Metadata

original_title: How Perplexity builds product

source: https://www.lennysnewsletter.com/p/how-perplexity-builds-product

Notes

Summary

摘要：文章探讨了Perplexity公司的产品构建方法，特别是其联合创始人兼产品负责人Johnny Ho的观点。Perplexity采用“AI优先”的策略，员工在处理问题时优先寻求AI的帮助，而不是直接咨询同事。公司组织结构灵活，像粘菌一样优化协调成本，大多数项目由两到三人小组完成，强调自我驱动的团队文化。Ho认为，随着时间推移，具备技术背景和产品敏感性的项目经理将成为公司中最具价值的人才。同时，Perplexity注重用户反馈，并在项目管理中采取去中心化决策，确保快速迭代和灵活应对变化。

要点总结：

AI优先策略：员工在产品构建过程中优先向AI咨询，提升工作效率。
灵活的组织结构：采用小团队模式，优化协调成本，鼓励平行工作。
自我驱动的团队文化：强调招聘自我驱动的个人贡献者，减少管理层级。
用户反馈的重要性：将用户反馈转化为直观产品，确保团队自主决策。
去中心化决策：项目管理中，鼓励团队成员在设计和实施阶段进行独立决策，快速迭代。

Origin

👋 Hey, Lenny here! Welcome to this month’s ✨ free edition ✨ of Lenny’s Newsletter. Each week I tackle reader questions about building product, driving growth, and accelerating your career.

👋 嘿，莱尼在这里！欢迎来到本月的 ✨ 免费版 ✨ Lenny 时事通讯。每周我都会回答读者关于构建产品、推动增长和加速职业生涯的问题。

If you’re not a subscriber, here’s what you missed this month:

如果您不是订阅者，以下是您本月错过的内容：

For $150 a year, get access to these posts and every prior post, along with an invite to a private Slack community with global meetups, a mentor matching program, interview prep support, live AMAs, and more. I guarantee you’ll get 100x the value of a subscription or your money back.

每年仅需 150 美元，即可访问这些帖子和之前的所有帖子，并受邀加入私人 Slack 社区，其中包括全球聚会、导师匹配计划、面试准备支持、现场 AMA 等。我保证您将获得 100 倍的订阅价值或退款。

Founded less than two years ago, Perplexity has become a many-times-a-day-use product for me, replacing many of my Google searches—and I’m not alone. With fewer than 50 employees, the company has a user base that’s grown to tens of millions. They’re also generating over 20 million ARR and taking on both Google and OpenAI in the battle for the future of search. Their recent fundraise of 63m values the company at more than $1 billion, and their investors include Nvidia, Jeff Bezos, Andrej Karpathy, Garry Tan, Dylan Field, Elad Gil, Nat Friedman, Daniel Gross, and Naval Ravikant (but sadly not me 😭). Nvidia CEO Jensen Huang said he uses the product “almost every day.”

Perplexity 成立不到两年，已经成为我每天多次使用的产品，取代了我的许多 Google 搜索，而且我并不孤单。该公司员工数量不足 50 人，但用户群已增长至数千万。他们还创造了超过 2000 万美元的 ARR，并在搜索的未来之战中与 Google 和 OpenAI 展开竞争。他们最近筹集了 6300 万美元，使公司估值超过 10 亿美元，他们的投资者包括 Nvidia、Jeff Bezos、Andrej Karpathy、Garry Tan、Dylan Field、Elad Gil、Nat Friedman、Daniel Gross 和 Naval Ravikant（但遗憾的是不是我） ??）。英伟达首席执行官黄仁勋表示，他“几乎每天”都使用该产品。

I sat down with Johnny Ho, the company’s co-founder and head of product, to give you an inside look at how Perplexity builds product—which to me feels like what the future of product development will look like for many companies:

我与该公司联合创始人兼产品负责人 Johnny Ho 坐下来，让您深入了解 Perplexity 如何构建产品——对我来说，这就像许多公司产品开发的未来：

AI-first: They’ve been asking AI questions about every step of the company-building process, including “How do I launch a product?” Employees are encouraged to ask AI before bothering colleagues.

人工智能优先：他们一直在询问有关公司创建过程的每一步的人工智能问题，包括“我如何推出产品？”我们鼓励员工在打扰同事之前询问人工智能。
Organized like slime mold: They optimize for minimizing coordination costs by parallelizing as much of each project as possible.

像粘菌一样组织起来：它们通过尽可能多地并行化每个项目来优化以最小化协调成本。
Small teams: Their typical team is two to three people. Their AI-generated (highly rated) podcast was built and is run by just one person.

小团队：他们的典型团队是两到三人。他们的人工智能生成（高评价）播客仅由一个人构建和运行。
Few managers: They hire self-driven ICs and actively avoid hiring people who are strongest at guiding other people’s work.

很少有经理：他们雇用自我驱动的IC，并积极避免雇用最擅长指导他人工作的人。
A prediction for the future: Johnny said, “If I had to guess, technical PMs or engineers with product taste will become the most valuable people at a company over time.”

对未来的预测：约翰尼说，“如果我不得不猜测的话，随着时间的推移，具有产品品味的技术产品经理或工程师将成为公司中最有价值的人。”

For more, check out Perplexity. And they’re hiring! For more stories of how the best product teams operate, don’t miss my deep dives into Linear, Shopify, Figma, Notion, Duolingo, Ramp, Miro, Coda, Gong, and Snowflake.

欲了解更多信息，请查看Perplexity。他们正在招聘！有关最佳产品团队如何运作的更多故事，请不要错过我对 Linear、Shopify、Figma、Notion、Duolingo、Ramp、Miro、Coda、Gong 和 Snowflake 的深入研究。

P.S. I’m collaborating with Perplexity on a deep dive into how product managers use Perplexity, and we’d love to hear from you. Fill out this short survey if you use Perplexity regularly, and they’ll reach out to conduct a user interview.

附：我正在与 Perplexity 合作，深入研究产品经理如何使用 Perplexity，我们很乐意听取您的意见。如果您经常使用 Perplexity，请填写这份简短的调查，他们会联系您进行用户访谈。

How Perplexity builds product Perplexity 如何构建产品

From left: Johnny Ho, Aravind Srinivas, and Denis Yarats, co-founders of Perplexity

左起：Perplexity 联合创始人 Johnny Ho、Aravind Srinivas 和 Denis Yarats

1. How do you use AI tools within Perplexity to build Perplexity? 1. 如何使用 Perplexity 中的 AI 工具来构建 Perplexity？

Honestly, at the very beginning, we didn’t know how to do all kinds of things, including product management, project management, finances, HR, etc. We had early access to GPT-3, and as we were figuring out how to build the company, we’d start everything by asking AI, “What is X?” and then “How do we do X properly?” For example, we asked questions like “How do you launch a product? What should be the steps in the launch process?” You get a rough step-by-step process, which for a startup was good enough. Obviously, it’s often not correct on the first try, but neither is a human, right? So we’d just iterate naturally from there.

老实说，一开始，我们不知道如何做各种事情，包括产品管理、项目管理、财务、人力资源等。我们很早就接触到了GPT-3，当我们在想如何做的时候，建立公司时，我们会通过询问 AI“X 是什么？”来开始一切。然后“我们如何正确地做某事？”例如，我们提出了诸如“您如何推出产品？启动过程应该采取哪些步骤？”你会得到一个粗略的逐步过程，这对于初创公司来说已经足够好了。显然，第一次尝试通常是不正确的，但人类也不是，对吧？所以我们就从那里自然地迭代。

Trying to figure things out by ourselves took days, but with AI and some prompting, we could get rolling in five minutes.Trying to figure things out by ourselves took days, but with AI and some prompting, we could get rolling in five minutes.

尝试自己解决问题需要几天时间，但有了人工智能和一些提示，我们可以在五分钟内开始工作。尝试自己解决问题需要几天时间，但有了人工智能和一些提示，我们可以在五分钟内开始工作。

We’re still doing this. This week, for example, I asked Perplexity, “How do I write an email inviting someone to Perplexity Pro?”We’re still doing this. This week, for example, I asked Perplexity, “How do I write an email inviting someone to Perplexity Pro?”

我们仍在这样做。例如，本周，我问 Perplexity，“我如何写一封电子邮件邀请某人使用 Perplexity Pro？”我们仍在这样做。例如，本周，我问 Perplexity：“我如何写一封邀请某人使用 Perplexity Pro 的电子邮件？”

We even tried to use it at times to build our product, but we found AI tooling wasn’t anywhere near good enough when it comes to coding. It could help us write scripts, but if you wanted sustainable code to build a platform on, it didn’t really work. Even today, with the advances and latest models, it still only writes templates. You can’t really design a new long-lived abstraction with it.

我们甚至有时尝试使用它来构建我们的产品，但我们发现人工智能工具在编码方面还不够好。它可以帮助我们编写脚本，但如果你想要可持续的代码来构建平台，它并没有真正起作用。即使在今天，随着技术的进步和最新的模型，它仍然只编写模板。你无法真正用它设计一个新的长期抽象。我们有时甚至尝试使用它来构建我们的产品，但我们发现人工智能工具在编码方面还远远不够好。它可以帮助我们编写脚本，但如果你想要可持续的代码来构建平台，它并没有真正起作用。即使在今天，随着技术的进步和最新的模型，它仍然只编写模板。你无法真正用它来设计一个新的、长期存在的抽象。

2. How many PMs do you have? 2. 你们有多少位PM？

We have only two full-time PMs, in an organization of 50.

在我们 50 名员工的组织中，只有两名全职产品经理。

Our two PMs 我们的两位 PM

Typical projects we work on only have one or two people on it. The hardest projects have three or four people, max. For example, our podcast is built by one person end to end. He’s a brand designer, but he does audio engineering and he’s doing all kinds of research to figure out how to build the most interactive and interesting podcast. I don’t think a PM has stepped into that process at any point.

我们从事的典型项目只有一两个人。最困难的项目最多需要三到四个人。例如，我们的播客是由一个人端到端构建的。他是一名品牌设计师，但他从事音频工程，并且正在进行各种研究，以找出如何构建最具互动性和最有趣的播客。我认为总理在任何时候都没有介入过这个过程。

We leverage product management most when there’s a really difficult decision that branches into many directions, and for more involved projects.

当需要做出一个非常困难的决策并涉及多个方向时，以及对于涉及更多的项目时，我们最会利用产品管理。

The hardest, and most important, part of the PM’s job is having taste around use cases. With AI, there are way too many possible use cases that you could work on. So the PM has to step in and make a branching qualitative decision based on the data, user research, and so on. For example, a big problem with AI is how you prioritize between more productivity-based use cases versus the engaging chatbot-type use cases. Pretty early on, we decided to focus on the former, but there are still ongoing discussions.

PM 工作中最困难也是最重要的部分是对用例进行品味。借助人工智能，您可以处理太多可能的用例。因此，PM 必须介入并根据数据、用户研究等做出分支定性决策。例如，人工智能的一个大问题是如何在基于生产力的用例与引人入胜的聊天机器人类型用例之间进行优先排序。很早就，我们决定关注前者，但仍在进行中的讨论。PM 工作中最困难也是最重要的部分是对用例进行品味。借助人工智能，您可以处理太多可能的用例。因此，PM 必须介入并根据数据、用户研究等做出分支定性决策。例如，人工智能的一个大问题是如何在基于生产力的用例与引人入胜的聊天机器人类型用例之间进行优先排序。很早就，我们决定重点关注前者，但仍在进行讨论。

We plan to hire one or two more PMs over the next year, but the bar for hiring is going to stay very high.We plan to hire one or two more PMs over the next year, but the bar for hiring is going to stay very high.

我们计划在明年再聘用一到两名产品经理，但招聘的门槛仍将非常高。我们计划在明年再聘用一到两名产品经理，但招聘的门槛仍将非常高高的。

3. I assume much of your success has been due to hiring well, and keeping a very high bar. What do you most look for when hiring (that maybe others don’t)? 3. 我认为你们的成功很大程度上归功于良好的招聘和保持很高的标准。您在招聘时最看重什么（也许其他人并不看重）？

Given the pace we are working at, we look foremost for flexibility and initiative. The ability to build constructively in a limited-resource environment (potentially having to wear several hats) is the most important to us.Given the pace we are working at, we look foremost for flexibility and initiative. The ability to build constructively in a limited-resource environment (potentially having to wear several hats) is the most important to us.

鉴于我们的工作节奏，我们首先寻求灵活性和主动性。在资源有限的环境中进行建设性建设的能力（可能需要身兼数职）对我们来说是最重要的。考虑到我们的工作节奏，我们最看重的是灵活性和主动性。在资源有限的环境中进行建设性建设的能力（可能需要身兼数职）对我们来说是最重要的。

When you take a look at resumes of PMs, a lot of them prioritize helping other people and finding alignment. I believe this becomes less important with the advent of AI. So you don’t necessarily need skills around managing processes or leading people as much. We look for strong ICs with clear quantitative impacts on users rather than within their company. If I see the terms “Agile expert” or “scrum master” in the resume, it’s probably not going to be a great fit.

当你查看产品经理的简历时，你会发现很多人都会优先考虑帮助其他人并找到一致性。我相信随着人工智能的出现，这一点变得不那么重要了。因此，您不一定需要那么多管理流程或领导人员的技能。我们寻找对用户而不是对公司内部有明显定量影响的强大 IC。如果我在简历中看到“敏捷专家”或“Scrum Master”等术语，可能不太合适。

Also, AI allows PMs to do a lot more IC work, especially for data analysis and customer insights. You still need some fundamental knowledge, of course (i.e. math, statistics, a basic grasp of programming), but it’s never been easier to be a truly “technical” PM.

此外，人工智能还允许产品经理做更多的 IC 工作，特别是数据分析和客户洞察。当然，您仍然需要一些基础知识（即数学、统计、编程的基本掌握），但成为真正的“技术”PM 从未如此简单。

We still select for culture fit and being easy to work with, but we’re less looking for people who guide other people’s efforts, because it’s not as necessary with AI. This might change as we get to a certain scale, but at the current scale, there are way more products to build than there are people to work on them.

我们仍然会选择文化契合度高、易于合作的人，但我们不再寻找指导他人工作的人，因为这对于人工智能来说并不那么必要。当我们达到一定规模时，这种情况可能会发生变化，但在目前的规模下，要构建的产品远多于开发这些产品的人员。

I think in the future, I expect fewer layers of management in the industry in general. And if I had to guess, a technical PM or an engineer with product taste will become the most valuable people at a company over time.

我认为未来，我预计整个行业的管理层级会减少。如果我不得不猜测，随着时间的推移，技术 PM 或具有产品品味的工程师将成为公司中最有价值的人员。

4. Do you structure your teams around products, user types, user journeys, outcomes, or something in between? Has this changed over the years? 4. 您是否围绕产品、用户类型、用户旅程、结果或介于两者之间的其他因素来构建团队？这些年来这种情况有变化吗？

My goal is to structure teams around minimizing “coordination headwind,” as described by Alex Komoroske in this deck on seeing organizations as slime mold. The rough idea is that coordination costs (caused by uncertainty and disagreements) increase with scale, and adding managers doesn’t improve things. People’s incentives become misaligned. People tend to lie to their manager, who lies to their manager. And if you want to talk to someone in another part of the org, you have to go up two levels and down two levels, asking everyone along the way.

我的目标是围绕最小化“协调逆风”来构建团队，正如 Alex Komoroske 在本套中将组织视为粘液霉菌时所描述的那样。粗略的想法是，协调成本（由不确定性和分歧引起）随着规模的扩大而增加，并且增加经理并不能改善情况。人们的激励变得失调。人们倾向于对他们的经理撒谎，而经理又对他们的经理撒谎。如果你想与组织中其他部门的人交谈，你必须上两层、下两层，一路询问每个人。

Instead, what you want to do is keep the overall goals aligned, and parallelize projects that point toward this goal by sharing reusable guides and processes. Especially with the advance of AI, it’s possible to minimize coordination costs by using AI for “rubber duck debugging” your ideas instead of relying on perfect alignment and consensus. We also keep a “who’s who” list updated in our internal docs, and if you feel the need to reach out to anyone, just do it. This requires a large degree of trust.

相反，您想要做的是保持总体目标一致，并通过共享可重用的指南和流程来并行化指向该目标的项目。尤其是随着人工智能的进步，通过使用人工智能“橡皮鸭调试”你的想法，而不是依赖完美的一致性和共识，可以最大限度地降低协调成本。我们还在内部文档中更新了“谁是谁”列表，如果您觉得需要联系任何人，就这样做吧。这需要高度的信任。

But even more importantly, with AI, you don’t have to reach out to people as often. Sometimes before asking the question you were going to ask someone else, you could first try spending one minute asking AI to reduce coordination costs and give everyone a reasonable jumping-off point to do it themselves.

但更重要的是，有了人工智能，你不必经常与人接触。有时，在问你要问别人的问题之前，你可以先尝试花一分钟要求人工智能降低协调成本，并给每个人一个合理的起点，让他们自己做这件事。

5. How far out do you plan in detail, and how has that evolved over the years? 5. 您的详细计划有多远？这些年来进展如何？

Perplexity has existed for less than two years, and things are changing so quickly in AI that it’s hard to commit beyond that. We create quarterly plans. Within quarters, we try to keep plans stable within a product roadmap. The roadmap has a few large projects that everyone is aware of, along with small tasks that we shift around as priorities change. Being nimble is critical as developments in AI often have unforeseeable impact. For example, the rapid developments in open-source models and context length have had downstream impact on the product, roadmap, and overall business. Just recently, Meta released Llama 3 and Mistral released 8x22B; we’re looking into creative ways to use those models in our product.

Perplexity已经存在了不到两年，而人工智能领域的变化如此之快，以至于很难再做出超出这一范围的承诺。我们制定季度计划。在几个季度内，我们尝试在产品路线图中保持计划稳定。该路线图有一些每个人都知道的大型项目，以及我们随着优先级变化而转移的小任务。灵活性至关重要，因为人工智能的发展往往会产生不可预见的影响。例如，开源模型和上下文长度的快速发展对产品、路线图和整体业务产生了下游影响。就在最近，Meta发布了Llama 3，Mistral发布了8x22B；我们正在寻找在我们的产品中使用这些模型的创造性方法。

The projects in the product roadmap also need to be flexible because new product development runs in parallel with a technical/model development roadmap. Engineers shift between maintaining existing products and building new products, depending on the week. The technical roadmap tends to grow quickly as we run into limitations of existing systems and accumulate tech debt, but we try to prioritize tech debt that unlocks product improvements.

产品路线图中的项目也需要灵活，因为新产品开发与技术/模型开发路线图并行运行。工程师根据周的情况在维护现有产品和开发新产品之间切换。当我们遇到现有系统的局限性并积累技术债务时，技术路线图往往会快速增长，但我们尝试优先考虑能够解锁产品改进的技术债务。

Within a given week, though, plans are fairly stable. Each week we have a kickoff meeting where everyone sets high-level expectations for their week. We have a culture of setting 75% weekly goals: everyone identifies their top priority for the week and tries to hit 75% of that by the end of the week. Just a few bullet points to make sure priorities are clear during the week.

不过，在一周内，计划相当稳定。每周我们都会举行一次启动会议，每个人都会为自己的一周设定高水平的期望。我们有一种设定 75% 每周目标的文化：每个人都确定本周的首要任务，并努力在周末前实现其中的 75%。只需几个要点即可确保本周的优先事项明确。

Taking a moment at the beginning of the week to reflect on meta tasks brings clarity and prevents overly reactive or hectic decision-making. Over time, our ability to estimate size and prioritize based on return on investment has also improved.

在本周初花点时间反思元任务可以带来清晰度，并防止过度反应或忙碌的决策。随着时间的推移，我们根据投资回报来估计规模和确定优先级的能力也得到了提高。

6. Do you use OKRs in some form? 6. 您是否以某种形式使用 OKR？

We try to be as rigorous and data-driven as possible in quarterly planning. All objectives are measurable, either in terms of quantifiable thresholds or Boolean “was X completed or not.” Our objectives are very aggressive, and often at the end of the quarter we only end up completing 70% in one direction or another. The remaining 30% helps identify gaps in prioritization and staffing. Underinvestments, for example, in hiring infra engineers become quickly apparent when infrastructural goals aren’t met.

我们在季度规划中尽可能做到严格和数据驱动。所有目标都是可衡量的，无论是可量化的阈值还是布尔值“X 是否完成”。我们的目标非常激进，通常在季度末我们只在一个方向或另一个方向完成 70%。剩下的 30% 有助于确定优先级和人员配置方面的差距。例如，当基础设施目标未实现时，在雇用基础设施工程师方面的投资不足很快就会显现出来。

7. How do your product/design review meetings work? 7. 你们的产品/设计评审会议如何进行？

After the central objectives and high-level designs are determined, we try to be fairly decentralized in our decision-making. Projects are driven by a single DRI, and execution steps are done in parallel as much as possible.

在确定了中心目标和高层设计之后，我们会尝试在决策过程中实现相当分散的权力。项目由单一 DRI 驱动，执行步骤尽可能并行完成。

The first step for any project is to break it down into parallel tasks as much as possible to reduce coordination issues. We do this in Linear, and I lead this work along with the PM on the team (or whoever is handling the PM duties). We strive for each task to be self-contained—you should be able to execute on it without blockers. And you will likely have to make some controversial decisions, but you can just work through the controversy later.

任何项目的第一步都是尽可能将其分解为并行任务，以减少协调问题。我们在 Linear 中完成这项工作，我与团队中的 PM（或任何负责 PM 职责的人）一起领导这项工作。我们努力使每项任务都是独立的——您应该能够在没有障碍的情况下执行它。你可能不得不做出一些有争议的决定，但你可以稍后解决这些争议。

At the beginning of each project, there is a quick kickoff for alignment, and afterward, iteration occurs in an asynchronous fashion, without constraints or review processes. When individuals feel ready for feedback on designs, implementation, or final product, they share it in Slack, and other members of the team give honest and constructive feedback. Iteration happens organically as needed, and the product doesn’t get launched until it gains internal traction via dogfooding.

在每个项目开始时，都会快速启动对齐，然后以异步方式进行迭代，没有任何限制或审查过程。当个人准备好接受有关设计、实施或最终产品的反馈时，他们会在 Slack 中分享，并且团队的其他成员会提供诚实和建设性的反馈。迭代会根据需要有机地进行，并且产品只有通过内部测试获得内部吸引力后才会发布。

I encourage people to try to work in parallel as much as they can. They should not be waiting for everyone to unblock them. Ideally, you have design, front end, and back end all working at the same time on the same project. And now that we have a business team, all four people could work in parallel, whereas conventionally you might wait for designs or mock-ups to show up first.

我鼓励人们尽可能多地尝试并行工作。他们不应该等待每个人都解锁他们。理想情况下，设计、前端和后端都在同一个项目上同时工作。现在我们有了一个业务团队，所有四个人都可以并行工作，而按照惯例，您可能会等待设计或模型首先出现。

8. How do reporting lines work? 8. 汇报关系如何运作？

The teams are currently structured by function (product, R&D, design, business, etc.), and different teams think about different layers of the company and stack. But all energy is directed toward improving the core product. We design objectives that translate to common top-level metrics and improve the user experience holistically. For example, all teams share common top-level metrics while A/B testing within their layer of the stack. Because the product can shift so quickly, we want to avoid political issues where anyone’s identity is bound to any given component of the product.

目前，团队是按职能（产品、研发、设计、业务等）构建的，不同的团队考虑公司和堆栈的不同层面。但所有精力都集中在改进核心产品上。我们设计的目标可转化为通用的顶级指标并全面改善用户体验。例如，所有团队在其堆栈层内进行 A/B 测试时共享通用的顶级指标。由于产品变化如此之快，我们希望避免任何人的身份与产品的任何给定组件绑定的政治问题。

At our current size, we are flat by design, and the reporting structure does not dictate priorities as much as commitments to top-level goals. Our two full-time PMs—one web and one mobile—report to me as head of product. We’ve found that when teams don’t have a PM, team members take on the PM responsibilities, like adjusting scope, making user-facing decisions, and trusting their own taste.

就我们目前的规模而言，我们的设计是扁平化的，报告结构并不像对顶层目标的承诺那样规定优先事项。我们的两名全职产品经理（一名网络产品经理和一名移动产品经理）向我汇报，担任产品主管。我们发现，当团队没有 PM 时，团队成员就会承担 PM 的职责，例如调整范围、做出面向用户的决策以及相信自己的品味。

9. You build one of the most beloved and successful products out there. What would you say is unique or central to your approach to product that leads to such success? 9. 你打造了最受欢迎、最成功的产品之一。您认为您的产品取得如此成功的独特之处或核心是什么？

Central to our approach is to take feedback, both from users and internally, and distill it into a few intuitive products that can work for many customers. We also try to distill the feedback in a way that motivates and informs our team, setting a broad vision but letting individuals control their own decisions about what would best serve the original goal. Our decentralized approach for decisions passes the torch of responsibility, enabling fast-paced iteration without the need for approval processes. Individuals make urgent, locally optimal decisions. Any misalignments are then ironed out quickly afterward.

我们方法的核心是收集用户和内部的反馈，并将其提炼成一些可以为许多客户服务的直观产品。我们还尝试以一种激励和告知我们团队的方式提炼反馈，设定广阔的愿景，但让个人控制自己的决定，决定什么最能服务于最初的目标。我们的去中心化决策方法传递了责任的火炬，无需审批流程即可实现快节奏的迭代。个人会做出紧急的、局部最优的决策。任何不一致的地方都会很快得到解决。

10. What’s your primary tool for task management, and bug tracking? 10. 您用于任务管理和错误跟踪的主要工具是什么？

Linear. For AI products, the line between tasks, bugs, and projects becomes blurred, but we’ve found many concepts in Linear, like Leads, Triage, Sizing, etc., to be extremely important. A favorite feature of mine is auto-archiving—if a task hasn’t been mentioned in a while, chances are it’s not actually important.

线性。对于人工智能产品来说，任务、错误和项目之间的界限变得模糊，但我们发现 Linear 中的许多概念非常重要，例如线索、分类、规模调整等。我最喜欢的一个功能是自动存档——如果某项任务有一段时间没有被提及，很可能它实际上并不重要。

The primary tool we use to store sources of truth like roadmaps and milestone planning is Notion. We use Notion during development for design docs and RFCs, and afterward for documentation, postmortems, and historical records. Putting thoughts on paper (documenting chain-of-thought) leads to much clearer decision-making, and makes it easier to align async and avoid meetings.

我们用来存储路线图和里程碑规划等事实来源的主要工具是 Notion。我们在设计文档和 RFC 的开发过程中使用 Notion，然后用于文档、事后分析和历史记录。将想法写在纸上（记录思路链）可以使决策更加清晰，并且可以更轻松地协调异步并避免会议。

Unwrap.ai is a tool we’ve also recently introduced to consolidate, document, and quantify qualitative feedback. Because of the nature of AI, many issues are not always deterministic enough to classify as bugs. Unwrap groups individual pieces of feedback into more concrete themes and areas of improvement.

Unwrap.ai 是我们最近推出的一款工具，用于整合、记录和量化定性反馈。由于人工智能的性质，许多问题并不总是具有足够的确定性来归类为错误。 Unwrap 将各个反馈分组为更具体的主题和改进领域。

11. Would you say roadmap ideas primarily come top-down (teams are told what to build) or bottom-up (teams generally come up with the ideas)? 11. 您认为路线图想法主要是自上而下（团队被告知要构建什么）还是自下而上（团队通常提出想法）？

High-level objectives and directions come top-down, but a large amount of new ideas are floated bottom-up. We believe strongly that engineering and design should have ownership over ideas and details, especially for an AI product where the constraints are not known until ideas are turned into code and mock-ups. Plenty of brainstorming is going on at all times. We have a dedicated brainstorm channel in Slack, follow-up ideas are collected in Linear, and often polishes go straight to code without anyone asking.

高层目标和方向是自上而下的，但大量新想法是自下而上提出的。我们坚信，工程和设计应该对想法和细节拥有所有权，特别是对于人工智能产品来说，在想法转化为代码和模型之前，限制是不知道的。大量的头脑风暴一直在进行。我们在 Slack 中有一个专门的头脑风暴频道，在 Linear 中收集后续想法，并且通常无需任何人询问就可以直接编写代码。

The best examples of bottom-up ideas can be seen in Perplexity’s Discovery, Collection, and Sharing experiences. For example, as I shared above, our brand designer Phi builds the Discover Daily podcast and simultaneously makes decisions around the script, ElevenLabs integration, brand, and audio engineering. With AI, it’s impossible to predict use cases until iterations of the product are released. A year ago, we would never have predicted that the Discover experience would eventually be built into a podcast.

自下而上的想法的最佳例子可以在 Perplexity 的发现、收集和共享体验中看到。例如，正如我上面分享的，我们的品牌设计师 Phi 构建了 Discover Daily 播客，同时围绕脚本、ElevenLabs 集成、品牌和音频工程做出决策。对于人工智能，在产品迭代发布之前不可能预测用例。一年前，我们绝不会预料到 Discover 体验最终会被内置到播客中。

12. When people see a company like yours from the outside, it all looks perfect and like you have everything figured out. What are some things that aren’t working well or have been big challenges? 12. 当人们从外面看到像你这样的公司时，一切看起来都很完美，就像你已经把一切都搞清楚了。有哪些事情进展不顺利或遇到了巨大的挑战？

Big challenges today revolve around scaling from our current size to the next level, both on the hiring side and in execution and planning. We don’t want to lose our core identity of working in a very flat and collaborative environment. Even small decisions, like how to organize Slack and Linear, can be tough to scale. Trying to stay transparent and scale the number of channels and projects without causing notifications to explode is something we’re currently trying to figure out.

今天的重大挑战围绕着将我们当前的规模扩大到新的水平，无论是在招聘方面还是在执行和规划方面。我们不想失去在非常扁平和协作的环境中工作的核心身份。即使是很小的决定，比如如何组织 Slack 和 Linear，也可能很难扩展。我们目前正在努力解决的问题是，尝试保持透明度并扩大渠道和项目的数量，同时又不会导致通知激增。

13. What are some fun/unique rituals or traditions you have on the product team or at the company? 13. 您在产品团队或公司有哪些有趣/独特的仪式或传统？

A lot of features and products at Perplexity were built during one-week (or less) hackathons. Focused sprints to build new features have proved to be the most exciting and memorable times. Our first interactive search prototype, Pro Search (formerly Copilot), was built in a few days, but it has improved over many iterations of polish and fine-tuning.

Perplexity 的许多功能和产品都是在为期一周（或更短）的黑客马拉松中构建的。事实证明，集中冲刺来构建新功能是最令人兴奋和难忘的时刻。我们的第一个交互式搜索原型 Pro Search（以前称为 Copilot）在几天内就构建完成，但经过多次打磨和微调，它已经得到了改进。

Thank you, Johnny! Also, a big thank-you to Phi Hoang for helping with visuals.

谢谢你，约翰尼！另外，非常感谢 Phi Hoang 在视觉效果方面的帮助。

For more, check out Perplexity, and they’re hiring!

欲了解更多信息，请查看 Perplexity，他们正在招聘！_

Have a fulfilling and productive week 🙏

度过充实而富有成效的一周🙏_

I run a white-glove recruiting service specializing in product roles, working with a few select companies at a time. If you’re hiring for senior product roles, apply to work with us:

我经营一家专门从事产品角色的白手套招聘服务，一次与几家选定的公司合作。如果您正在招聘高级产品职位，请申请与我们合作：

Apply to join 申请加入

If you’re exploring new opportunities yourself, use the same button above to sign up. We’ll send over personalized opportunities from hand-selected companies if we think there’s a fit. Nobody gets your info until you allow them to, and you can leave anytime.

如果您自己正在探索新的机会，请使用上面的相同按钮进行注册。如果我们认为合适，我们将从精心挑选的公司中发送个性化机会。除非您允许，否则没有人会获取您的信息，并且您可以随时离开。

**If you’re finding this newsletter valuable, share it with a friend, and consider subscribing if you haven’t already. There are group discounts, gift options, and referral bonuses available. **

如果您发现此新闻通讯很有价值，请与朋友分享，如果您还没有订阅，请考虑订阅。有团体折扣、礼品选项和推荐奖金。

Sincerely, 真挚地，

Lenny 👋

Vandee's Digital Garden

Explorer

Perplexity 如何构建产品