Mike Krieger

Mike Krieger builds AI products at Anthropic. He joined as Chief Product Officer in May 2024, overseeing product engineering, management and design, then moved into Labs alongside Ben Mann in January 2026 to develop experimental products at the frontier of Claude’s capabilities. By July 2026, Latent Space identified him as Anthropic’s Head of Labs.[3][4][1]

Krieger’s 2024 appointment gave him responsibility for expanding Anthropic’s enterprise applications and bringing Claude to a wider audience. He said he admired the company’s focus on “capable and trustworthy AI systems that empower humans.” The January 2026 Labs transition placed him alongside Ben Mann in experimental development, while Ami Vora took leadership of Product, partnering with CTO Rahul Patil to scale Claude experiences.[3][4]

In his interview with swyx at the AI Engineer World’s Fair, reported by Richard MacManus on July 3, 2026, Krieger described his team’s use of Claude Tag as increasingly delegated, asynchronous and proactive. The instruction went beyond fixing an individual bug: “Now you are responsible for this part of the codebase, and I want you to monitor this feedback channel and proactively take on tasks.” He described the resulting workflow as “multiplayer, async, proactive,” but also said the team was “bottlenecked on reviews” and on the “human ability to fully conceptualize what we’re doing.”[1]

Before Anthropic, Krieger co-founded Instagram with Kevin Systrom and served as its CTO, then worked with Systrom on Artifact, a personalized news app acquired by Yahoo. His earlier accounts of ranking systems and startup discovery provide context for his experience turning machine learning into consumer products.[3][5]

In a February 2023 Stratechery interview, Krieger explained Instagram’s initial feed-ranking goal as helping people keep up with friends. A friend’s evening post could disappear beneath frequent posts from brands and influencers by the next morning. Chronology already imposed an ordering: “Your feed is already ranked, it’s just ranked by order by created at descending.” He said Instagram could have explained that purpose better publicly.[5]

Krieger recalled that integrating Facebook’s existing feed system was estimated to take two years because it contained so much Facebook-specific logic. Instead, a New York team built Instagram’s implementation in three or four months, learning from Facebook’s experience while designing for Instagram. He also recalled mistaking the missing chart results from an early machine-learning Explore experiment for a testing-tool bug; the response was that the results exceeded the display’s range.[5]

Artifact gave Krieger a different lesson about early product discovery. Instagram had launched with just the two founders; at Artifact, he judged even an initial seven-person team too large and suggested four might have been better. He distinguished the shifting assumptions of a new product from the greater certainty of an established one. He also acknowledged that parts of Artifact had felt ready six months before launch, with perfectionism and expectations surrounding the founders’ second venture contributing to the delay.[5]

Krieger described examining Artifact’s retention cohorts, return triggers and dropoff points, while asking whether readers gained value beyond tapping articles. Reading a longer piece and developing a lasting relationship with the product mattered to his stated approach. He characterized his and Systrom’s work as applying emerging technology through product sense, engineering and practicality to solve consumer problems.[5]

His 2023 view of AI in publishing emphasized writers’ distinctive perspectives and voices, which he considered difficult to replace even if models could imitate their ideas. “So I put my vote or my belief firmly in the writers’ corner,” he said. He supported helping writers find audiences and build long-term reader relationships, while also seeing uses for automated summarization and coverage of niche interests that were uneconomical for humans to serve.[5]