Podcast Summary: AI & I – "How to Build an Agent-native Product"

Host: Dan Shipper
Guest: Mike Krieger (Co-founder of Instagram, at Anthropic Labs)
Date: March 25, 2026

Episode Overview

In this episode, Dan Shipper sits down with Mike Krieger to discuss the evolving landscape of software development in the age of agent-native AI products. Drawing on experience from Instagram’s early days to current work at Anthropic, Krieger explores what has fundamentally changed—and what hasn't—in product building, especially as AI tools accelerate prototyping, iteration, and scale. They dive into product design philosophy, team structure, feature curation, and the future of agent-powered interfaces, all while remaining candid about the challenges and trade-offs of rapid progress.

Key Discussion Points & Insights

The Changing Nature of Product Development

• Prototyping & Acceleration:

  • Today’s AI models (like Claude) greatly accelerate how quickly products can move from concept to completion.
  • "You can get it to go…not zero to one, but zero to end pretty quickly over the matter of hours." (Mike, 02:31)
  • Prototyping is less bottlenecked by engineering; the challenge is what to build, not how quickly.

• Pitfalls of Overbuilding

  • The acceleration tempers intuitive, incremental product development.
  • "What would normally be this sort of incremental thing…You can actually kind of grow an entire tree indoors. And then you have this whole thing that…doesn't have the same level of intuition and exposure to experience at each step." (Dan, 04:47)
  • Modern tools make it easy to add features, but not to curate—or cut—them.

• Product Intuition Still Matters

  • Intuitions and hard-earned lessons from real-world use are irreplaceable.
  • "There haven't been a lot of breakout consumer products…even in the age of accelerated AI building…It just still takes time to hone your view about what sort of intervention you want to make on the world and then build from there." (Mike, 02:31)

Product Simplicity and the Art of Subtraction

• Simplicity as Competitive Advantage:

  • Story of Dan rewriting an overbuilt agent-native app from scratch, inspired by a much simpler, focused app (Monologue).
  • "I just basically threw out the product and started over with…just a shareable markdown link…and…now we launched it and it just blew up." (Dan, 07:09)

• Lean Startup & YAGNI Principles:

  • Overbuilding in v1 is tempting with AI, but creates a brittle, hard-to-test matrix of features.
  • "We way overbuilt for V1...just because you can doesn't necessarily mean that it should be in at least the first version." (Mike, 05:51)

Rethinking the Development Cycle

• Embracing Rewrites & Iteration:

  • Not taboo to significantly rewrite products, thanks to AI acceleration.
  • "It’s no longer…a year long rewrite that might have killed a company like…Netscape…We’ve actually had several initiatives…built the full blown thing, realized we've overcomplicated…tore it down, done a V2…" (Mike, 08:59)
  • Prototyping and iteration cycles are measured in days or weeks, not months or years.

• Learning from Real User Contact:

  • The value of shipping minimally useful versions early to allow real-world learning and validation.

Agent-native Products: Philosophy and Implementation

• Definition and Principles:

  • "Agent-native" means that anything a user can do, an AI agent can also do seamlessly.
  • Importance of designing software primitives so agents can modify and operate on every function natively.

• Evolving the Paradigm:

  • Comparing Claude Code’s architecture (agent-native) to previous iterations.
  • "It should have knowledge about itself and that unlocks so much capability…how do you imbue the software that cloud builds to be more cloud aware…and even just cloud agent native…because it still won't [by default]…" (Mike, 12:48)

• Agent Native as a Test Case:

  • Hard to anticipate or test for all emergent agent-native behaviors.
  • Need for robust architecture to handle unexpected agent interactions.
  • "It's much harder to sort of write an end-to-end functional test around an agent native product because part of it is that unpredictability." (Mike, 16:49)

Proof of Use and Robustness

• Proof of Use as Value:

  • Preference for proof of actual use/demo (e.g., Loom videos) over just passing tests for feature validation.
  • "You want to have a playground within a safe environment. That's the only way you can have a playground is if it's safe around the edges." (Dan, 19:07)
  • "It's not just proof of work, but it's like proof of thoughtfulness. Like, did you think this through?" (Mike, 20:13)

• Robustness Over Time:

  • "Does it feel like it's built on sand or does it feel robust? And the agent native part adds something totally even beyond that." (Mike, 22:47)

Team Structure, Roles, and Changing Skills

• Smaller, Conviction-led Teams:

  • Early phases benefit from solo or paired founders who hold the context and drive conviction.
  • "Scaling the teams too quickly actually is a net negative…you [just] end up in this sort of meta coordination game." (Mike, 33:23)

• Shifting Skillsets:

  • With AI, hiring can now focus on great product sense and the ability to use AI tools, not just deep engineering.
  • "We just hired a new GM who's…lightly technical, but he spikes super high on product and writing sense…and now we can hire someone like that where a year ago we wouldn't have been able to." (Dan, 24:21)
  • Designers are moving into more hybrid builder-roles, writing as much code as engineers in some experiments.

• Agency-Like Resource Model:

  • Both Anthropic and Every use lightweight core teams, supplemented with shared designers, marketers, etc., as projects require.

Managing Features & Enterprise Adoption

• The Difficulty of Unshipping:

  • Need to be ruthless about removing features, but hard when serving enterprise customers.
  • "Deleting features as a sort of imperative…if this is not working, let's go unship that." (Mike, 35:33)
  • Legacy features may become load-bearing for certain customers; plugins/skills may allow keeping complexity out of the core.

• Enterprise vs. Startups:

  • Enterprise buyers may lag behind state-of-the-art; startups risk becoming outmoded if catering only to current needs.
  • "You have to be willing to put out the V3 or the V4…a big rethink…cloud can help…host both for a little while…but then also be willing to cut over…" (Mike, 38:50)

Personal Agents, Ownership, and Trust

• Emergence of Personal Agents:

  • Open-source tools like OpenClaw demonstrate how quickly users form personal/identity-based relationships with agents.
  • "My girlfriend's claw is called Shelly…there’s this thing that happens where it feels like it's mine…mirrors me in this way…" (Dan, 43:26)

• Social Dynamics of Agents:

  • Agents can reflect both the skills and trustworthiness of their owners, creating "shadow org charts" within organizations.
  • "...everyone has a claw, their claw becomes known for and used for the thing that they're specialized at that per their owner is specialized at…" (Dan, 46:39)

Notable Quotes & Memorable Moments

• "You can get it to go zero, not zero to one, but zero to end pretty quickly over the matter of hours."
  (Mike, 02:31)

• "If you grow a tree indoors, without it being exposed to wind, it doesn't get as strong…we've accelerated the pace of development so drastically, [but] you can actually kind of grow an entire tree indoors…doesn't have the same level of intuition and exposure."
  (Dan, 04:47)

• "Because vibe coding is so fun and so addictive, I just found myself being like, yeah, like I'll do this and I'll do this…and it just created this monstrosity that wasn't that good."
  (Dan, 07:09)

• "Instead [of adding the right feature], it just made for…something that felt really complicated."
  (Mike, 08:59)

• "Agent native means…the agent can do anything on your computer that you can do, and it's customizable and flexible and extensible."
  (Dan, 11:38)

• "It should have knowledge about itself and that unlocks so much capability in there as well."
  (Mike, 12:48)

• "It's not just proof of work, but it's like proof of thoughtfulness. Like, did you think this through?"
  (Mike, 20:13)

• "Scaling the teams too quickly actually is a net negative because they end up spending all this time on coordination…alignment conversations."
  (Mike, 33:23)

• "Being willing to delete code…they have deleting features as a sort of imperative."
  (Mike, 35:33)

• "It's like you have to be willing to put out the V3 or the V4 that is a big rethink of how the existing piece worked and then maybe have a transition period."
  (Mike, 38:50)

• "My girlfriend's claw is called Shelly. There's this thing that happens where it feels like it's mine, like it's really mine…has a personality that sort of like, mirrors me in this way."
  (Dan, 43:26)

Important Timestamps

• [02:31] – Challenges of AI-accelerated product building
• [05:51] – The pitfalls of overbuilding and lessons from Lean Startup
• [08:59] – Accepting rewrites and embracing iteration
• [11:38] – Defining agent-native products
• [12:48] – Claude Code and engineering for agent-native functionality
• [16:49] – The need for robust, unpredictable, and testable agent behaviors
• [19:07] – Proof of use, not just proof of work
• [24:21] – How AI is changing who and how you hire
• [33:23] – Team size, scale, and coordination challenges
• [35:33] – Feature deletion core to modern product thinking
• [38:50] – Startups, enterprise, and relentless product iteration
• [43:26] – The psychology of personal AI agents

Summary

This episode provides a candid, deep-dive into the craft of building agent-native products in the AI era. The discussion balances optimism about rapid innovation with hard-won lessons on simplicity, team dynamics, and the critical importance of product intuition. If you want a playbook for harnessing AI to shape the future of software, this episode delivers—while also acknowledging where human experience and taste are irreplaceable.