Josh Miller
CEO and co-founder, The Browser Company
Where to hear and read about Josh Miller1 episode or test
Josh Miller is the CEO and co-founder of The Browser Company, maker of the Arc browser, and previously worked at Facebook and Thrive Capital. The thing most worth learning from him is how he wins a commodity category by building for how people feel, with a team of intense makers, and uses one honest metric to check the work.
I view our product as our team
Competing with giants: An inside look at how The Browser Company builds product | Josh Miller (CEO)
Josh Miller: the story and the lessons
1Josh Miller's path to Arc and why he chose browsers
Miller learned product at Facebook after selling his first company there. He and co-founder Hursh started a company that Facebook acquired, and he spent years there around 2014. It was his first time as a PM, and he calls Facebook one of the best-executing organizations he has seen.
Numbers drove every debate. When Snapchat rose, the question became how often people post each week, which later shaped his push to start from feelings.
He spent two years at Thrive Capital studying what makes ideas win. Thrive invested in Slack and GitHub. He came away believing different ideas carry different attributes that decide whether they succeed.
He and Hursh were deliberate about what would need to matter for their own company before choosing it.
Miller started the company to work with favorite people, not to build a browser. He says he never cared about web browsers. He and Hursh, friends for a decade, wanted to build something that would attract the best minds, because their best memories were of groups of people at a moment in time.
He treats the team as the product and picked a problem ambitious enough to draw people he admired.
He picked browsers because they are interchangeable and very profitable. Most browsers are near copies of Chromium with small tweaks, and Google, Apple and Microsoft own them because margins are high. If people can be made to love one, the money follows.
He says this logic does not fit every category. Cybersecurity software sold to governments would not care about rounded edges or button colors.
Miller bets that computing is moving onto the internet. Files, apps and photos already live online, so he wants Arc to be an internet computer and a developer platform, the iPhone to the web browser.
He says he underestimated mobile at his first company and thinks cloud computing is the shift being underestimated now. He expects the platform side to matter most within a decade, if Arc still exists.
He knows a challenger must be far better than the default. If Arc were only 20% better than Chrome, people would stay put because they do not want to learn something new.
He calls that a high bar for consumer products and points to Tuple, whose audio beats Zoom's, as a product that cleared it.
2How Miller builds for feelings and tracks one metric
Miller names the feeling he wants before the team builds. Each project starts by asking how the person on the other end should feel. For Peek, which pops a link up from Hacker News without leaving the page, the targets were lightness, airiness, speed and agility.
This happens in conversation, not in a PRD. He admits it is not formally operationalized and will need to be as the team grows.
He picks feelings that track the business metric. Arc has no growth team and no growth projects. Surprise and joy make people send screenshots to friends, and that drives word of mouth.
"If you pick the right feeling it typically tracks pretty closely with the metric you care about."
Miller uses numbers to stay honest, not to decide what to make. At Facebook the Snapchat debate turned on posts per week, when the better question was whether people feel closer to friends and family.
He admits that optimizing purely for feelings is flawed too. He pushes it because Silicon Valley has leaned so far toward graphs.
The team obsesses over one number: people active five days a week. D5/D7 counts a person only if they open a tab on each active day. It captures retention, engagement and growth together and is hard to game. Arc grew over 10% a week on it for about eight months.
The team does not track daily or weekly active retention at all. He does not expect the 10% weekly growth to last another eight months.
He judges retention by whether each cohort beats the last. D5/D7 retention sits between the low 30s and low 40s depending on cohort. Miller is proudest that it keeps improving even as new users are less passionate than the earliest adopters.
Because success would make any single point in time look small, the team watches the trend and week-over-week growth rate more than the level.
3How Miller hires, sets values and builds trust
Miller hires for heartfelt intensity, then gets out of the way. Intensity looks different in each person, from UI craft to doubling performance with a quarter of the engineers. His recipe is an ambitious promise and then no interference.
The values "assume you don't know" and "start by asking what could be" push people to act. He began removing himself from product development that January and calls it healthy.
Proven senior people join Miller's team as individual contributors. The co-creator of Chrome, who ran it for 16 years, joined, along with Tara, a former Vimeo senior product leader, and Peter Vidani, Tumblr's first designer and a former Slack SVP of design. Each wanted one more shot at career-defining work.
Once the first great people joined, more followed. He calls this the tumbleweed.
He uses interviews to talk candidates out of joining. He opens with "ask me anything" and says most people should not want to work there. What candidates ask and how they follow up shows what they care about.
He screens mainly for authenticity, meaning whether people really believe what they say.
Miller gives credit to the makers, not to himself. New hires are announced like product launches. When someone tweets about an animation, he tags the designer who built it.
He says he does little of what people fall in love with. Praise reaching the person who did the work tells new hires they will be recognized.
He wrote the values from the team's own words after distrusting them. He saw values as corporate propaganda and resisted for the first one to two years. Then one-year check-ins kept surfacing the same traits, so the team interviewed everyone and turned the words into the essay "Notes on Road Trips".
It took about three months and colleague Abby did much of the work. Nearly every value word came from someone on the team. He calls his early stance naive and hard-headed.
Miller builds in public to earn trust, and worries about it. A browser holds sensitive data, so the team shows its meetings and people in videos. He calls it radical trust building, not radical transparency, because not everything can be shared.
Their best-performing YouTube video was his reaction to MKBHD on Arc, which made him ask if he was becoming a YouTuber. He asks listeners to tell him if it goes too far.
4The odd teams Miller runs as prototypes
Miller runs a membership team that owns the whole user relationship. It combines support, success and research from the first touch of the product. Day 1 is how to use it, day 37 is a bug, day 58 is what people want on mobile.
The idea came from seeing Facebook treat user research as a service assigned to a PM. He took knowing your users deeply to the extreme.
A single storytelling team covers everyone who is not yet a user. It handles the work normally split into investor relations, PR and marketing. Its three members are Nash, a poet-writer lead, Ellis, a former Verge reporter who ran Snapchat marketing strategy for seven years, and Josh Lee, a filmmaker who edits video.
He hires different archetypes on purpose after watching Facebook drift toward sameness. The cost is that it is harder to say "we hire designers like this".
Miller picks the project leader per project instead of running a PM org. A performance project is led by an infrastructure engineer, and a project about serving members by someone from membership. About three people could hold a PM title elsewhere.
"Leaders of the project should be picked for the project not because they're in a certain org." He admits it breaks "in 18 ways" and notes Snapchat swore off PMs and later hired them.
He hires multidisciplinary "mutts" who say "I make things". Rebecca, a former Quora data scientist with an MIT PhD and Stripe engineering experience, runs data and plays the PM role on the multiplayer team.
He treats PM duties as verbs anyone could do and keeps data inside a feelings-first company by hiring humane people.
Miller uses made-up team names to force first-principles thinking. A new name sheds assumptions that come with a familiar one like "browser history feature". He calls it a rhetorical tactic, not novelty for its own sake.
Airbnb called PMs producers for about a year and a half. If they end up called PMs, he says that is fine, since the change is reversible.
He expects to be wrong about PMs and says so. He agrees PMs will almost certainly arrive, since PM work crowds out designers' time to think and design. He would likely grow the data team into a technical PM role.
He calls this the idea he will most likely "come crawling back" to admit he got wrong, and he has no idea how any of this scales to 500 people.
Competing with giants: An inside look at how The Browser Company builds product | Josh Miller (CEO)
Josh Miller explains how The Browser Company builds Arc by optimizing for feelings over metrics, hiring for heartfelt intensity, and running odd, prototype-driven teams. The result is a browser with 10%+ weekly growth for about eight months and D5/D7 retention in the low 30s to low 40s that keeps improving cohort by cohort.
Miller learned product at Facebook after selling his first company there. He and co-founder Hursh started a company that Facebook acquired, and he spent years there around 2014. It was his first time as a PM, and he calls Facebook one of the best-executing organizations he has seen.
Numbers drove every debate. When Snapchat rose, the question became how often people post each week, which later shaped his push to start from feelings.
He spent two years at Thrive Capital studying what makes ideas win. Thrive invested in Slack and GitHub. He came away believing different ideas carry different attributes that decide whether they succeed.
He and Hursh were deliberate about what would need to matter for their own company before choosing it.
Miller started the company to work with favorite people, not to build a browser. He says he never cared about web browsers. He and Hursh, friends for a decade, wanted to build something that would attract the best minds, because their best memories were of groups of people at a moment in time.
He treats the team as the product and picked a problem ambitious enough to draw people he admired.
He picked browsers because they are interchangeable and very profitable. Most browsers are near copies of Chromium with small tweaks, and Google, Apple and Microsoft own them because margins are high. If people can be made to love one, the money follows.
He says this logic does not fit every category. Cybersecurity software sold to governments would not care about rounded edges or button colors.
Miller bets that computing is moving onto the internet. Files, apps and photos already live online, so he wants Arc to be an internet computer and a developer platform, the iPhone to the web browser.
He says he underestimated mobile at his first company and thinks cloud computing is the shift being underestimated now. He expects the platform side to matter most within a decade, if Arc still exists.
He knows a challenger must be far better than the default. If Arc were only 20% better than Chrome, people would stay put because they do not want to learn something new.
He calls that a high bar for consumer products and points to Tuple, whose audio beats Zoom's, as a product that cleared it.
Miller names the feeling he wants before the team builds. Each project starts by asking how the person on the other end should feel. For Peek, which pops a link up from Hacker News without leaving the page, the targets were lightness, airiness, speed and agility.
This happens in conversation, not in a PRD. He admits it is not formally operationalized and will need to be as the team grows.
He picks feelings that track the business metric. Arc has no growth team and no growth projects. Surprise and joy make people send screenshots to friends, and that drives word of mouth.
"If you pick the right feeling it typically tracks pretty closely with the metric you care about."
Miller uses numbers to stay honest, not to decide what to make. At Facebook the Snapchat debate turned on posts per week, when the better question was whether people feel closer to friends and family.
He admits that optimizing purely for feelings is flawed too. He pushes it because Silicon Valley has leaned so far toward graphs.
The team obsesses over one number: people active five days a week. D5/D7 counts a person only if they open a tab on each active day. It captures retention, engagement and growth together and is hard to game. Arc grew over 10% a week on it for about eight months.
The team does not track daily or weekly active retention at all. He does not expect the 10% weekly growth to last another eight months.
He judges retention by whether each cohort beats the last. D5/D7 retention sits between the low 30s and low 40s depending on cohort. Miller is proudest that it keeps improving even as new users are less passionate than the earliest adopters.
Because success would make any single point in time look small, the team watches the trend and week-over-week growth rate more than the level.
Miller hires for heartfelt intensity, then gets out of the way. Intensity looks different in each person, from UI craft to doubling performance with a quarter of the engineers. His recipe is an ambitious promise and then no interference.
The values "assume you don't know" and "start by asking what could be" push people to act. He began removing himself from product development that January and calls it healthy.
Proven senior people join Miller's team as individual contributors. The co-creator of Chrome, who ran it for 16 years, joined, along with Tara, a former Vimeo senior product leader, and Peter Vidani, Tumblr's first designer and a former Slack SVP of design. Each wanted one more shot at career-defining work.
Once the first great people joined, more followed. He calls this the tumbleweed.
He uses interviews to talk candidates out of joining. He opens with "ask me anything" and says most people should not want to work there. What candidates ask and how they follow up shows what they care about.
He screens mainly for authenticity, meaning whether people really believe what they say.
Miller gives credit to the makers, not to himself. New hires are announced like product launches. When someone tweets about an animation, he tags the designer who built it.
He says he does little of what people fall in love with. Praise reaching the person who did the work tells new hires they will be recognized.
He wrote the values from the team's own words after distrusting them. He saw values as corporate propaganda and resisted for the first one to two years. Then one-year check-ins kept surfacing the same traits, so the team interviewed everyone and turned the words into the essay "Notes on Road Trips".
It took about three months and colleague Abby did much of the work. Nearly every value word came from someone on the team. He calls his early stance naive and hard-headed.
Miller builds in public to earn trust, and worries about it. A browser holds sensitive data, so the team shows its meetings and people in videos. He calls it radical trust building, not radical transparency, because not everything can be shared.
Their best-performing YouTube video was his reaction to MKBHD on Arc, which made him ask if he was becoming a YouTuber. He asks listeners to tell him if it goes too far.
Miller runs a membership team that owns the whole user relationship. It combines support, success and research from the first touch of the product. Day 1 is how to use it, day 37 is a bug, day 58 is what people want on mobile.
The idea came from seeing Facebook treat user research as a service assigned to a PM. He took knowing your users deeply to the extreme.
A single storytelling team covers everyone who is not yet a user. It handles the work normally split into investor relations, PR and marketing. Its three members are Nash, a poet-writer lead, Ellis, a former Verge reporter who ran Snapchat marketing strategy for seven years, and Josh Lee, a filmmaker who edits video.
He hires different archetypes on purpose after watching Facebook drift toward sameness. The cost is that it is harder to say "we hire designers like this".
Miller picks the project leader per project instead of running a PM org. A performance project is led by an infrastructure engineer, and a project about serving members by someone from membership. About three people could hold a PM title elsewhere.
"Leaders of the project should be picked for the project not because they're in a certain org." He admits it breaks "in 18 ways" and notes Snapchat swore off PMs and later hired them.
He hires multidisciplinary "mutts" who say "I make things". Rebecca, a former Quora data scientist with an MIT PhD and Stripe engineering experience, runs data and plays the PM role on the multiplayer team.
He treats PM duties as verbs anyone could do and keeps data inside a feelings-first company by hiring humane people.
Miller uses made-up team names to force first-principles thinking. A new name sheds assumptions that come with a familiar one like "browser history feature". He calls it a rhetorical tactic, not novelty for its own sake.
Airbnb called PMs producers for about a year and a half. If they end up called PMs, he says that is fine, since the change is reversible.
He expects to be wrong about PMs and says so. He agrees PMs will almost certainly arrive, since PM work crowds out designers' time to think and design. He would likely grow the data team into a technical PM role.
He calls this the idea he will most likely "come crawling back" to admit he got wrong, and he has no idea how any of this scales to 500 people.