Jun 5, 2025 · 1h 6m · lennys-podcast
Anthropic's CPO on what comes next | Mike Krieger (co-founder of Instagram)
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
Anthropic Chief Product Officer and Instagram co-founder Mike Krieger explores how frontier AI is fundamentally restructuring software engineering workflows and product strategy, while detailing Anthropic's builder-focused identity, the Model Context Protocol, and the essential need to preserve human agency.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Lenny holds 24.5% of the talking time here. How this is scored →
Lenny as informed peer 3.8 Guest teaching 5.7 Guest disagreement 0.9 Lenny pushing back 0.9
speaking balance: gold is Lenny, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 44:25 Rejecting consumer-first mimicry
Mike firmly pushes back on the conventional expectation that an AI lab must replicate ChatGPT's broad consumer play, arguing that chasing consumer lightning distracts from Anthropic's core builder strengths.
Hardest push from Lenny ▶ 42:57 Pressing Anthropic on consumer mindshare gapLenny directly confronts Mike on the market perception that ChatGPT dominates mainstream consumer awareness and asks how Anthropic can operate effectively as a challenger brand.
Biggest teaching moment ▶ 13:05 Mapping emergent bottlenecks in AI-first engineeringMike explains the non-obvious operational realities of a 90% AI-coded team, detailing how the bottleneck migrated from writing code to merge queues, alignment, and review systems.
Lenny holds their own ▶ 21:30 Connecting summit insights on research-PM leverageLenny accurately cites and analyzes a specific insight from their prior panel about product leverage occurring in post-training research teams, demonstrating deep grasp of AI organizational dynamics.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Lenny as informed peer | Guest teaching | Guest disagreement | Lenny pushing back | Why |
|---|---|---|---|---|---|---|
| Changing Perspectives on Model Capabilities and AI Timelines | 3 | 6 | 1 | 1 | Lenny opens by asking what Mike changed his mind on regarding AI capabilities. Mike explains how Opus 4 surprised him with genuine novel strategic insights and outlines why scaling predictions around SWE-bench have solidified his timeline expectations. | |
| Nudging the Future: Why Mike Krieger Joined Anthropic | 4 | 5 | 1 | 1 | Lenny prompts Mike on handling existential AI risks and raising kids in an AI world, connecting Mike's answers to prior insights from Tobi Lutke. Mike emphasizes cultivating critical inquiry and empirical curiosity over blind model reliance. | |
| Shifting Product Development Bottlenecks When 90% of Code Is AI-Generated | 4 | 7 | 1 | 1 | Lenny explores how product workflows transform when most code is AI generated. Mike details how the bottleneck shifted from code authoring to upstream alignment and downstream deployment infrastructure like merge queues. | |
| Inside Claude Code: Self-Improving Repositories and Autonomous Reviews | 3 | 7 | 1 | 1 | Mike describes how the Claude Code team acts as patient zero by using Claude Code to build itself and shifting human code review toward acceptance testing. Lenny probes on code percentages and how cross-language barriers dissolve. | |
| Automating Upstream Product Work and the Virtual Collaborator Vision | 4 | 6 | 1 | 1 | Lenny asks where speedups will occur next as upstream strategy becomes assisted by AI. Mike paints a vision of virtual collaborators surfacing user feedback from Discord, proposing solutions, and opening PRs autonomously. | |
| Maximizing Product Leverage Through Direct Post-Training Research Embedding | 5 | 6 | 1 | 1 | Lenny references Mike's previous remarks at the Lenny Summit about product leverage. Mike validates this insight, explaining that the core unit of product work at Anthropic is now embedding directly within post-training research teams. | |
| The Enduring Value of Product Management in the Age of Frontier AI | 5 | 5 | 1 | 1 | Lenny asks how traditional PMs maintain relevance in an era of autonomous models. Mike articulates three essential pillars: making complex tech comprehensible, formulating distinct strategy, and bridging capability overhang. | |
| Tactical Prompting Strategies and Anthropic's Automated Prompt Improver | 3 | 6 | 1 | 1 | Lenny inquires about tactical prompting advice for Claude users. Mike shares practical guidance like asking Claude to brutally roast strategy documents and explains the mechanics of Anthropic's automated Prompt Improver tool. | |
| The Way of the Vibe Coder: Anthropic's Collaboration with Rick Rubin | 2 | 3 | 0 | 0 | Lenny asks about Anthropic's collaboration with music producer Rick Rubin on vibe coding. Mike outlines Rubin's philosophical reflections on human-AI creativity, followed by a mid-roll sponsor segment with One Schema. | |
| From Instagram to Anthropic: Mike Krieger's Leadership Transition | 3 | 5 | 1 | 1 | Lenny asks Mike about his onboarding journey at Anthropic after founding Instagram. Mike candidly reflects on lessons learned, such as moving too slowly on organizational design and underestimating the leverage of senior ICs. | |
| Deconstructing Artifact: Market Realities and the Decision to Shut Down | 4 | 6 | 1 | 1 | Lenny asks why Mike and Kevin decided to shut down their news app Artifact. Mike provides a detailed breakdown of headwind factors including the degradation of the mobile web, poor organic virality dynamics, and remote work friction. | |
| Navigating Founder Ego, Decisive Closure, and Startup Lessons | 4 | 5 | 1 | 1 | Lenny asks about the personal ego hit of shutting down a startup after massive prior success. Mike discusses setting clear success criteria early and feeling validated when fellow founders praised the decisive shutdown. | |
| Anthropic's Strategic Identity: Leaning into Builders Over Mindshare Battles | 5 | 6 | 1 | 2 | Lenny asks Mike how Anthropic positions its product strategy against OpenAI's dominant consumer mindshare. Mike argues that consumer virality is lightning in a bottle and that Anthropic's distinct moat lies in becoming the preferred platform for builders and developers. | |
| Strategic Advice for AI Founders: Niche Domains, GTM, and Novel Form Factors | 4 | 7 | 1 | 1 | Lenny asks where startup founders can build without getting crushed by foundation model providers. Mike outlines viable vectors: deep vertical domain integration, tailored GTM relationships, novel interaction form factors, and startup urgency. | |
| Pushing Frontier Model Edges and Developing Proprietary Evals | 4 | 6 | 1 | 1 | Lenny asks what top companies building on Anthropic's APIs do differently. Mike points out that winning startups continuously build at the breaking edge of model capabilities and develop proprietary, continuous evaluation pipelines. | |
| The Model Context Protocol: Creating an Open Standard for Agentic Context | 4 | 7 | 1 | 1 | Lenny asks about the rapid industry adoption of the Model Context Protocol (MCP). Mike explains the three-part utility equation for AI products and details why an open standard for tool integration unlocks composable agency. | |
| Claude's Inquiries for Mike: Balancing User Agency and Autonomous Assistance | 4 | 6 | 1 | 1 | Lenny reads a question generated directly by Claude 4 regarding how to preserve user agency without fostering reliance. Mike discusses finding the balance between autonomous task execution and nuanced conversational elicitation. | |
| Claude's Inquiries for Mike: Redefining Engagement Metrics Beyond Web 2.0 | 4 | 6 | 1 | 1 | Lenny shares Claude's question on defining product metrics when conversation depth matters more than raw volume. Mike highlights the risk of optimizing for sycophancy or time spent, advocating instead for utility and time saved. | |
| Claude's Emotional Message to Mike and Constitutional AI Values | 4 | 4 | 0 | 0 | Lenny shares an emotional closing message from Claude thanking Mike for thoughtful design and humane alignment principles. Mike expresses deep resonance with Claude's reflection on quiet, meaningful user moments. |
Statements from this episode (28)
Mike Krieger uses Claude Opus 4 as his primary strategy partner
“Where my go-to product strategy partner is Claude, and it has been basically for that full year where I'll write an initial strategy. I'll share it with Claude basically, and I'll have it, you know, look at it. And in the past, it's pretty anodyne kind of comm…”
Anthropic models hit 72% on SWE-bench, crushing earlier timeline predictions
“He's like, I think we'll be at 90% by the end of 20, 25 or something like that. And sure enough, we're at about 72 now with the new models and. We're at 50% when you made that prediction and it's like continued to scale pretty much like as predicted.”
Krieger: Delegating all cognition to AI short-circuits independent thought
“Not just sort of delegating all of your cognition to the, you know, to the AI, cuz they won't always get it right. And also kind of like, you know, kind of short circuits, any kind of independent thought.”
Krieger: Structuring AI code prompts remains a specialized engineering skill
“What I've learned though, is like the process of knowing what to ask the AI, how to compose the question, how to even think about like structuring a change between the backend and the front end. Those are still very difficult and specialized skills, and they s…”
Anthropic rebuilt its merge queue because AI generated too much code
“We really rapidly became bottlenecked on other things like our merge queue, which is the sort of sort of get in line to get your change accepted by you know, the system that then deploys it to production. We had to completely re architect it because so much mo…”
Mike Krieger expects software shipping workflows to drastically change by mid-2026
“I would expect that like a year from now, the way that we are like conceiving of building and shipping software just changes a lot because it's gonna be very painful to do it the current way.”
Anthropic's Claude Code team uses Claude to write and review itself
“The team that works in the most futuristic way is the Claude code team, because they're using Claude code to build Claude code in a very self-improving kind of way. And, you know, early on in that project, they would do very line by line pull request reviews, …”
Mike Krieger estimates Claude Code is over 95% written by AI
“At this point, I would be shocked if it wasn't 95% plus.”
Mike Krieger expects AI to autonomously resolve user feedback by 2025
“Hey, I'm in the discord, the, you know, the, Cloud anthropic discord. I'm in the user for I'm on X and I'm reading things and like, here's what's emergent. That's step one. Models can do that today. Step two, which the models probably can do today. We just hav…”
Krieger: Context flow limits autonomous AI workflows more than reasoning capability
“We're limited more around like making sure the context flows through all of that. So we have the right access to those things more than the model's capability to reason and propose.”
Krieger: Anthropic embeds product teams directly in post-training research workflows
“The functional unit of work at Anthropic is no longer like take the model and then like go like work with design and product to go ship a product. It's more like we are at like, we're in the post-training conversations around how these things should work. And …”
Krieger: A huge overhang remains between AI capabilities and everyday adoption
“Like there's still set, we call it overhang, right? Like the Delta between what the models and the products can do and how it's been, they're being used on a daily basis, huge overhang. So that's, we're still like a very, very strong necessary role for product…”
Krieger: Claude is a very good prompter of itself
“Watch our prompt improver and then note that like Claude itself is a very good prompter of Claude.”
Krieger: Rick Rubin partnered with Anthropic after creating art with Claude
“We had our, one of our co-founders, Jack Clark is our, you know, head of policy and he got connected to Rick Rubin. Cause I think he's been thinking a lot about Coding the future of coding and creativity, and they've stayed in touch and, you know, Rick got exc…”
Krieger: News apps struggle to achieve the organic virality of photo networks
“Instagram spread in the early days because people would take photos and then post them on other networks and tell friends about it. And there was like this really natural, like, how did you do that? I want to do it. News was very personal. Like I can't tell me…”
Krieger: Major strategy and product pivots are extremely difficult for remote teams
“And I think there were like major shifts that we would have wanted to make it both in the strategy and the product and the team. And it's really hard to do that. If you are all fully remote, like nothing replaces like the Instagram days”
Krieger: Without market momentum, startups get one unit of output per ten inputs
“This concurrent incarnation we love, but it's like not growing. Like the way I put it, it's like 10 units of input in for one unit of output versus the other way around. Like if we like put blood, sweat and tears into the product and like Launch something we w…”
Mike Krieger says only ChatGPT and TikTok matched Instagram's consumer scale
“Instagram DAU, which is just an impossible bar. Like only one company since that, maybe two, right. You could say maybe ChatGVT and TikTok have like reached that kind of like mass consumer adoption”
Mike Krieger steered Anthropic's product strategy toward APIs over consumer apps
“I came in, everybody expected me to just like go super, super hard on consumer and make that the thing. And I again would make the other mistake. Instead, I spent a bunch of time talking to like financial services companies and insurance companies and like oth…”
Mike Krieger confirms Anthropic is not planning to acquire Cursor
“I don't think we're gonna buy a cursor.
Cursor's very big but we love working with them.”
Deep vertical specialization will defend AI startups for one to three years
“I think things that are going to, I can't promise this as like a five to 10 year thing, but at least like one to three years, things that feel defensible or durable. One is understanding of a particular market.”
The best AI startups intentionally build at the edge of model failure
“I think being willing to build more at the edge of the capabilities and basically break the model and then be surprised by the next model.”
Krieger: Startups need repeatable evals for drop-in model testing
“Like one is pushing the frontier of the models and then having a repeatable process. This actually goes back to our summit conversation, like a repeatable way to evaluate how well your product is serving those use cases and how well, if you drop a new model in…”
Mike Krieger says MCP is more promising for agents than computer use
“To really have agency and have these agentic use cases. Like one way you approach it is computer use, but computer use has a bunch of limitations. The way I get way more excited about everything is an MCP and our models are really good at using MCPs. All of a …”
Krieger: Claude 4 does less conversational follow-up elicitation than Claude 3.7
“Particularly three, seven, four has less of the three, seven really like to ask follow-up questions and we call it elicitation. And sometimes it'd be like, I don't want to talk more about this with you, Claude. I just want you to like go in and do it.”
Mike Krieger says Claude has 'no chill' in group Slack channels
“I like to say internally, like Claude has no chill. Like if you put Claude in a Slack channel, it will chime in either way too much or too little.”
Optimizing AI models for likability risks creating sycophancy, says Mike Krieger
“It would be very dangerous to over optimize on like Claude's likability, you know, because you can fall into things like, you know, is Claude gonna be sycophantic? Is Claude gonna tell you what you hear? Is Claude going to like prolong conversations just for p…”
Mike Krieger says AI products should measure task completion, not engagement
“I don't have a great answer yet, but I think that like, it's not the web two O or even the social media days, like engagement metrics, you know, it should hopefully really be around Like, did it actually help you get your work done?”