How Kayvon Beykpour changed Twitter's product culture and what he learned building Periscope

Executive overview

Twitter had calcified into a risk-averse, stagnant product org — users complained the product hadn't changed in eight years. Kayvon Beykpour, as head of product and then GM of consumer, spent years dismantling sacred cows and shifting the org toward shipping real bets.

The core lever was changing who owned the work: bringing in acquired founders, building protected silos for speculative bets, and staffing projects only with people who believed in them.

Sacred cows aren't blockers — they're a free roadmap of everything worth building.

Getting fired and meeting Elon

• Parag Moheajeri became CEO after Jack resigned and restructured the company from functional to GM model, promoting Kayvon to GM of consumer
• One month later Kayvon went on paternity leave; the day after returning from the hospital, Parag called and fired him, citing a new direction
• That same night, Twitter signed the term sheet with Elon
• Scott Belsky connected Kayvon with Elon during the chaotic first days of the acquisition; they met in-person at Twitter HQ — with Walter Isaacson silently observing
• Elon invited Kayvon to "just hang out" and "swipe left or right" on ideas; Kayvon declined, deciding it was someone else's turn
• Kayvon flagged people he thought were exceptional who are still at the company; projects he was bullish on (Community Notes, Spaces, Communities) continued under Elon

Changing a stagnant product culture

• The prior strategy — "refine the core" — had worked: moving to a ranked timeline restored DAU growth, but it further calcified the org's fear of risk
• The goal wasn't to abandon core refinement but to add a portfolio of ambitious bets alongside it
• Functional org structure (separate product, engineering, design, research leads) required consensus, not execution — this was the root cause of slowness
• Kayvon's first year was "walking through mud" — politically and bureaucratically exhausting; it got easier as early wins built momentum
• A PM was told by an engineer not to work on the hide-replies feature because it was bad for her career — a microcosm of the culture problem
• Storytelling mattered: internally repeating the vision, and externally signaling to users and advertisers that things were changing

Using acqui-hires to break through

• Bringing in acquired founders accelerated cultural change: they drove urgency, ambition, and had startup instincts — but also the savviness to operate inside a large org
• Spaces, Communities, Community Notes (then Birdwatch), Fleets, and creator monetisation (Super Follows, tipping) were all led by founders from acquisitions
• Keith Coleman (Community Notes) was Kayvon's predecessor as head of product; he ran Birdwatch as a protected silo insulated from OKR pressure
• Esther Crawford (creator economy) and Mo Oladam (Fleets) were both acquired entrepreneurs given latitude to lead
• The key staffing lesson: people who believe in the idea will work harder, be more creative, and will the thing into existence — misaligned teams are toxic to speculative projects
• Staffing based on who's available rather than who's right for the role is a pattern Kayvon learned to avoid the hard way

Frameworks, OKRs, and jobs to be done

• Jobs to be done was rolled out at Twitter in a way Kayvon found exhausting and unhelpful — he had to defend something he didn't believe in
• The critique isn't unique to jobs to be done: any framework followed religiously loses nuance and becomes process for process's sake
• The charitable version of jobs to be done: it forces thinking through the customer lens and their true alternatives — useful, but you don't need a framework for that
• OKRs had the same failure mode: the org became so focused on driving DAU quarter-over-quarter that it couldn't accommodate bets that would hurt short-term metrics
• Spaces required putting a live bar at the top of the app — which pushed tweets and ads down, hurting DAU and revenue; a pure OKR lens would kill it
• The ranked timeline was great for DAU but produced hostile experiences: the "swish" toggle would silently revert users from reverse-chron back to ranked after 24 hours — users hated it
• Amazon burying order details in confirmation emails is the same failure mode: metrics-optimised, customer-hostile
• Signs a framework is failing: you're making decisions that are bad for users, or you're being disincentivised from taking bets you believe in

Periscope: what went wrong and what it inspired

• Periscope hit retention problems early; surges of growth in new markets (US, France, Turkey, Middle East) masked the underlying decay
• The core thesis — "teleportation, seeing through someone else's eyes" — turned out to be wrong; the vast majority of users just wanted to talk to people
• A standalone live-only consumer product is not durable; it needs asynchronous scaffolding (like Instagram or TikTok) to retain users between broadcasts
• Integration with Twitter was too slow; Twitter was distracted with "refining the core"
• Facebook's response was decisive: directed ~300 people to build live into the core product and bought out prominent Periscope creators with exclusivity deals
• Twitter made the same mistake with both Vine and Periscope: acquired the right insight, then competed with the acquisition internally by building a parallel native video team on a separate stack
• The same pattern repeated with live video: UGC live (Periscope stack) and premium live (Thursday Night Football/NFL) were run as separate teams with separate architectures
• Eventual resolution: the stacks were merged, so ESPN going live and a user going live from their phone now share the same tech

Building Spaces after Clubhouse

• Long before Clubhouse, Twitter had a project called Hydra (multiple heads = conversation participants) exploring audio and video formats — none felt right
• Clubhouse crystallised the right UX for longer-form, multi-person audio conversations; Kayvon credits Paul and Rohan's team for that
• Twitter took what worked from Clubhouse and added Twitter-native mechanics, making it a different product
• Spaces went from six people to the company's top priority; the speed was driven by having already felt the pain of being too slow with Vine and Periscope
• Kayvon considers Spaces the project he's most proud of from his time at Twitter

Advice on building consumer products

• The best cheat code: be a voracious user of products — try new things, notice what works, what's ugly but useful, what's beautiful and useful
• Don't judge quickly; things that seem dumb early sometimes become meaningful
• Copying ideas can be done with or without taste; what matters is whether you put your own spin and serve the customer better
• Build the product you want to use: finding your own pain points is one of the best inputs to building something wonderful
• Synchronous communication (live video, audio rooms) needs asynchronous scaffolding around it to be durable as a consumer product