as
The $1B Pivot That Saved Twitch…
@ a16z speedrun
06/16/25
Takeaways
ES
Emmett Shear
06/16/25
@ a16z speedrun
Before Twitch became a billion-dollar empire, it started as Justin.tv, a 24/7 live stream of one person's life without any script or plan, created by founders who were just trying to survive after selling a calendar app on eBay.
ES
Emmett Shear
06/16/25
@ a16z speedrun
The initial idea for Justin.tv emerged from discussions about what to do with Yahoo, leading to the unconventional decision to live stream their lives, which eventually evolved into a platform for others.
ES
Emmett Shear
06/16/25
@ a16z speedrun
After launching Justin.tv, we quickly realized that while we gained initial excitement and media coverage, we struggled with viewer retention and decided to pivot to a tech company that opened the platform for everyone instead of trying to create entertainment ourselves.
ES
Emmett Shear
06/16/25
@ a16z speedrun
At Twitch, we made a clear decision to prioritize the needs of streamers over viewers, focusing on streamers because they were the key to our platform's success, especially in the early days when there were only about 200 gaming streamers who mattered.
ES
Emmett Shear
06/16/25
@ a16z speedrun
Through interviews with streamers, I learned that they were primarily motivated by the desire to make money, even small amounts, which made them feel like professional streamers rather than just hobbyists.
ES
Emmett Shear
06/16/25
@ a16z speedrun
When we pivoted from Justin.tv to Twitch, we had about 400,000 monthly active users, which was only 2% of Justin.tv's total user base, but it was the only content I was genuinely interested in running.
ES
Emmett Shear
06/16/25
@ a16z speedrun
Within two months of launching Twitch, we realized we had product-market fit, experiencing growth rates of 30-40% per month, which was a clear indicator of our success.
ES
Emmett Shear
06/16/25
@ a16z speedrun
As Twitch scaled, we focused 70-80% of our engineering efforts on improving our video system because if the video didn't work, it didn't matter how good the rest of our product was.
ES
Emmett Shear
06/16/25
@ a16z speedrun
In the early stages of hiring, be cautious about adding people when you don't have product-market fit, as each new hire can add momentum in the wrong direction; once you find product-market fit, you should hire rapidly to build energy behind your direction.
ES
Emmett Shear
06/16/25
@ a16z speedrun
When evaluating new employees, consider their output and velocity; if they are producing a high volume of work, even if it's not perfect, you can guide them to improve over time. I learned that it's much easier to redirect high volume output than to increase the volume of output.
ES
Emmett Shear
06/16/25
@ a16z speedrun
To get the most out of your best people, clearly articulate what you need and then let them figure out how to achieve it.
ES
Emmett Shear
06/16/25
@ a16z speedrun
As a CEO, my priority is to not run out of money, and equally important is to be the keeper of the vision, constantly rearticulating where we're going and why.
ES
Emmett Shear
06/16/25
@ a16z speedrun
The hardest part about being a CEO is balancing the need to inspire with the operational demands of running a business.
ES
Emmett Shear
06/16/25
@ a16z speedrun
When it comes to valuation, don't get hung up on your first round's valuation; the most important thing is to close the round and get back to work.
ES
Emmett Shear
06/16/25
@ a16z speedrun
The biggest decision we made at Twitch was to be generous with revenue share, which was painful for us but crucial for attracting and retaining streamers.
ES
Emmett Shear
06/16/25
@ a16z speedrun
Partnerships that worked for us were those where we had a product that people loved so much that they demanded the partnership, rather than us doing all the selling.
ES
Emmett Shear
06/16/25
@ a16z speedrun
I believe alignment in AI is key to building safe and capable systems, and it should be viewed as a mutual relationship rather than a tool to control.
ES
Emmett Shear
06/16/25
@ a16z speedrun
I'm building reinforcement learning simulations to understand the fundamental nature of alignment, which I believe is crucial for the future of AI.