
Paul Buchheit on Google, Friendfeed, and Start-ups 09/21/2009
What this covers
Paul Buchheit, the developer of Gmail and founder of FriendFeed, sits down with Russ Roberts to explore how products get built at scale and how companies lose the ability to build them well. The conversation moves across Gmail's evolution from experimental project to industrial tool, the mechanics of innovation inside large corporations, and the broader transformation of Google and Silicon Valley culture as organizations grow. Buchheit anchors his observations in concrete product histories—Gmail's JavaScript-first architecture, FriendFeed's missteps around feature perception, the specific reliability constraints that shaped email infrastructure—rather than abstraction.
The core tension Buchheit develops is between two kinds of incentive systems: startups where everyone wins or loses together, versus mature companies where individual contributions blur into collective stock price movements. He traces how large organizations accumulate what he calls "negative incentives"—the pressure to avoid looking foolish outweighing the motivation to swing for ambitious bets—and how industrial-era management methods designed for factories fail at creative work that resists measurement. He contrasts this with what he identifies as Silicon Valley's distinctive asset: a culture that respects risk-taking enough to make leaving a secure job for an unknown startup feel rational rather than reckless. The discussion widens into social technology, where Paul Buchheit argues that lowered friction for sharing and feedback shifts power to consumers and changes who participates in these systems. He closes with observations on how barriers between people and direct feedback—whether padding in running shoes or measurement systems in organizations—degrade decision-making.
Buchheit argues that building great products like Gmail comes from iterative experimentation, fast feedback loops, and incentives that reward risk-taking—and that as companies scale, industrial-era management and negative incentives erode the very culture that made them innovative.
- Gmail succeeded by iterating on real user behavior rather than following a fixed plan
- large companies accumulate negative incentives that punish bold bets
- the best signal of individual contribution comes from peers choosing who to work with, not top-down measurement
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
Every time the friction of sharing is lowered, people share more and qualitatively different things—cell phone cameras enabling spontaneous photo sharing connects distant people in ways impractical with film cameras and mailed envelopes.
“every time you lower the friction essentially to sharing people share more and more stuff”
Building Gmail entirely in front-end JavaScript demonstrated that serious web applications were possible, inspiring developers to build office suites and other rich applications in the browser and helping shift the industry away from desktop clients.
“by building what was actually a really great email experience right in the browser I think that a lot of people saw that and said wow you know the web is real we can build applications on top of this”
Silicon Valley's enduring core strength is a culture comfortable with risk-taking, where leaving a stable big-company job (like Intel) for a tiny unknown startup (like Google) is respected rather than seen as foolish—and this social acceptance of risk is decisive because it tips people on the fence toward bold choices.
“it's a place where people are very comfortable with the notion of taking risks and just trying things that may not work and that's really the heart of it”
Over time large companies accumulate negative incentives that motivate people to avoid screwing up while positive incentives disappear, so employees rationally avoid bets with a 10% chance of creating a billion-dollar business and a 90% chance of looking dumb—and their bosses won't back such gambles either.
“if there's something you could do that maybe has a 10% chance of creating a billion dollar business and a 90% chance of making you look dumb you know which one are you gonna do”
Industrial-era management methods that measure and monitor output work for factories but fail for creative product work, where individual output cannot be measured well, requiring a fundamentally different incentive and management system.
“in a factory you can usually measure and monitor everyone's output... in a more creative endeavor like creating new products all of that it's not none of that work”
Social sharing technology shifts power to consumers and forces companies to actually make customers happy rather than relying on advertising spend and customer ignorance, because dissatisfied customers can now broadcast complaints to thousands instantly (e.g., Yelp reviews prompting direct apologies from business owners).
“the entire way that I think the consumer economy works is changing because previously it was all about just telling people what they needed to think... now you actually have to make your customers happy”
As communication moves from ephemeral media (face-to-face) to recorded media (email, IM), offhand comments become permanent evidence, which both penalizes casual speech and creates perverse incentives for companies to destroy historical records via retention policies.
“as a lot of our communication moves from kind of ephemeral media medium... but it's all of a sudden I started doing it online over IM or email then suddenly these just offhand conversations could become evidence in a trial”
A culture of real identities and real relationships (as in Facebook) increases the trustworthiness of online information because reviews come from known friends rather than anonymous strangers, and it makes fake reviews harder to pull off since friend-of-owner relationships become visible.
“because it has a culture of real identities and real relationships you start to actually see the truth... instead of just being a review from someone... it's from my friend”
Targeted advertising in Gmail is economically efficient because directed ads deliver interested customers with a high payoff per conversion, and the system was optimized to keep operational costs low enough that it need not generate much revenue per user.
“the nature of that kind of directed advertising is that the payoff is pretty good I think if you actually deliver an interested customer”
Automatically grouping email into conversations by thread eliminates the wasteful manual labor of reassembling discussions and lets users catch up on a 30-message thread as a single inbox line, making the experience far more efficient.
“by just putting it all together it's a it's a much more efficient experience I can just read the whole conversation all at once”
Company values and mottos must be short and memorable to actually shape behavior, because forgettable, perfectly-crafted value statements that capture everything still don't matter if no one remembers them.
“you can create some perfectly crafted statement that has all of your values... but if it's forgettable it doesn't matter”
A highly productive individual can still be a net negative to a team if they cause everyone else to quit, meaning individual output is not the only thing that matters for team output.
“even if someone is very productive if they end up causing everyone else to quit they're probably in that negative”
In a startup everyone's incentives are aligned around collective success or failure, but as a company grows, the stock price becomes a poor proxy for an individual's contribution because it is driven by a few key markets (like ads), leaving most employees with little direct impact.
“the stock prices it's a very good proxy for anything because you could be creating some cool new product but you know really the stock price is going to be driven by like something in the ad market”
A market leader with a strong lead that stays focused and continually innovates is extremely difficult to overcome, so competitors are better off pursuing lower-hanging fruit elsewhere rather than attacking the incumbent's franchise.
“if the company has a lead the way Google has and they remain focused on it and they're able to continually innovate I think that's tremendously difficult to overcome that lead”
Email infrastructure requires fundamentally different reliability assumptions than web search—losing a crawled web page is recoverable, but an email system can never lose data, and this constraint propagates down to the lowest-level technical decisions in the stack.
“when you're crawling the web if something goes wrong and you lose a few web pages it isn't really a big problem because you can just go fetch them again later but when you're building an email system obviously you can't ever lose anything”
Storing personal email with Google is safer than the common alternatives because Google has strong business incentives to protect data and employs world-class security people; the greater privacy threat is legal subpoena, which applies to any system including self-hosted email.
“I think Google has a lot of very strong incentives to be careful with the data you know if they did something bad it causes a very very severe you know a lot of damage to their business”
The padding in modern running shoes isolates the feet from environmental feedback, causing people to land on their heels and run in a damaging style; running with minimal insulation gives the body better feedback, encourages forefoot landing, and allows running long distances without injury.
“by actually having less insulation you your body gets better feedback from the environment you run in a different style where you're not landing on your heel”
The 'don't be evil' value is valuable beyond being a catchy phrase because it gives every employee license to stop and question decisions, creating a more thoughtful organization where the option to question is better than quitting in protest.
“it actually gives everyone in the company license to question the decisions”
Many email users maintain two systems—an aspirational elaborate folder hierarchy they think they should use, and the reality of leaving everything in their inbox because they fall behind—which most existing mail systems failed to support.
“a lot of people had essentially two different systems they had a system they wanted to be using and their system actually using today”
Competitors often appear to 'catch up' to a market leader by matching what the leader had years earlier, while the leader has already advanced privately—a perception problem illustrated by Yahoo's Panama ad project matching Google's capabilities from three years prior.
“what they were coming out with was something that was equivalent to what Google has you know three years prior and so it was the kind of hopeless endeavor”
The most valuable information about who is contributing on a team comes from team members themselves choosing who they want to work with—a peer-selection signal that beats top-down measurement and captures subtle qualities like complementary skills and who compromises at the right times.
“the people on the team know who the really good people are and who's really contributing and who's not and that's the most valuable information”
FriendFeed was conceived around the concept of sharing, not aggregation; the auto-import features were meant to surface content already shared elsewhere, but they backfired by making people perceive FriendFeed as merely an aggregator—illustrating how a feature can hurt a product by changing how it is perceived.
“the core concept that it was based on is sharing... but the downside of that is that it creates this notion that well it's just this thing that aggregates”
Lowering the friction of feedback changes who provides it—shifting from extreme personalities or 'cranks' with lots of time to ordinary people doing it as part of everyday life—which changes the nature and representativeness of the feedback businesses receive.
“if you're writing a letter to a restaurant like that's just not the kind of thing that I think most people would do... now a lot of these things are so easy that these normal people are doing it”
Putting up too many barriers between ourselves and reality—whether in running shoes or in organizations—leads to bad decisions because we stop getting the right feedback.
“the dangers of insulating ourselves from reality too much where... we end up making bad decisions because we're not really getting the right feedback”
FriendFeed and Twitter were on a convergent evolutionary path: Twitter began as status updates but users evolved it toward link-sharing, while FriendFeed added richer conversational features—causing FriendFeed to fall into the same mental 'bucket' as the much larger Twitter, which was a competitive disadvantage.
“Twitter started out just being about status updates... but over time the users of Twitter have kind of evolved it to be a lot of other things”
The right approach to email for people who don't enjoy organizing is to abandon manual sorting, archive messages, and rely on search to retrieve them later, since there is no single 'right way' to do email and it largely comes down to personality.
“instead of having to spend a lot of time sorting my mail upfront I could just search for it later on if I needed to find it”
Within Google, Gmail was initially seen as an unnecessary distraction because the company identified as a search-technology provider powering Yahoo and others, and a consumer email product threatened those partnerships and didn't fit the self-conception.
“a lot of people questioned why we were even building Gmail they thought it was a bad idea you know an unnecessary distraction”
FriendFeed's key difference from Twitter was that its comment threads make participants aware of each other—creating a discussion like a group in a room rather than independent replies—which fundamentally changed the social dynamic of the service.
“the comments are all aware of each other so it's more like a bunch of people in a room talking together versus you know you just kind of hear back independent messages”
Facebook has the potential to be extremely influential in the way Google has been—not just commercially but in setting the standard for Silicon Valley.
“Facebook is a company that has the potential to be extremely influential as much in the same way that Google has been”
Gmail was developed over roughly two years of iterative dogfooding, starting with Buchheit alone searching his own email in late 2001 and growing to a core team of about a dozen by its April 1, 2004 invite-only launch.
“the initial launch was April 1st 2004”
Buchheit was Google employee number 23, joining in June 1999 when the search market was crowded and dominated by AltaVista, and he initially had low expectations that the little startup could survive against the incumbent.
“I had fairly low expectations to be honest you know the search market was pretty crowded back in 99”
Gmail invitations were so scarce in the early invite-only period that people sold them on eBay for as much as hundreds of dollars.
“people were actually selling them on ebay for in some cases I think even hundreds of dollars”
Buchheit took roughly a year to a year and a half off after leaving Google before co-founding FriendFeed, a break prompted partly by personal events including his daughter's medical issues and his brother's death from cancer.
“I took off for maybe about a year and a half or a year”
Facebook acquired FriendFeed, with the team staying largely intact to bring lessons from building FriendFeed into Facebook, because Facebook and FriendFeed were on a convergent path toward the same kind of product.
“FriendFeed and Facebook are presently very different products we had a lot of the same ideas about the kind of product that we ultimately wanted to build”
FriendFeed lowered the barrier to entry by letting users sign in with one click via Twitter or Facebook to pull in their existing contacts and content, so the experience varies dramatically depending on how active a user's friends are on the service.
“there's just a one-click button on the homepage if you want you can click on Twitter or Facebook and it'll just pull in your information from there”
Concern about Gmail outages is often overstated; in practice an email outage every couple of months is comparable to or less frequent than residential power loss, and rarely causes day-to-day problems.
“I lose electricity more often than I lose access to Gmail”