By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
At a glance
WHAT IT’S REALLY ABOUT
How Fable 5 changes building: delegation, verification, and agent-native apps
- Mike Krieger explains how extended, day-to-day use of Fable changes work from turn-by-turn prompting into delegating longer, higher-context tasks that can run for hours or overnight.
- He describes a workflow shift toward earlier architectural planning with the model and using generated artifacts (documents/diagrams/HTML) to create team alignment before rapid implementation.
- He contrasts Fable with Sonnet, noting he prefers Sonnet for quick questions and reserves Fable (often with tuned effort levels) for deep, high-leverage work where follow-up turns would otherwise pile up.
- He demos a weekend-built media tracker that reflects “agent-native architecture,” including a mechanism to preview and apply code edits from within the app itself via managed agents.
- He outlines how Anthropic teams adapt through ownership (DRIs), parallel Claude sessions, dashboards to track agent work, and stronger verification practices to maintain accountability in production.
IDEAS WORTH REMEMBERING
5 ideasUsing Fable effectively requires rethinking task decomposition and time horizons.
He described moving from short, tightly guided prompts to expressing intent and delegating larger chunks of work, including setting tasks to run for hours or overnight and reviewing results later.
Up-front architecture and alignment become more important as build speed increases.
He said he now spends more time planning with the model early, iterating on architecture, and then having it generate shareable artifacts like HTML/markdown/diagrams to align humans before execution.
Model choice and effort settings matter: Sonnet for speed, Fable for heavy lifting.
He said he uses Sonnet for fast, low-stakes questions because Fable “thinks really hard,” and he also adjusts “effort levels” when a task doesn’t warrant maximum deliberation.
Agent-native architecture evolves from tool access to software that can update itself.
In his weekend media-tracker project, he described an “agent-native” approach where the agent can not only operate the app through tool calls, but also propose code edits from inside the product with previews and diffs.
The collapsed build cost expands who can build, not just how fast experts ship.
He framed the big shift as bringing “the thing that’s in my head” closer to “the thing that exists in the world,” citing an internal example of a recruiter feeling newly empowered to build tools without waiting on overloaded engineers.
WORDS WORTH SAVING
5 quotesI sat there and I, and I was like, "I feel like a total newbie again," because I feel like the way that I am prompting or even thinking about decomposing a task is really out of date now with this model.
— Mike Krieger
I think the most impressive thing for me is like you're just being able to like delegate that kind of level of task and just trust that the right thing will happen by the end.
— Mike Krieger
I should never have to like navigate a menu to do anything ever again.
— Mike Krieger
It is the first time in my life, and she works in recruiting. She's like, "The first time in life where like I feel like the thing that's in my head and the thing that exists in the world is now like, uh, they're right next to each other, like I can just do it."
— Mike Krieger
You would dream about code, and if you ever had the experience of, like, you would dream about the thing that you're working on, then you, like, uh, wake up in the morning and be like, "Ah, I figured out how to solve this thing really elegantly." Um, and, and that for sure has, has, has passed.
— Mike Krieger
SPEAKERS
Dan Shipper
hostWriter and entrepreneur; host at Every.
Mike Krieger
guestHead of Anthropic Labs and co-founder of Instagram.
High quality AI-generated summary created from speaker-labeled transcript.
