How we built Grok Bot in a month | Roman Ugarte (SpaceXAI)
发布时间 来源
Episode 设置
在最近的一期播客中，GrokBot 的产品负责人、Cursor 的前第15号员工罗曼·乌加特（Roman Ugarte）讨论了 GrokBot 的迅速崛起和核心差异化优势，该产品被誉为“目前世界上最热门的AI产品”。乌加特揭示了该产品的起源、理念和未来愿景，将其成功归因于几个关键的、尽管最初并不显而易见的决定。
GrokBot 的旅程始于一个小型、专注的内部团队，他们的任务是“进入洞穴约一个月”，目标单一：构建一款“能将代理引入公司其他部门的卓越知识工作产品”。这种紧张而封闭的孵化在仅仅一个月内就形成了一个功能原型。一个关键的早期决定是将 GrokBot 作为一个独立产品构建，而不是将其整合到现有的 Cursor 平台中。这使得其拥有清新的愿景，摆脱了以开发者为中心的工具的限制和品牌关联，确保“用户体验的每一个像素”都为通用知识工作量身定制。
乌加特强调了支撑 GrokBot 效率的两个关键早期决定。首先，致力于**云优先、持久化模型**，即机器人存在于云端，独立于用户的本地设备。这使得从任何设备进行无缝交互成为可能，并确保机器人可以在用户计算机无需激活的情况下自主执行任务。这与许多当前AI产品形成对比，后者通常将代理与本地环境绑定，给用户造成“痛点”。
其次，GrokBot 的设计赋予每个机器人**“自己的计算机”**。这不仅仅是API访问，它模仿了人类的工作方式：点击像素，与界面交互，并使用缺乏良好API支持的工具。乌加特将其比作一位不会与你共享笔记本电脑的人类同事的入职，强调了专用资源对于真正独立工作的重要性。这种“同事视角”（colleague-pilled）的心态——持续追问“人类队友会怎么做？”——指导着产品决策，偏爱自然的、对话式界面，而非复杂、充满旋钮的仪表盘。
该产品的成功也源于不懈地专注于使其“就是能用”。这包括大力“卸载”可能让用户不知所措的不必要特性，转而专注于强大的后端改进。例如，解决代理与Salesforce等销售工具交互的特定技术障碍，为市场团队开辟了重要的工作流程。这种对100%完成任务而非90%完成任务的承诺，是 GrokBot 的独特之处，它允许用户真正委派工作而无需微管理。正如乌加特所发推文：“一个能完成100%工作的AI，与一个只能帮你完成90%工作的AI，在本质上是不同的。”
对200-300名早期用户（包括咖啡店老板等非传统用户）的手动引导，提供了无价的、无偏见的反馈。这个过程帮助团队识别了关键模式，例如出现了管理其他专业机器人的“幕僚长机器人”，并证实了用户倾向于抽象化内部机制（工具调用、思维链），以获得更简单、更直观的体验。
展望未来，GrokBot 描绘了一个愿景：“你应该拥有一支AI机器人团队，它们能帮助你完成工作，也能帮助你打理生活。”这一抱负延伸到将机器人整合到业务团队中，解决复杂的组织环境问题，甚至开发能够就紧急事务提醒用户的主动型机器人。这种市场进入策略映照了 Cursor 在编程领域的成功：构建出如此不可或缺的产品，以至于早期用户在他们的职业生活中强烈需求它。
乌加特还谈到了公司快速迭代和持续创新的文化。尽管规模显著增长，团队仍保持着“创业公司感觉”——一种混乱、快速变化的能量，优先将新的AI能力转化为今天的有用产品，即使这意味着每六个月“删除产品”以适应不断变化的环境。这种适应性，加上对提供一流用户体验的专注，对于驾驭竞争激烈的市场至关重要。
对于新用户，乌加特建议为 GrokBot 提供充足的上下文和工具，然后直接询问它能减轻哪些任务。对于资深用户，秘诀是构建交互框架，引导机器人将输出写入一个单一、有组织的存储位置，以便于审查，从而培养一个真正协作的数字工作空间。
In a recent podcast, Roman Ugarte, Head of Product for GrokBot and former employee number 15 at Cursor, discussed the rapid ascent and core differentiators of GrokBot, hailed as the "hottest AI product in the world right now." Ugarte revealed the product's origin, philosophy, and future vision, attributing its success to a few crucial, albeit initially non-obvious, decisions.
GrokBot’s journey began with a small, dedicated internal team, tasked to "go off into a cave for about a month" with the singular goal of building an "amazing knowledge work product that brings agents to the rest of the company." This intense, isolated incubation led to a functional prototype in just a month. A key early decision was to build GrokBot as a standalone product, rather than integrating it into the existing Cursor platform. This allowed for a fresh vision, free from the constraints and brand associations of a developer-focused tool, ensuring every "pixel of the experience" was tailored for general knowledge work.
Ugarte highlighted two critical early decisions that underpin GrokBot's effectiveness. First, the commitment to a **cloud-first, persistent model** where bots live in the cloud, separate from the user's local device. This allows for seamless interaction from any device and ensures bots can autonomously perform tasks without the user's computer needing to be active. This contrasts with many current AI products that often tie agents to local environments, creating "paper cuts" for users.
Second, GrokBot's design empowers each bot with **its own "computer."** This isn't just about API access; it mimics how humans work, clicking on pixels, interacting with interfaces, and using tools that lack well-supported APIs. Ugarte likened it to onboarding a human colleague who wouldn't share your laptop, emphasizing the importance of dedicated resources for truly independent work. This "colleague-pilled" mindset – continually asking "what would a human teammate do?" – guides product decisions, favoring natural, conversational interfaces over complex, knob-filled dashboards.
The product's success also stems from a relentless focus on making it "just work." This involved aggressively "unshipping" unnecessary features that might overwhelm users and instead concentrating on robust backend improvements. For instance, addressing specific technical hurdles for agents interacting with sales tools like Salesforce unlocked significant workflows for the go-to-market team. This commitment to 100% task completion, rather than 90%, distinguishes GrokBot, allowing users to truly delegate work without micromanagement. As Ugarte tweeted, "an AI that does 100% of the job feels categorically different from one that gets you 90% there."
Manual onboarding of 200-300 early users, including non-traditional profiles like a coffee shop owner, provided invaluable, unbiased feedback. This process helped the team identify key patterns, such as the emergence of a "Chief of Staff bot" managing other specialized bots, and confirmed the user preference for abstracting away internal mechanics (tool calls, chain-of-thought) for a simpler, more intuitive experience.
Looking ahead, GrokBot envisions a future where "you should have a team of AI bots that help you with your job and help you with your life." This ambition extends to integrating bots into business teams, tackling complex organizational contexts, and even developing proactive bots that can page users for urgent matters. This go-to-market strategy mirrors Cursor's success in the coding world: building something so indispensable that early adopters demand it in their professional lives.
Ugarte also touched on the company's culture of speed and continuous reinvention. Despite growing significantly, the team maintains a "startup feeling" – a chaotic, fast-moving energy that prioritizes translating new AI capabilities into useful products today, even if it means "deleting the product" every six months to meet the evolving landscape. This adaptability, combined with a focus on delivering best-in-class user experience, has been crucial for navigating a hyper-competitive market.
For new users, Ugarte advises giving GrokBot ample context and tools, then asking it directly what tasks it can alleviate. For power users, the tip is to scaffold interactions, directing bots to write outputs to a single, organized store for easier review, fostering a truly collaborative digital workspace.
摘要
Roman Ugarte helped incubate and build Grok Bot, the popular new knowledge-work agent from SpaceXAI. A small, isolated team took it from first line of code to a working internal product in four weeks, and to a hugely successful public launch just three weeks later. Before Grok Bot, Roman led Growth at Cursor, where he helped scale the company from 15 people to over 1,000 before its acquisition by SpaceX.
*In our in-depth conversation, we discuss:*
1. The origin story of Grok Bot
2. The key decision to build it from scratch instead of adding it to Cursor
3. Why the team personally onboarded nearly 300 of its first users
4. The two early product decisions that made Grok Bot so successful
5. Their “colleague-pilled” product philosophy
6. Roman’s advice on moats, and what has allowed Cursor to keep winning in the most competitive market in the world
*Brought to you by:*
WorkOS—Make your app enterprise-ready, with SSO, SCIM, RBAC, and more: https://workos.com/lenny
Mercury—Radically different banking, now with Command: https://mercury.com/
*Episode transcript:* https://www.lennysnewsletter.com/p/how-we-built-grok-bot-in-a-month
*Archive of all Lenny's Podcast transcripts:* https://www.dropbox.com/scl/fo/yxi4s2w998p1gvtpu4193/AMdNPR8AOw0lMklwtnC0TrQ?rlkey=j06x0nipoti519e0xgm23zsn9&st=ahz0fj11&dl=0
*Where to find Roman Ugarte:*
• X: https://x.com/romanugarte_
• LinkedIn: https://www.linkedin.com/in/romanugarte
• Website: https://x.ai
*Where to find Lenny:*
• Newsletter: https://www.lennysnewsletter.com
• X: https://twitter.com/lennysan
• LinkedIn: https://www.linkedin.com/in/lennyrachitsky/
*In this episode, we cover:*
(00:00) Introduction
(02:09) The origin story: building from scratch in one month
(08:40) Why Grok Bot was built as a separate product
(11:20) Manually onboarding a couple hundred people
(14:29) Hiding internal mechanics from users
(18:41) Timeline from beta to public launch
(19:14) Unshipping features and simplifying
(23:50) Early use cases and feedback
(26:50) Product philosophy: “Grok Bot can now”
(30:02) Cloud-first architecture
(33:12) The fresh-start advantage
(35:54) The vision: a true team of AI colleagues
(39:20) The “colleague-pilled” framework
(42:36) Work versus personal: one product or two?
(47:14) Long-lived agents, persistent memory, and the computer abstraction
(51:04) Grok Bot as an always-on infovore and chief of staff
(53:35) How fast the team moves and what preserves the startup feeling
(58:20) SpaceXAI pillars
(1:00:44) The first 90% vs. the last 10%
(1:03:30) Moving fast at scale
(1:06:40) How Cursor kept winning in the most competitive market in the world
(1:10:04) Company values: “deleting the product” and “just do the thing”
(1:11:45) Moats: discovered, not planned
(1:15:11) Tips for new users and power users
(1:18:00) Lightning round and final thoughts
*Referenced:*
• Grok Bot for iOS: https://apps.apple.com/us/app/grok-bot/id6794501026
• Grok Bot for Android: https://play.google.com/store/apps/details?id=ai.x.grok.bot&hl=en_US
• How I AI: Grok Bot + Grok 4.6—what’s great (and what’s still hype) & Lessons from spending $20,000 on Devin in one month: https://www.lennysnewsletter.com/p/how-i-ai-grok-bot-grok-46whats-great
• Codex: https://chatgpt.com/codex
• ChatGPT Work: https://chatgpt.com
• Shopify: https://www.shopify.com
• Salesforce: https://www.salesforce.com
• The playbook for building high-talent-density teams | Adam Ward, Head of Talent at Cursor: https://www.lennysnewsletter.com/p/the-playbook-for-building-high-talent
• Superhuman: https://superhuman.com
• Stripe: https://stripe.com
• OpenClaw: https://openclaw.ai
• Listen: OpenClaw: A power user’s guide to the most powerful personal AI tool since ChatGPT: https://www.lennysnewsletter.com/p/listen-openclaw-a-power-users-guide
• Hermes: https://hermes-agent.nousresearch.com
• From skeptic to true believer: How OpenClaw changed my life | Claire Vo: https://www.lennysnewsletter.com/p/how-openclaw-changed-my-life-claire-vo
• Cursor: https://cursor.com
• Grok Build: https://x.ai/build
• Roman’s post on X, “An AI that does 100% of the job feels categorically different from one that gets you 90% there”: https://x.com/romanugarte_/status/2087344044435505175
• Notion: https://www.notion.com
• Casablanca: https://www.imdb.com/title/tt0034583
• Monk: https://www.imdb.com/title/tt0312172
• Exa: https://exa.ai
• Desiderata - Words for Life: https://allpoetry.com/desiderata---words-for-life
*Recommended books:*
• Cat’s Cradle: https://www.amazon.com/dp/038533348X
• The War of Art: Break Through the Blocks and Win Your Inner Creative Battles: https://www.amazon.com/War-Art-Through-Creative-Battles/dp/1936891026
_Production and marketing by https://penname.co/._
_For inquiries about sponsoring the podcast, email podcast@lennyrachitsky.com._
Lenny may be an investor in the companies discussed.
GPT-4正在为你翻译摘要中......
