为感受而做，不为曲线：Arc 浏览器背后的产品哲学

Miller admits that The Browser Company still struggles to describe Arc in one sentence. Objectively it is a replacement for your default web browser, and people who try it find it hard to go back to the old way of using the internet. He says sharpening that explanation is on the to-do list, and he comes back to the problem at the end of the interview.

For metrics, the company obsesses over a single number it calls D5/D7 (some companies say L5/L7): how many people turn to Arc at least five days a week. Miller likes it because it bundles retention, engagement and growth, and you cannot game it by opening the app accidentally once a week. He adds that the team ignores absolute numbers and watches week-over-week growth, which he says has been above 10% for roughly eight months, though he does not expect that to last.

On retention, he says D5/D7 sits somewhere between the low 30s and low 40s depending on the cohort. What he is proudest of is that the curve has been inching up over twelve months even as new users are further from the earliest, most passionate adopters. Cohort-over-cohort improvement, not a single snapshot, is what the team looks at.

Miller says his approach is a reaction to the way modern Silicon Valley, as he saw it at Facebook around 2014, became obsessed with graphs and metrics. He acknowledges that numbers are an effective, objective way to reach outcomes, but argues that optimizing for them "leaves a lot on the table." At The Browser Company the question for any feature, project or piece of storytelling is what feeling it should evoke: joy, speed, organization, focus.

He points to Nike, Disney and Apple as brands that first won him over as a child, and asks what Walt Disney, Phil Knight or Steve Jobs were optimizing for at the moment of creation. His Facebook example is the way the company discussed Snapchat's rise through a share-count metric, while Snapchat itself was thinking about something more human. A better question, he says, would have been whether people feel closer to friends and family.

Asked how this works in practice, Miller says that if you pick the right feeling it tends to track the business metric anyway. Features meant to surprise people get screenshots dropped into Slack and recommendations to friends, which is why the company has no growth team. He is careful to say that both extremes are flawed and that the answer depends on the project, but he wants to pull a metrics-heavy industry back a little.

The process itself is informal. For the Peek feature, which lets you pop open a link preview without leaving the page you are on, the team talked about making it feel light, airy and effortless for someone skimming five to seven links on Hacker News. There is no product requirements document listing emotions, just a conversation among a tight group of people who have worked together for a long time.

Asked how the team ships quickly and closes the loop with users, Miller says it comes down to values. The first, which he considers the most important, is hiring people who show up with "heartfelt intensity": some fire in the belly that might be UI craft for one person and doubling performance with a quarter of the headcount for another. Give such a team an ambitious prompt and get out of the way, he says, and they do remarkable work.

Other values include "assume you don't know," which applies even to an engineer who built the first version of Chrome and ran it for 16 years, and a bias to action that follows from it. Another is "start by asking what could be." Peek, for example, was meant to blur the line between native and web software, not just fix context switching.

Miller says he personally hates corporate values and resisted them for the first couple of years, despite pressure from experienced advisors. He changed his mind after noticing in one-year check-ins that the team kept describing the same traits. The company then interviewed everyone, one by one, and almost every word in the final list came from a team member.

Rather than a landing page of five titles and subtitles, they wrote an essay called "Notes on Road Trips," a semi-biographical, mostly fictional story of a teenager and their father, inspired by how Miller likes to travel. It took about three months, and he credits a teammate named Abby with doing much of the work.

Responding to the most popular listener question, about how the company attracts remarkable people, Miller deflects and suggests asking the employees. He then reveals that he never cared about web browsers. He and co-founder Hursh decided long before the company began that they wanted to build somewhere they could hire anyone they admired, and that the best memories of their decade together were about groups of people, not products.

That is why he calls the team the product. Interviews are meant to talk candidates out of joining: he does not pitch and tells them most people shouldn't want to work there. Recent hires include Darren, who co-created the first Chrome prototype; Tara, a former Vimeo executive who worked on Paper by FiftyThree; and Peter Vidani, Tumblr's first designer and later Slack's SVP of design. All joined as individual contributors.

Miller believes such people come because they want this work to define their careers. In return they need an ambitious aspiration, real empowerment, trust and a kind culture, since people with a decade or more of experience have little tolerance for nonsense. Lenny adds that great hires attract more great hires, the way a clear mission makes recruiting easier.

A practical lever is public celebration: the company announced its very first hire as if it were a product launch and publicly credits the people behind every shipped feature. Miller points out that he did not make Arc or Peek and dislikes the CEO-hero worship he sees in Silicon Valley. Tagging the animator who made something fun is a small, authentic gesture that reinforces the cycle.

The company films board meetings and internal discussions about delayed projects. Miller says the practice is a prototype that he rates high on the list of things the team could come to regret. His worry is that public storytelling pulls the CEO to the center, and he admits that after watching the team's best-performing YouTube video, a reaction to MKBHD discussing Arc, he asked himself why he was trying to become a YouTuber.

The motivation is trust. Miller, who studied sociology, says people have culturally lost trust in tech companies, and a browser holds your most sensitive personal and professional data. His hypothesis is that even the best privacy policy is ultimately just people behind a screen, so if users get to know the team, imperfections included, they will be more willing to believe the company is acting in good faith.

He calls the goal "radical trust building" rather than radical transparency, since not everything can be shared. He says the risk of looking like internet celebrities is worth taking, and invites listeners to tell him if the company goes too far. He also says Lenny's podcast is an inspiration, because listening to someone for a year creates a sense of trust in a person you have never met.

Miller describes a culture driven by prototypes, including how the company is organized. A membership team replaces customer support, success and user research: its job is to maintain a genuine relationship with a person from the moment they touch the product. He says he took inspiration from Facebook's user researchers, but felt that research there often acted like a service organization for PMs.

The storytelling team covers people who do not yet use Arc, which could mean investors, press or the wider world, handling what others would call investor relations, PR and marketing. Miller says that giving something a new name sheds preconceived notions. Even the idea of calling a feature "browser history" brings expectations about what it has to be, so invented names make people think from first principles.

The company has no PM organization. Project leaders are chosen for the project, so an infrastructure engineer might lead a performance effort and someone from membership might lead a project aimed at serving members. Miller says that two or three people could hold a PM title elsewhere, and he likes to hire "mutts," such as Rebecca, an MIT-trained data lead who effectively PMs the multiplayer work.

Lenny predicts that, as at other companies, designers and engineers will eventually tire of doing PM work and want a specialist, and that great PMs make everything better. Miller agrees it will probably happen, noting that Snapchat once said it would never hire PMs and now has many. The team already feels the "verbs" crowding out design time, and he expects to come back and admit Lenny was right if it all blows up.

Lenny compares The Browser Company's storytelling team to Airbnb's early video team, which many insiders joked accounted for half of its market cap. Miller describes his team of three: Nash, who leads it and has a background that includes NBC and writing poetry; Ellis, a former Verge reporter who ran marketing strategy at Snapchat for seven years; and Josh Lee, a former designer turned indie filmmaker who edits the videos.

Miller says he saw a tendency toward homogeneity at Facebook and has purposely hired very different archetypes on storytelling and design. That can be hard, since it is easier to hire a uniform type of designer or storyteller, but he finds the diversity worth it. He says large companies tend to drift toward a single "our way or no way" model.

The Airbnb thread leads to the Snow White storyboards, in which a Pixar storyboard artist drew the journey of a host and guest. Miller says seeing them on the office walls early in his career influenced his thinking about how someone should feel opening a door for the first time. He stresses that the team's beliefs build on lessons absorbed from others, and hopes that The Browser Company can become an inspiration in the way the documentary General Magic still inspires him.

Lenny asks about a cohort of products, including Linear, Raycast and Cron, that seem obsessed with experience over revenue. Miller, a former investor at Thrive Capital, says different businesses depend on different attributes, and that he and Hursh chose the browser category deliberately. In his view the browser is among the most universal consumer products, used by parents, nieces and nephews alike, and one of few pieces of software in the middle of that Venn diagram.

The second attribute is that browsers are commodities: increasingly copies of Chromium with small tweaks. The third is that they are extremely lucrative, with high margins and low marginal costs, which is why Google, Apple and Microsoft own them. Put together, he says, you must win on how much people love your product, almost emotionally rather than rationally, which is why surprise and delight are a practical, even capitalistic, strategy.

Miller cautions that this does not apply to every category. Nobody selling cybersecurity software to government agencies wins on rounded corners. He gives Tuple, a pair-programming tool, as an example of balance: its design is lovely, but it also has the best audio quality he has ever experienced in a communication tool, and he says remote-work tools with good design but poor latency lost out.

Miller explains the company's name as something of a misdirection: Arc replaces your default browser, but the larger goal is an "internet computer." He says he has repeatedly underestimated big shifts in technology, such as mobile, and that today's underestimated shift is to cloud computing. Applications are web-based, files are URLs, and photos live on a server, so if you smashed your device you would lose little beyond the hardware.

If hardware becomes commoditized empty shells, he asks, what will his two-year-old son's computer be at ten? The company's answer is to be to the web browser what the iPhone was to the cell phone. What Arc builds now is the equivalent of multi-touch gestures and a first-party notes app, but he says the developer platform on top is the more interesting and lucrative opportunity.

He argues the web is the best development platform: it works on every device, is free for users and developers, has no 30% tax and lets the best frameworks win. The gap is that native software feels better, and the companies controlling browsers have perverse incentives. Apple wants developers on its app store, and Google wants content indexed in search, which is why Miller says Chrome extensions are underused.

In the closing lightning round, Miller recommends the books Harold and the Purple Crayon and Seeing Is Forgetting the Name of the Thing One Sees, praises episode three of The Last of Us, and suggests Tuple as his favorite tool. His interview question is "What would you like to ask me?", and he looks for authenticity. His top Arc tip is not to think of it as a browser, and he says the team must get much better at explaining what it built.