Feb 22, 2023 · 1h 20m · news
Kevin Systrom & Mike Krieger: The Founders of Instagram Reveal their New Project "Artifact" | E981 · 20VC with Harry Stebbings
Kevin Systrom · 39m spoken Mike Krieger · 23m spoken Harry Stebbings · 10m spoken Dave Clark · 0s spoken
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Instagram co-founders Kevin Systrom and Mike Krieger discuss their enduring professional partnership, the operational and personal lessons learned from scaling Instagram, and their new venture, Artifact, which leverages machine learning to redefine content discovery.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 14.7% of the talking time here. How this is scored →
Harry as informed peer 2.9 Guest teaching 4.5 Guest disagreement 2.3 Harry pushing back 2.2
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 44:43 Kevin dismisses vertical investing as BS
Kevin forcefully rejects prevailing VC wisdom by declaring that investing in specific verticals is total BS and that outcomes are 100% driven by team quality.
Hardest push from Harry ▶ 47:38 Harry challenges Kevin's anti-vertical stanceHarry refuses to accept Kevin's blanket dismissal of vertical investing, stepping up to defend specialized VCs who leverage deep domain expertise to identify superior founders.
Biggest teaching moment ▶ 37:56 Kevin explains the breakdown of the social graphKevin systematically dismantles the foundational assumption of major social networks, explaining why relying on friends as content filters was a fundamental mistake.
Harry holds his own ▶ 47:38 Harry leverages VC domain knowledge to counter KevinHarry demonstrates strong industry fluency by framing the investor perspective on specialization, LP expectations, and deal detection to push back against Kevin's generalist team argument.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| The Foundations of the Systrom-Krieger Partnership | 2 | 2 | 1 | 1 | Harry opens with warm-up questions regarding the founders' 13-year partnership and how their dynamic evolved. Mike and Kevin emphasize mutual trust and complementary skills, keeping the conversation collaborative. | |
| Aligning Visions and First-Time Parenting Empathy | 2 | 3 | 1 | 1 | Mike explains how becoming first-time parents gave them shared empathy and how they intentionally spent weeks aligning on expectations before starting Artifact. | |
| The Reality of Failure, Suffering, and the Marathon Analogy | 2 | 5 | 2 | 2 | Harry shares a personal feeling of constantly failing across roles. Kevin reframes this by explaining that startups are inherently broken and quotes Ray Dalio about enjoying the suffering. | |
| The Hedonic Treadmill and Finding True Happiness | 2 | 4 | 1 | 2 | Harry asks about their relationship with wealth, and Kevin explains the hedonic treadmill before contrasting material wealth with meaningful family moments. Harry pushes slightly on whether wealth reduces operational urgency. | |
| Balancing Pragmatism and Perfectionism in Product Launches | 2 | 3 | 2 | 1 | Mike contrasts his pragmatism with Kevin's perfectionism during product launches. He gently challenges the startup cliché that launching late means you waited too long. | |
| Expectations and the Cold Start Data Problem | 2 | 5 | 2 | 2 | Harry asks if they felt pressure from expectations. Kevin admits they overbuilt the product and describes the cold start problem in machine learning apps where data loops require initial scale. | |
| Building Recommendation Science Over News Services | 3 | 6 | 3 | 3 | When Harry asks for Artifact's Trojan horse insertion point, Kevin explicitly reframes the premise, clarifying that Artifact is not a news service but a recommendation science platform. | |
| Defining Retained Users and Beta Feedback Loops | 3 | 3 | 1 | 1 | Harry asks about key retention metrics. Mike details how they track content consumption, category breadth, and messaging responsiveness across beta cohorts. | |
| Preventing Feature Creep: The Jobs-to-be-Done Framework | 2 | 5 | 1 | 1 | Harry asks how they prevent feature creep. Kevin uses Christensen's Jobs-to-be-Done framework and a music theory analogy regarding dissonant 13th chords to explain product focus. | |
| Positioning, App Store Limits, and the Vision of Social ML | 3 | 5 | 3 | 3 | Harry points out the tension between Kevin saying they aren't a news app while App Store copy says personalized news. Kevin and Mike explain tactical App Store positioning versus strategic long-term ML ambitions. | |
| The Flaw of the Social Graph and the TikTok Wave | 4 | 6 | 3 | 2 | Harry asks if consumer social is abandoning the social graph. Kevin strongly agrees, arguing that assuming friends share content tastes was the biggest mistake social networks made in the last 20 years. | |
| Vertical Communities, Defensibility, and Why Verticals Are "BS" | 7 | 6 | 6 | 7 | Kevin calls vertical investing 'BS' and insists success is 100% team-dependent. Harry aggressively pushes back, citing specialized VCs who claim domain expertise gives them superior founder detection. | |
| Tailoring Startup Processes and Sizing Down | 3 | 4 | 1 | 2 | Harry asks how hiring processes adapt to large applicant volumes. Mike explains the need to right-size processes for early-stage startups rather than blindly adopting big-tech frameworks. | |
| Trust in Hiring, Firing Fast, and the "Boring" Executor | 4 | 6 | 5 | 3 | Harry asks how to build trust quickly in hiring. Kevin reframes the problem, arguing 100% trust upfront is unrealistic and founders must fire fast instead. Kevin also rebuts Harry's complaint about 'boring' candidates, calling quiet execution a feature. | |
| Sourcing Raw Talent and Breaking Down Boundaries | 3 | 5 | 4 | 4 | Harry asks if startups still belong in SF despite tax revenues leaving. Kevin vigorously defends SF as the 'kingdom of nerds' and rejects doom-loop narratives. | |
| Bouncing Back, Career Worry, and the Value of Randomness | 2 | 4 | 1 | 2 | During quickfire questions, Kevin advises his 20-year-old self not to worry and to incorporate randomness. Harry gently calls out sample bias, noting Kevin only interviews billionaire founders. | |
| Toddler Routines, Angel Portfolio Regrets, and Gritty Founders | 3 | 4 | 1 | 1 | Mike reflects on angel investing mistakes, noting he regretted following consensus hype over backing gritty founders who can push through adversity. | |
| Startup Volatility, Zero Interest Founders, and Instruments | 4 | 6 | 4 | 3 | Kevin criticizes entitlement among 'Zero Interest Founders' (ZIPs) and uses a flight-by-instruments metaphor to explain how experienced founders manage startup volatility. | |
| The Five-Year Vision for Artifact | 2 | 3 | 1 | 1 | Mike and Kevin share their five-year vision for Artifact, emphasizing publisher partnerships, creator empowerment, and building user trust in machine learning algorithms. |
Statements from this episode (32)
Systrom: Machine learning is fundamentally changing social product consumption
“We think fundamentally machine learning is changing the way people consume social products, and we want to ride that wave and build the best social product using machine learning we can.”
Krieger: Unresolved initial disagreements inevitably destroy founder partnerships
“I've definitely seen that, you know, having, we've done some angel investing and I've seen partnerships go south where there was some initial disagreement It either got lightly talked about, never fully resolved, or never talked about, and then it inevitably b…”
Systrom: Instagram co-founders met in person only five times last year
“Mike and I haven't done a remote company before. We've seen each other how many times in person in the last year? Like, five?”
Krieger: Instagram's seed funding was contingent on finding a technical co-founder
“You know, with Instagram, Kevin had some basically initial like seed funding contingent on finding a technical co-founder.”
Systrom: Companies never reach stable perfection; founders must choose their suffering
“The myth that things are ever just in the stasis of good is totally BS, and what people need to realize is that, like Ray Dalio likes to say it's, like, not about avoiding suffering, and I'm paraphrasing here, it's about, like, learning to enjoy the suffering,…”
Systrom spends at least 16 hours a day working at his desk
“I mean, I'm in this chair Gosh, 16 hours a day at least.”
Systrom would not work 16-hour days if he were in an office
“And if I were working this hard elsewhere, first of all, I wouldn't work this hard in an office. It just wouldn't happen.”
Krieger: Instagram's UI broke at launch when posts exceeded 30 likes
“The second something got more than 30 likes on a post, it totally broke the UI on launch because we didn't summarize it.”
Krieger: Artifact delayed launch by months to perfect its recommendation algorithm
“With artifact, the interesting thing is I feel like the sort of core UI, you know, we probably could have shipped several months ago, but like this is product lives or dies by its recommendation.”
Systrom: Artifact overbuilt social features and could have launched much earlier
“There's parts of this product that exist, you can DM people, you can post socially, like, Facebook or Twitter, like, they're entire, there's entire pieces of this app that people can use, it's just locked away and not launched. We have over, we could have laun…”
Systrom: Artifact should have launched its core kernel quickly for feedback
“If anything, I think we should have started with the kernel of the idea, launched that thing really quickly and then gotten feedback.”
Systrom: Artifact is a recommendation engine, not a news service
“I think what people don't understand is we're not building a news service. Like, maybe I shouldn't be saying this as much as I say it, but like, no, I'm interested in core recommendations. I'm interested in the science of utilizing people's tastes and profiles…”
Systrom: News is ideal for bootstrapping machine learning taste profiles quickly
“News is a great place to start because there's a ton of information out on the web. Machine learning is very, very good at analyzing it. And we can build portraits very, very quickly”
Systrom: Artifact achieved hundreds of thousands of first-day sign-ups
“We had hundreds of thousands of people sign up the first day.”
Systrom: Artifact could not have been started by unknown founders in 2010
“This is not a company we could have started at a Dogpatch Labs in 2010. It's just not, we couldn't have done it.”
Krieger: Day-one Instagram growth relied on filters, not network effects
“In day one of Instagram, you know, there wasn't really a network there. You were there because the filters made your, you know, photos from your iPhone three G. Or three GS look a little better or a lot better”
Systrom: Features that contradict an app's core job grate on users
“They can just grate at you if they include things that are dissonant to the core job. So you need to be very careful about forming that core job, and only adding things in when they support the core job.”
Systrom: Every successful social app begins as a utility before scaling
“Every app that is successful in social starts off as a utility, and it becomes a large social behemoth over time.”
Systrom: Believing friends share content interests was social media's biggest mistake
“And the fundamental thesis is that your friends are kind of into the stuff that you're into, and I would argue it is the greatest mistake Of social networks in the last 20 years, is to believe that is true.”
Systrom: Social networks must become less social and rely on behavioral affinity
“So, long time, like, long story short, I think that social networks need to become less social, and I think that something that will make up for it is this idea of learned association or learned affinity through your behavior.”
Systrom: Investing in industry verticals is 'BS' compared to team quality
“This is gonna be heresy to a VC's ears, but like, I think investing in verticals and spaces is BS. Like, yes, in general, generative AI is gonna produce something pretty cool, Like, yes, the stuff, like, these verticalized, like, the sharing economy, it's like…”
Systrom: Facebook's key legacy is teaching PMs scientific growth methods
“The one thing that I think Facebook should be remembered for as a professional development, like, incubator. Is its ability to teach young PMs the scientific method of growth. How do you grow things? Maybe to a fault, by the way.”
Krieger: Artifact launched with a team of just seven people
“Our team has seven people, you know, we're just getting started.”
Krieger: Personalizing Instagram's Explore feed resulted in a 2x usage boost
“When we launched Explore and made it personalized rather than just like, most popular photos, like, that was a two X delta in usage.”
Systrom: Second-time founders fire fast while first-time founders struggle
“And if it's not the right fit for whatever reason, you just, you move on it. And I think that's something a second-time founder will do that a first-time founder has a lot of trouble doing.”
Systrom: Early Instagram mistakenly prioritized 'mission hipsters' over top talent
“As a small company, one of our mistakes in Instagram is hiring too many, like, mission hipsters that wore plaid and really liked drinking bourbon. Like, that was stupid. Like, we should have just been hiring the greatest people out there.”
Systrom: San Francisco will rebound as top startup center within five years
“But like, mark my words, look back five years from now, and I, you're even seeing the glimmers of it now. Just all these companies getting started. If you look at history, history, what is it? It's like it never repeats perfectly, but it rhymes. And there's no…”
Systrom: Injecting margin randomness into decisions is optimal long-term
“Be a little bit random. And actually, machine learning would tell us that, which is that being a little random on the margin is actually optimal in the long run.”
Krieger admits investing in deals based on hype without conviction
“There were like a few either, like, LP positions I took or individual investments. I was like, man, like, I did this because it felt like this was a thing that a lot of people were excited about, and I wasn't”
Systrom: Startups grind people down through workload trade-offs and emotional volatility
“This stuff's just hard, and that's why the startup process just chews up and spits out people, like, no process I've seen before. It is just a grinder. Like, it goes through people like crazy, because it either screens you out because it's too hard work-wise, …”
Systrom: Founders must rely on dashboards over gut feelings during volatility
“Just trust the instruments. And like, my god, we have instruments, dashboards, et cetera, just trust the instruments. Forget about what people are saying, forget about what issue you had yesterday with the server, trust the instruments. And it doesn't make it …”