How Anthropic Uses Claude Fable 5 With Mike Krieger
Anthropic Labs head and Instagram co-founder Mike Krieger talks with Dan Shipper about how to work effectively with Claude's new frontier model for complex, autonomous software tasks.
The gist
-
Krieger does architectural planning conversations with the model before jumping to implementation, then delegates overnight work and reviews completed features in the morning.
-
Verification loops using screenshots, videos, and real-world testing are essential because static tests alone miss issues like UI jank and edge cases.
-
A media tracker app that would have taken weeks of late-night coding took a weekend, but Krieger says cost discipline matters when choosing this model over cheaper options.
-
Ownership, architectural judgment, and production incident response stay with humans; the bottleneck has shifted from writing code to understanding intent and trade-offs.
◎
Related episodes
Episodes covering similar topics
The 4-Step Agentic Engineering Workflow I Use to Ship Real Apps | Ras Mic
software architecture
You're not behind (yet): How to learn AI in 22 minutes
productivity optimization
Microsoft’s Vision for an Internet Made for Agents With CTO Kevin Scott (Best of the Pod)
ai-assisted software development
VIBE CHECK: GEMINI 4—GOOGLE IS BACK
Oct 14, 2026
Working With Agents in Slack
Oct 10, 2026
OpenAI Dots Took Over Our Company
Oct 07, 2026
LIVE: DevDay 2026
Sep 30, 2026