Building Is the Easy Part Now | Mike Krieger on What AI Changed
Mike Krieger, Instagram co-founder and head of Anthropic Labs, discusses how AI models have fundamentally changed product development—making building easier but strategy harder. The episode explores the paradox that while engineers can now prototype features in hours, the critical challenge remains knowing what to build and what to cut, requiring deep product intuition that can't be accelerated by AI alone.
Key takeaways
- • Building is now trivial, but product strategy remains hard—AI models are excellent at adding features but poor at determining what should be removed, making the ability to simplify and focus more valuable than execution speed.
- • Agent-native architecture should be baked into products from day one, allowing agents to perform any action a human user can perform, rather than treating agent capabilities as an afterthought or constraint.
- • Products built too quickly without user feedback become "indoor trees"—overbuilt and fragile—so launching minimal versions early to test real-world assumptions is still essential despite accelerated development cycles.
- • Small, founder-led teams stay more agile during rapid AI iteration cycles; adding people too early creates coordination overhead and slows down the frequent rewrites necessary as models improve every few months.
- • Robustness and architecture still require senior technical expertise—distributed systems knowledge, database design, and thoughtful tool architecture cannot be outsourced to AI and remain critical for products that feel reliable rather than fragile.
- • The unit of value in code review has shifted from "tests passing" to "proof of use"—requesting videos or agent demonstrations showing the feature working as intended, rather than relying on automated test suites alone.
Recommendations (7)
"I think that Claude Code is the canonical thing that taught me about how that kind of product can work so well where it's like it's an agent."
Every · ▶ 12:08
"I loved your write up on agent native and I was like I want this as a skill. So whenever I'm prototyping something it thinks in an agent native way. So I had it packaged it up as a skill."
Mike Krieger · ▶ 15:05
"One of the things that we like have now is a skill about the Claude API, which sounds super obvious, but even just having that is really valuable."
Mike Krieger · ▶ 17:11
Mentioned (2)
More from these creators
Most SaaS Companies Got AI Wrong. Linear Waited.
What Happens When Beginners Start Building With Claude Code—With Mike Taylor and Kate Lee
Reviewing Everything on my Desk! (2026)
How We Use Proof, a Collaborative Editor for Humans and AI
Vibe Check: GPT-5.4—OpenAI is Back
Meet the Slowest Startup Incubator in the World—Pumping Out Billion-dollar Companies