圣牛就是现成的路线图：Kayvon Beykpour 如何重振 Twitter 的产品文化

Beykpour first met Elon Musk over FaceTime in the chaotic first days after the takeover, when Scott Belsky recommended him as someone Musk should talk to. A two-hour in-person meeting at Twitter headquarters followed. He was brought in through the back door to avoid rumors, since he had been fired, and Walter Isaacson sat silently at the end of the room throughout.

Beykpour says they never discussed running the company or the layoffs. Instead he shared who he thought were exceptional people at Twitter and brainstormed products. He says he is glad that most of those people are still there and that projects he cared about, including Birdwatch (now Community Notes), Spaces, Communities and creator monetization, kept getting investment.

He found it odd and exciting to see another person with big dreams for Twitter, now able to "do whatever the hell I want" as the owner. Musk even invited him to "just hang out" with a swipe-left or swipe-right choice. Beykpour decided it was time to move on, and it was now Musk's turn to shape the product.

Beykpour explains that when Jack Dorsey resigned and Parag Agrawal became CEO, Agrawal quickly fixed what Beykpour considered the company's biggest problem: its functional structure. Heads of consumer product, revenue product, engineering, design and research had to reach consensus, and without a strong tie-breaker, deadlock slowed everything. He argues that a functional organization needs a GM or CEO who resolves conflicts, and Dorsey did not operate that way.

Agrawal moved to a GM structure and promoted Beykpour to GM of consumer, about a month before his paternity leave. While Beykpour was on leave, Musk joined the board, left it and made an offer. The day after Beykpour's family came home from the hospital with their newborn daughter, Agrawal called to let him go, saying Twitter needed different skills from his. That night, Twitter signed a term sheet with Musk.

Beykpour says it took time to come to terms with, especially since he had been optimistic after the reorganization. He sees a silver lining in spending the first year of his daughter's life with his family and avoiding the deal-on, deal-off drama. Lenny adds that a strongly opinionated founder would have put any product leader in an awkward spot, and Beykpour agrees that it was no longer his canvas.

Beykpour calls his first year as head of product one of the hardest of his career, less because of the work than because of the politics and consensus-building. He credits earlier experience as a teenage executive at Blackboard, a public company that had acquired his first startup, for his patience. He says the change became addictive once alignment grew and execution showed results, but that it cannot happen without alignment from the top.

For his first two years the stated strategy was "refine the core", and that focus returned Twitter to user growth through a ranked timeline and better recommendations and notifications. Beykpour says this success calcified the organization's reluctance to take risks. Users were telling the team the product had not changed in eight years, and he wanted to fix the lack of ambition without abandoning what worked.

His method was to treat sacred cows as a built-in roadmap: ranked feeds, the 140-character limit, and the idea that users could not control replies to their own tweets. The "hide replies" feature shows how deep the resistance went. An engineer told a colleague not to work on it because it would be bad for their career, a reaction Beykpour says made his blood boil.

Beykpour stresses that optimizing the core kept paying off, such as when instrumenting onboarding revealed an SMS verification bug that blocked sign-ups in countries like the UAE. His goal was a portfolio in which some bets were low-risk growth work and others added new capabilities. He admits that Twitter never quite got that balance right during his tenure.

On top of momentum and trust with executives, he says repeated storytelling about the vision and why the bets were being made was essential, both internally and externally to users and advertisers. Leaders, he says, must quickly identify who is "on the wagon" or off it. Twitter had neither the structure nor the fortitude to move off-the-wagon people out, which made the change slower, and he sees Musk as the opposite extreme.

Beykpour says small acquisitions were an effective way to speed up culture change. Founder types bring urgency and ambition and can still work in a large company. Birdwatch, later Community Notes, was led by Keith Coleman, whose company Twitter had bought. Esther Crawford led Super Follows and tipping, Fleets was run by Mo Oladam, and Communities came from the team from Chroma Labs. Each team got a little silo, and leadership's job was to keep the bet from being suffocated.

He has benefited from this approach himself, since Periscope's acquisition led to him running product. He believes that betting on people and throwing them into the deep end is one of the best ways to drive change, though buying companies is expensive. Community Notes was structured like a seed investment, with the team told not to worry about OKRs.

A pet peeve is staffing projects by availability instead of belief. Under a functional structure, some teams did not even agree that a project like hide replies was worth trying. Beykpour says speculative work should be staffed with people who have the skills and, more importantly, an obsession with the idea, because they will work harder and be more creative.

Beykpour says he was not a fan of how Twitter used Jobs-to-be-Done, which he found exhausting and unhelpful, and that he was charged with defending it. His critique is that every framework, pushed to its limit and followed religiously, loses nuance. The most charitable view is that it makes you see needs and true alternatives through the customer's eyes, which is just good product thinking.

OKRs can fail the same way. He cites Amazon's order confirmation emails that hide order details as a customer-hostile choice that may serve metrics. At Twitter, a focus on DAU led to a toggle that sent users to a reverse-chronological timeline and then returned them to the ranked feed after about a day because that kept people in the app longer.

He gives two signs that a framework has gone too far: subjectively bad decisions that you would not accept as a user, and bets the organization is disincentivized to pursue. Spaces, for example, needed a live bar at the top of the app that pushed down tweets and ads, hurting DAU in the short term. The answer, he says, is judgment and product taste, along with willingness to adjust the framework.

Beykpour tells the story of onboarding every beta user personally, including Kobe Bryant. Kobe commented, "Why the fuck would anyone want to watch someone else stream live?" and, before Beykpour could answer, added that he was joking and thought the app was incredible. Beykpour says he remembers the moment well.

He lists several reasons the app shut down. Growth surged market by market in the US, France, Turkey and the Middle East, but underneath it retention was poor, and the team shipped changes that made it worse. Integration with Twitter took far too long because Twitter was busy with its own roadmap. He also does not believe a generalized, live-only consumer app can be durable, unlike products such as Twitch or Whatnot that focus on a vertical.

Facebook, as he tells it, put hundreds of people on live video, copied the features and paid prominent streamers to stream exclusively. Beykpour sees a pattern in which Twitter spotted Vine, Periscope and Instagram early but botched the follow-through. He says he did not want to repeat that mistake in his own role.

Beykpour says Twitter did believe in video but made a classic mistake: it bought Vine and then competed with it internally by building a separate native video stack. The same happened with Periscope, when Twitter acquired premium live rights to Thursday Night Football and put a separate team and technology stack on it. Eventually, after persistence and table-pounding, the two were integrated, and ESPN and individual users now go live on the same stack.

Asked about copying, he says everyone is shamelessly inspired by others, and the question is whether it is done with taste. Twitter had worked on audio internally under a project called Hydra, and when Clubhouse showed a more right-feeling experience, the team took inspiration and put its own spin on it. Spaces went from a six-person effort to the company's number-one priority, and he says he is proud of the radical execution.

Beykpour's best advice for building consumer products is to use many products, including those that seem silly, to hone taste about what is superficial, ugly but useful, or beautiful and useful. He says things that seem dumb at first sometimes become very meaningful. He is now building a consumer company with a small team and is not yet ready to say what it is.

In the lightning round he recommends sci-fi and fantasy, noting that Star Trek's teleportation was a metaphor for Periscope. He asks interview candidates about something that failed and something that succeeded, and he names Perplexity as a rare AI consumer product that has retained him. He also mentions his wife's news startup Particle and a first-job lesson from a boss: "When you got nothing to do, sweep."