Kevin Systrom and Mike Krieger on Partnership, Recommendations and Artifact

Mike Krieger says his partnership with Kevin Systrom rests on trust and integrity built over roughly 13 years. When fundraising or other problems arise, neither has to wonder whether the other has his back. They can disagree, take time to digest the argument and return with a better understanding of the other's view.

Krieger prefers raising a problem promptly without assuming it must be resolved immediately. He has seen partnerships deteriorate when an initial disagreement was left unspoken or only superficially addressed.

Systrom praises Krieger's ability to diagnose and solve difficult technical problems, and values the complementarity of their skills. He says even a long partnership needs work: their new remote company places the relationship in a different context from Instagram.

They built Instagram together in person, whereas they had seen each other only a handful of times during the previous year. Both now have families and different responsibilities; Systrom sees the partnership's ability to adapt to those changes as part of its strength.

Krieger says both still wanted to build after leaving Instagram. Before starting Artifact, they spent weeks aligning on what each wanted from the experience, rather than assuming that an earlier successful partnership guaranteed the same expectations now.

Asked how to manage work and family without feeling that something is always failing, Systrom says building a company means constantly fixing things that break. He recommends choosing work whose difficulties are worth bearing, rather than expecting a permanent state in which everything is settled.

Krieger emphasizes explicit agreements with a spouse or partner about schedules and commitments. Someone who can call out missed priorities and help manage them makes the competing demands less isolating.

Systrom describes the hedonic treadmill: an upgrade in spending quickly becomes the new normal. He finds happiness in moments with family and friends and says financial security allows the founders to focus on building something worthwhile instead of making another valuation the only goal.

Stebbings asks whether having money reduces urgency. Systrom says external pressure can increase speed, but his own motivation comes largely from challenging himself; he says working remotely lets him work intensely while still seeing his family, even though he does not particularly enjoy home working.

Krieger describes balancing pragmatism and perfectionism. Instagram's first release polished its core features but also contained obvious flaws, such as an overflowing list of names when a post received many likes; waiting for every detail to be perfect would have delayed real customer feedback.

For Artifact, the founders believed recommendation quality was central to launch readiness. Krieger says their previous success also raised expectations: attention was helpful, but it made them more conscious that the first public version should feel considered.

Systrom says they overbuilt Artifact, including social and direct-message features that were not yet public. In retrospect he thinks they could have launched a smaller kernel sooner; moderation and the attention the product would attract also made a broader initial release difficult.

He calls a recommendation product a particularly hard second startup because better recommendations need usage data, while users need a compelling first experience to generate that data. Some early users may see an ordinary feed and leave before personalization improves.

Systrom says the underlying ambition is broader than a news app: use behavioral data to recommend things in the user's interest. News was the initial use case because there was abundant material on the web and machine learning could analyze it.

His hypothesis is that patterns shared by groups of readers can help initial recommendations, while more specialized interests require more data and scale. At the time of the interview, he says the service was not yet where the founders wanted it to be for niche tastes.

The founders' Instagram history gave Artifact a launch advantage: Systrom says hundreds of thousands signed up on the first day. He hopes that initial attention can generate enough usage to improve recommendations, but acknowledges that the same company would have been much harder for them to start without that history.

Krieger defines early retention around people returning, browsing and reading. Unlike early Instagram, a news product has fresh material every day even before a user's friends join; relevant notifications give the service another chance to bring readers back.

For the small social beta, the team also looked at posts without reactions and unanswered messages, because publishing into a void can feel lonely. Krieger cautions against drawing overly precise conclusions from a small test group.

He says early public users were reading more stories per day than the invited beta group, a reassuring signal because beta testers were relatively close to the founders' network. He does not present the early metric as proof of long-term success.

Systrom uses a jobs-to-be-done lens to guard against feature creep. A customer opens a product for a reason; new features should support that job rather than introduce unrelated functions that confuse the product's purpose.

He compares conflicting features to dissonance in music. At roughly two weeks old, Artifact was deliberately focused on a simple core experience, with additional features to come later if they strengthened it.

The consumer-facing App Store description was personalized news, even though Systrom saw the longer-term ambition as a machine-learning-driven social product. He distinguishes a clear explanation of what customers can use today from a company's broader, uncertain aspirations.

Krieger says interests extend well beyond politics and a product's full ambition is hard to fit into a short App Store description. Launching forced the team to choose a simple category and message while allowing more advanced possibilities to emerge over time.

Systrom argues that friends do not necessarily share the same interests, making a social graph an imperfect filter for content. His and Krieger's tastes in electronic music and Formula 1 illustrate the problem: a strong friendship does not make every article relevant to both people.

He credits TikTok with showing the potential of interest-based discovery and predicts that social services will rely more on learned preferences. He does not foresee the social graph disappearing entirely, and recognizes risks including filter bubbles and the spread of false information.

Krieger separates discovery from conversation: a recommendation may find an article independently of friends, while a user can still share and discuss it with them. He also sees potential in communities formed around common interests rather than existing relationships.

Asked about copying and defensibility, Systrom challenges the idea that an entire sector is automatically unattractive. He puts great weight on the team and its skills, while recognizing that the area also matters.

He credits Facebook with teaching young product managers a scientific approach to growth, including experiments and statistical significance, though he notes that some tactics can be taken too far. The Instagram founders, he says, had not understood those methods before joining Facebook.

Stebbings argues that specialist investors may recognize talent and help founders more effectively in their field. Systrom accepts the point, but remains interested in investors who can identify exceptional people and assess whether a founder can assemble a strong team.

Krieger says Artifact recruited largely through existing networks while its beta was quiet. He is still learning to identify people who will thrive amid a startup's uncertain outcomes rather than expect the structures and fallback options of a large company.

The team also has to match methods to its stage. At seven people, it cannot simply reproduce the processes or technology used by thousands of engineers; prior experience must be filtered through the present problem.

An A/B test in a small beta would have needed a very large effect to produce a statistically significant result, Krieger says. Once public usage grew, the team could run a test with a measurable result; the lesson was to use the right tool for the amount of data available.

On hiring and trust, Systrom says it is unrealistic to know with certainty from a few interviews whether every hire will work. His approach is to make the best decision possible and act promptly, compassionately and clearly if the fit proves wrong.

Krieger adds that keeping someone in a role that is not working is not necessarily the kinder choice. Systrom says founders should improve their ability to recognize fit early while accepting responsibility for difficult decisions when necessary.

Stebbings asks whether to hire someone capable but quiet. Systrom says being entertaining is not a job requirement and recalls a mistake at Instagram of hiring too many people who resembled the founders socially, rather than simply finding the strongest people for the work.

Krieger values a team with some people who generate energy and others who execute quietly; he does not think everyone must have the same personality. Both say highly talented people who mistreat others have been damaging in their experience.

Krieger looks for talent, curiosity and willingness to cross role boundaries in a small company, while recognizing that expertise and clearer boundaries become important as an organization scales. Enthusiasm alone is not enough to guarantee competence.

In quick-fire answers, Krieger says parenthood has brought more breaks and a longer time horizon; Systrom says it forces prioritization because family needs can interrupt even important work. They describe themselves as curious and always learning.

Systrom predicts that San Francisco will remain a major startup center despite its downturn, citing its universities and concentration of technical talent. He also expects a substantial remote component; this is his 2023 forecast, not a demonstrated outcome.

Krieger says prior success does not predict future success and that a second-time founder must still work hard and build a thoughtful team. Systrom would tell his younger self to worry less about status, while acknowledging that many people face serious financial pressures.

Reviewing angel investments, Krieger regrets some decisions driven by excitement around a trend rather than his own conviction. He looks for founders able to respond to adversity, while recognizing that hindsight can distort the lessons an investor draws.

Stebbings criticizes millennials as insufficiently resilient; Systrom recounts a mismatch in expectations with a former Artifact employee, but also notes that every generation complains about the next. Their exchange is personal experience and opinion, not evidence about a generation as a whole.

Systrom describes startup work as emotionally volatile and says prior experience helps the founders interpret it. He compares relying on product metrics to a pilot trusting instruments when sensations are misleading, while acknowledging that doing so can still feel uncomfortable.

He warns against simply extrapolating a period of easy capital and says fashionable investments need scrutiny. Experience of earlier cycles can help people recognize and correct mistakes, though it does not prevent all of them.

Looking five years ahead from 2023, Krieger hopes Artifact could be part of a portfolio of products that make complex information feel useful and personal. Systrom hopes to build trust with publishers, support independent voices and make recommendation algorithms serve readers; he recognizes that achieving this would be a long, uncertain effort.

Stebbings thanks both founders for the discussion, and they close the interview.