Podcast Summary: Building One with Tomer Cohen

Episode: Building Anthropic with Mike Krieger: Product Playbooks In The Age Of AI, Why Memory Is Key, And Instagram Lessons

Date: December 9, 2025
Host: Tomer Cohen (LinkedIn CPO)
Guest: Mike Krieger (Co-founder of Instagram & Chief Product Officer, Anthropic)

Episode Overview

This episode features Mike Krieger, best known for co-founding Instagram and now serving as Chief Product Officer at Anthropic, a leading AI company. Tomer and Mike weave together lessons from building Instagram, insights from Mike’s second startup Artifact, and pioneering approaches Anthropic is taking to build AI-native products. They explore themes like product-market fit, measuring product success, product development in the age of AI, and the centrality of "memory" in creating effective AI assistants.

Key Discussion Points & Insights

1. Connecting Early Influences to Entrepreneurial Success

[02:12 – 04:50]

• Symbolic Systems at Stanford: Mike shares how an interdisciplinary degree in computer science, linguistics, psychology, and early AI helped shape his approach to building products grounded in human behavior and technical innovation.
  • “It was like, in some ways the perfect degree if you want to be an entrepreneur...Also combining how to build, but also who you’re building for and how people and machines think.” [02:36]
• Design Thinking: Exposure to Stanford’s d.school ingrained a user-centered process for problem-solving and discovery.
• Mayfield Fellows Program: Provided practical, accelerated exposure to startup case studies, which later allowed Mike and Instagram co-founder Kevin Systrom to share a language for tough product decisions.

2. Lessons From Instagram: Finding & Focusing the Product Core

[06:06 – 08:25]

• Iterative Beginnings: Instagram began as a check-in app (“Bourbon”) but pivoted after realizing the product’s core wasn’t driving growth.
  • “It’s very rare that a product succeeds because of the incremental N+1 feature if the core is not actually working.” – Mike [06:24]
• Crisis & Pivot: Investor skepticism prompted Mike and Kevin to re-examine Bourbon’s core, stripping it down to two key functions—photos and friend plans. Focusing on photos resulted in the clear product fit that became Instagram.
  • “It's always tempting to say, ‘and then this, and then this,’ without...rethinking the core.” [07:12]
  • “With Instagram, within a week of having a better prototype…it was clear that was going to be more intuitive and exciting.” [07:55]
• Importance of Intuitive UI: Creating recognizable patterns (e.g., tab bar with photo button) made Instagram’s usage immediately obvious.

3. Knowing When to Persist or Shut Down: Lessons from Artifact

[09:52 – 12:10]

• The "Energy in the System" Metric: Mike describes how Artifact, an AI news app, was shut down not because of user numbers, but the diminishing energy: “10 units of effort resulted in 1-2 units of output.”
• A Systematic Approach to Minimizing Regret:
  • Make a concrete list of things to try before considering shutdown.
  • Set a clear timeline to experiment and ensure nothing is left unevaluated.
  • “We tried them all basically and…we kind of let us enter 2024 and say...it's time to call it now.” [11:40]

4. Building with AI: Rethinking Product Playbooks

[12:10 – 17:11]

• Simplicity and Chat as the UX:
  • While Instagram focused on simple, intuitive interfaces, Anthropic’s Claude uses a chat box—surprisingly persistent and effective given the complex potential of AI.
    • “The thing I’ve most changed my mind on is...chat boxes are actually a very useful way of expressing an open-ended problem...The thing that needs to change though is what happens after you hit enter.” [14:47]
• Progressive Disclosure:
  • A challenge is exposing the AI’s range of capabilities without overwhelming users.
    • “You want to provide a fair amount of room to run, but at the same time also express the full list of potential or capabilities...” [13:40]
  • Rather than UI-based tips, empowering the model to self-suggest relevant capabilities in appropriate contexts leads to better user discovery and mastery.
• Non-Deterministic Roadmapping:
  • Traditional product roadmaps (as at Instagram) don’t translate to the flexible, evolving landscape of AI model capabilities.
  • Anthropic now reserves more roadmap space for continuous experimentation and prototypes, ready to turn breakthroughs into new product features.
    • “It’s preserving space in the roadmaps for prototyping experimentation...there's more room now than definitely there was at Instagram around experimentation.” [16:12]

5. The Central Importance of Memory in AI Product Design

[17:11 – 19:42]

• Building “Relationship” in AI:
  • Lasting “stickiness” comes from the AI remembering a user’s context and preferences—establishing a sense of trust and ongoing investment.
    • “They will learn a lot about you. And there are these memories and I think those are really important to developing that trust and long horizon relationship...But using that judiciously...feels empowering overall.” [18:29]
• Memory as a Multi-Layered Capability:
  • In-context memory: What’s ongoing in a conversation.
  • User-level memory: Knowledge about the individual user.
  • Procedural memory (“skills”): Remembering how to execute recurring tasks.
  • Organizational memory: What an organization knows collectively, who knows what, and which agents have done what before.
    • “I think of it as essential to everything we’re doing...It’s a big unlock around really feeling like a kind of trusted collaborator...” [19:24]

6. Measuring Success and Product Metrics in AI

[19:42 – 22:24]

• True North as “Relationship,” Operationalized Through Participation & Retention:
  • Focus on whether people are enabled to use features (e.g., connecting data sources), and whether engagement persists.
    • “We look at a lot of, have people connected their Google Drive...have they used memory? Do we even have enough context...to do useful things for them?” [20:24]
  • Evaluate both the uptake of core features and the retention loop (participation rate).
• Role of Evals:
  • Unique to AI, upstream model evaluations (“evals”) assess model performance on key user-centric tasks (like generating office documents) before user engagement metrics come into play.

7. Personal Practices, Curiosity & Advice for Product Leaders

[22:24 – 24:00]

• Favorite Non-Software Product:
  • Mike loves simple, high-quality physical tools that improve wellness, like a “self-massaging back thing” found on Amazon.
• Advice for Aspiring Builders:
  • Ruthless curiosity and experimentation—try early versions, join feature flags, and stay hands-on with nascent products.
    • “Remaining absolutely curious and engaged I think is the way to do it.” [23:45]

Notable Quotes & Memorable Moments

• On Product Market Fit:
  • “It’s very rare that a product succeeds because of the incremental N+1 feature if the core is not actually working.” – Mike Krieger [06:24]
• On Simplicity in Product Design:
  • “The same way as if you pick up a tool or you pick up a coffee mug…you would know how to hold it and what to use it. I think what Bourbon had become was you could kind of intuit how you might use it, but there were a lot of questions…” – Mike Krieger [08:39]
• On AI User Interfaces:
  • “The thing that needs to change though is what happens after you hit enter...is it just the model answers your question or is it the model's guiding you, connecting the right sources, doing some work independently, showing its state. And it's much more of a collaboration...” – Mike Krieger [14:47]
• On When to Shut Down a Product Idea:
  • “The leading indicator for me was, and I can't quantify this, but it's something around like the energy in the system.” – Mike Krieger [10:12]
• On Building AI Products:
  • “Roadmaps essentially become hypotheses, not contracts.” – Tomer Cohen [24:14, host wrap-up]
• On Memory as a Product Lever:
  • “I think of it [memory] as essential to everything we're doing...It’s a big unlock around really feeling like a trusted collaborator.” – Mike Krieger [19:24]

Timestamps for Key Segments

• Intro to Mike and episode theme – [00:39]
• Stanford & Symbolic Systems – [02:12]
• Instagram's pivot and product-market fit – [06:06]
• Artifact & knowing when to quit – [09:52]
• AI UX and chat interfaces – [12:10]
• How Anthropic’s product playbook differs from classic tech – [15:24]
• Memory as foundational to AI assistants – [18:29]
• Metrics for AI products – [19:48]
• Favorite non-software product & advice to builders – [22:24]

Tone & Takeaway

Richly reflective, candid, and pragmatic, this episode distills hard-won product lessons from one of tech’s best builders. Mike Krieger emphasizes the power of simplicity, the necessity of “energy” between user and product, and how memory and flexibility are at the core of AI-native product development. Builders at all levels will find valuable lessons on curiosity, when to pivot (or quit), and why roadmaps are now more fluid hypotheses in the age of AI.