to things that happened in the public bond markets and the public equity markets. And then there will be new innovations in other ways. But I believe that things will become much more transparent it's a certainty in the future and this is i guess more from just an investment standpoint but like what is the investors source of edge in the future you know one point maybe there was a source of edge because you were early to jump on and see the potential of stringing together some microsystems and you could replicate whatever are these workstations. Just looking forward for the portfolio manager, et cetera, what constitutes edge? It's a great question. And I think across industries, if you think of them as a treadmill, everyone's going to have to run faster. There's no question that that's the case. I think the edge, you could break up into three categories. First, I think the edge is going to be helping clients with their whole portfolio, as opposed to just pieces. And I think if you look at the asset management industry, a very significant, I would say, evolution aspect of the asset management industry is that the industry organized itself inconsistent with how clients build portfolios. you had fixed income shops you had equity shops you had index managers you had active managers you had systematic manager you're in public markets you had private markets and then you force the client to put all this stuff together so the industry is going to pivot more towards helping with the whole thing okay which is a different type of edge I think that the ability to use these tools is going to be an edge in and of itself So notwithstanding coding is going to be easier and there will be multiples more Like the ability to build technology is going to become more, not less important, even if the frictions to build technology go down. I would argue, I was at a conference and someone asked me like, who would you like to hire coming out of university? And I said, English majors. And that was a big mistake because then like thousands of people emailed me that their child's an English major while I speak to. But the reason I said English majors, and I believe this, like we're living through a time where those who could have imagination and articulate it, the ability to implement that has never been as fast yeah and literally if it used to be like years now it's like days so the ability to have those ideas the implementability of those ideas is going to be different than any point so that creativity and imagination and then lastly obviously the world is becoming very complicated You basically have this collision between national security technology and capital You have this fragmentation of geopolitics. You have the changing demographics of the world. You have what's happening with this alien technology that's now been found on the planet Earth. So if you think about the global connectivity that's required to really manage portfolios and how geopolitics is going to implement within portfolios going forward, I think that requires on-the-ground networks as much as it requires technology. And it's interesting. I'm sure you guys have had a similar experience. I read the paper. I read Bloomberg. Thank you. I watch a lot of stuff. But when you talk to, and I was actually supposed to be traveling last week in the Gulf region, and instead I did a virtual tour, when you speak to our clients there, you get a very different picture of what's going on than what you're reading. And I think those networks are going to become more... [50:00] are important in terms of edge as we look forward. Nothing will replace being on the ground. Well, actually, on that note, you're in a position where you're hiring people constantly for very specific roles. Are the processes or the questions that you're asking people now different in the age of AI versus what they were like four years ago? It's interesting. I hope so. I know the questions I'm asking people are. I hope that's at scale. Maybe I should follow up on that. But the questions I'm asking, I couldn't be more excited about the opportunities ahead for BlackRock, but I couldn't be more aware of the requirement in today's world. everything needs to be reimagined. And what we need and what I think leaders are going to be faced with, and ironically, I think a lot of the reimagination is going to come bottom up, but how do you reimagine what you do How do you reimagine how you grow How do you reimagine and how you interact with clients in this new world And I think that again I don think that started I think right now it a bit of a brain teaser a lot of experiments, but it has not started. Almost every great company, if you walk in in the year 2030, is going to be fundamentally different than today. And the way I think of this, this is not an overnight reimagination but it's not a five-year reimagination it's somewhere in between those two you know we said in the beginning we could go for hours and there's a million other things that we could ask one last question on my mind i've been i've tried i've asked this version to a few different people and it maybe kind of relates to the other kind of tokens tokenization but just this idea that's gonna get confusing i know it's really annoying already because there's Well, can I just say one thing before you ask your question? It's going to get confusing. It already is. But importantly, AI and digital assets are very related topics, extremely related topics. Well, this kind of maybe maybe you can fold that into the answer here. But you know like an interesting thing There a lot of very exciting private companies that people want to get access to But in many cases they already kind of trading in some way And there might be these SPVs that people have already or some token somewhere that trades on Hyperliquid on the weekend that represents somehow shares of Anthropic or whatever. You need other hobbies. But I'm curious from your perspective, like, and another thing that relates to this is the disclosure obligations for existing public companies seem like they're going to come down over time. And maybe companies will only have to report every six months or a year, maybe never. Like, will there always be a bright line between what's a public and what's a private asset? Or is it just going to be this spectrum of liquidity and disclosure, but no clear definition of what it means anymore between public and private? It's a great question. I think over time, it'll be more of a spectrum. I think that when I look at the broad industry over the past multiple decades, the lines, one of the defining themes, the lines almost across everything you could imagine have been getting more and more blurry And I would argue that been because of technology So if you think about the old style boxes that existed they were convenience technologies because you couldn bottom up model things You basically said, okay, mid cap equities have this attribute and like that's a Lego piece as opposed to let me model what the individual stocks are. And then I could think of it in the context of a whole portfolio. So I think that what's happening is that technology is enabling those lines to be less discrete and more blurry across spectrums and across how do clients build portfolios? How do these Lego pieces get put together to actually achieve their objectives? I think that there's a lot going on embedded in your question, including you see, and to me, this is one of the most remarkable things that is happening, is that two, three, four years ago, the whole narrative was private for longer. And certainly, I don't know how this is true that there's fewer public companies today than when I started. I've checked it. It's actually true because I didn't believe it. Yeah. But you wonder, like, how is that possible? Now, that's... [55:00] said, the flip side of it is you see for these companies like OpenAI, Anthropic Space, their race to go public, because at some point, the public markets provide a lot of value propositions that are beyond what the private markets can provide. So I think you see this sort of tale of two cities that's happening. But there will be a spectrum, including a spectrum of certain end investors and institutions that want to have a digital wallet and certain end investors and institutions that want to have a traditional custody account. And it's not a binary thing. It's about technology enabling that personalization. I have one more question, sort of a wild card, but what do you and Larry Fink disagree about the most? and the reason i ask is no i'm genuinely interested you you've been working alongside each other for years and years and years now i have a personal interest in close collaborative yeah tracy and i disagree on a lot so it a very legit question i would say often we both see things similarly in terms of the end point Larry is more of a tomorrow person and I more of a well there work to do here This is going to take a few years. I would say that that's typically it. Okay, Larry is a Joe, and I'm a Rob, I guess. All right, Rob Goldstein, thank you so much for coming on All Thoughts. That was great. That was a lot of fun. That was great. Thank you so much. Awesome. My pleasure. Yeah. Joe, that was a really fun conversation and a lot to think about. I mean, I do think when we're talking about moats around some of these businesses, you brought up the power dynamics law very early on. And it does feel like a lot of the moat is basically size and data and capability that you have. And I can't imagine that you hear this. You hear people saying that they're like vibe coding a bunch of different programs and they all look kind of interesting and cool. But like it hard for me to imagine a big moat around those businesses if you just like plugged in a few instructions into cloud code And no one is going to be putting anything of importance or sensitivity into some homemade I mean that definitely true That's what was interesting, this idea that actually the regulatory moat around finance actually becomes a valuable thing in the age of AI. Totally. You know, it's interesting, too, not really AI-related, but the idea of being able to provide a solution, all of portfolio visibility. What is one reason why the big get bigger within finance? One reason is because only a really large entity would have the capacity to be able to, like, and here's what we can offer you with private credit, and here's what we can offer you with indexing, et cetera. So, OK, the end investor has this big portfolio consisting of lots of different types of asset classes. As he mentioned, historically, the industry has sort of been verticalized by asset class. If you want some entity that has how do all the puzzle pieces fit together, which is the essence of portfolio construction, then theoretically, you just want like a really big company that understands all of it. Yeah I also thought and it was kind of also on the source of edge question is we talked about this before It always seems like one of the fun jobs in finance would be the uh the channel check person the person who goes to the mall to oh yeah the field trips yeah the field trips it's like okay how many sweaters are on the gap shelf whatever maybe that becomes even more valuable because like things that have not yet been put into a model yeah and look i love that people listen to odd lots i love that people read the news, particularly financial news on Bloomberg. I've never thought, by and large, that people read an article and it's like, I'm going to make an investment decision based on that. Hopefully it helps inform their thinking or their processes in some way. But no one listens to a podcast and then goes out and buys the stock, by and large. But it's because once it's out there in the digital world, it's kind of priced in already. So there's going to be an intense hunt for information that is like truly like has not been put into the model. You are such an EMH purist. I am. It's amazing. But right, like the value of people who can find information that has not turned into training data yet for a model, that data is going to get like super valuable and maybe like more valuable and more reason to like just like get out on the road and stuff like that. No, I largely agree with that. The other thing I thought was really interesting on... [60:00] to your last question, was the sort of melding of public and private markets. And it's also what I was kind of getting at with the whole portfolio thing. If all these private assets are tokenized and treated in the same way in a portfolio, or at least visible in a portfolio in the same way as a public asset would be, it seems like that distinction starts to get really, really blurry. Oh, totally. Not to go back to a point that I just made, but on this, like, I think we're I'm sorry. I moved on too quickly. No, no, this just clicked to me too. But it's like, okay, where are we? Maybe there'll be more value in people who go out on the road and get information that's not in the model. We saw this recently with all the people going crazy for Cetrini's analyst in the Strait of Hormuz. And this idea that field trips out to the world to collect information that has not been digitized yet is going to be where all the action is. It feels like that's going to be a big thing. No, it sounds trite, but I feel like the pendulum has swung from, you know, for the past 20 years, if you were a smart person who could think in terms of numbers and code, you were probably very valued by society. And now the pendulum sort of swings to those on the ground relationship building social skills. It's it's an interesting transition. It's an interesting time. All right. Shall we leave it there? Let's leave it there. This has been another episode of the All Thoughts podcast. I'm Tracy Allaway. You can follow me at Tracy Allaway. And I'm Joe Weisenthal. You can follow me at The Stalwart. Follow our producers, Carmen Rodriguez at Carmen Armand, Dashiell Bennett at Dashbot, Kale Brooks at Kale Brooks, and Kevin Lozano at Kevin Lloyd Lozano. And for more OddLots content, go to Bloomberg.com slash OddLots, where the daily newsletter and all of our episodes. And you can chat about all these topics 24-7 in our Discord, discord.gg slash OddLots. And if you enjoy Odd Lots if you want us to do more on the ground investigative reporting and analysis then please leave us a positive review on your favorite podcast platform And remember if you are a Bloomberg subscriber you can listen to all of our episodes absolutely ad free All you need to do is find the Bloomberg channel on Apple Podcasts and follow the instructions there. Thanks for listening. Thank you. choices at every price point. And with their 360 degree guarantee, you can be 4imprint certain your order will show up just right, right on time. Explore more at 4imprint.com. 4imprint, 4certain. Healthcare doesn't always work great. If you've ever waited on a refill or couldn't schedule an appointment, you get it. That's the kind of stuff Optum is changing. They're using data and technology to integrate patient care, pharmacy, and everything else. So healthcare is connected, not complicated. What's that look like? Cheaper prescriptions that are easier to get and care that looks at the whole person how you need it. Optum is helping make health care work as one for everyone. Learn more at business.optum.com. Welcome to Alaska Airlines. We're now flying nonstop from Seattle to London. So enjoy a cup of tea in your lie flat suite and get ready to experience all London has to offer. Lie back, relax, and let your dreams take flight. Welcome to Alaska Airlines.