Anthropic Expert: Rethink What's Possible With Fable 5
TL;DR
Fable 5 changes the unit of work from prompts to delegation: Krieger says he now sets up complex tasks before bed, "wishes Claude a good night," and expects it to complete the full swing, including workarounds when a service goes down.
The best use is often upfront planning, not just coding: He spends more time having architectural conversations, generating HTML mockups, markdown docs, and diagrams so teams can align before Fable executes chunks of work in parallel.
He actively routes tasks across models because Fable is slow and expensive: Quick questions on iOS go to Sonnet, while Fable is reserved for long-horizon, high-context work where the total cost can be lower because it avoids 9 or 10 corrective follow-up turns.
A weekend side project showed agent-native software in miniature: Using Claude Code in the terminal, Krieger built a personal media tracker whose in-app chat can not only add items by URL but also modify the app itself through managed agents and Vercel live previews.
Software engineering is not over, but the craft has shifted upward: He says the old work of typing code and fixing framework quirks is collapsing into product and systems thinking, while humans still own intent, verification, incident response, and production accountability.
Verification is now a first-class skill: Krieger wants every AI-generated PR to include screenshots or video, uses staging flows with real data, and even gives Claude video captures plus ffmpeg so it can spot UI jank a screenshot would miss.
The Breakdown
Mike Krieger says Fable 5 is the first model that feels less like a copilot and more like a teammate he can hand overnight work to, with enough judgment to reroute around outages, push back in code review, and finish the job by 2 a.m. He argues the real shift is not just better code generation, but a collapse in the distance between intent and execution that changes who can build software at all.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.

Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.

Playbook
The Art of Tasteful Prompting
Learn how tasteful prompting helps you move beyond generic AI output by shaping context, style, and judgment from the start.