May Contain Jab
← All episodes

Episode 9

Is Claude Fable Worth the Hype? Guardrails, Sub-Agents & the LLM Wiki Every Owner Needs

October 27, 2026 ·29 min

Is Claude Fable worth the hype? We test it on real small-business builds, then explain LLM Wikis vs RAG in Obsidian in plain English.

Show Notes

Is Claude Fable worth the hype? May spent the week building real projects with Anthropic's Claude Fable: website rebuilds, back ends and a personal LLM Wiki. In this episode of May Contain Jab: The Owner's Edge, she and Jab whiteboard what actually happened: where Fable is genuinely better, where its new guardrails get in the way, what it costs, and the knowledge-management setup every small business owner should know about.

May's headline lesson: "It's not the tool, it's what you do with it." Pick the AI tool you like, stick with it, and put your energy into how you use it.

Key takeaways

  • Claude Fable reasons better than earlier Claude models and is less likely to break a project when you make changes.
  • Since its relaunch, Fable has stricter guardrails. Anything touching emails or personal data can push you back to Opus.
  • A workaround May has tested: run Opus on ultra mode and have it use Fable sub-agents in Claude Code. You can't fully audit what the sub-agents do.
  • Fable is credit-intensive, so budget for it before you build with it.
  • An LLM Wiki turns your raw business documents into an organised, searchable "personal Wikipedia" you can view in Obsidian.
  • RAG and LLM Wikis both make years of SOPs, transcripts and processes findable in seconds.

Chapters

  • 00:00 Welcome: vibe coding with Claude Fable
  • 02:30 It's not the tool, it's what you do with it
  • 03:09 Whiteboard: putting Fable on the track
  • 03:50 Fable's strengths: reasoning and building websites
  • 05:31 Does Claude Fable live up to the hype? Guardrails and privacy
  • 06:54 The trick: Opus ultra mode + Fable sub-agents
  • 08:47 Before vs after the ban
  • 09:07 Why Fable was restricted
  • 09:48 Is the government killing the vibe?
  • 11:08 Let us party: the Google AdWords lesson
  • 12:29 Credit-intensive: the cost of Fable
  • 13:44 What is an LLM Wiki?
  • 15:44 RAG explained: retrieval augmented generation
  • 17:00 "Detective May AI": searching your business files
  • 19:22 LLM Wiki vs RAG: two structures
  • 20:54 Demo: May's LLM Wiki in Obsidian
  • 22:13 Why owners need this: SOPs, onboarding and process
  • 25:43 How to get in touch
  • 27:27 Coming next episode: Codex and ChatGPT
  • 28:08 Recap

Tools and resources mentioned

Claude by Anthropic: https://www.anthropic.com/claude

Anthropic's statement on Fable access: https://www.anthropic.com/news/fable-mythos-access

Obsidian: https://obsidian.md

The website stack May used: Vercel, Sanity, Formspree and Kit

About the hosts

May Zhan specialises in AI automation and strategy, and builds and tests AI systems hands-on before recommending them to business owners. Jab Siwela is the founder of Revalty Digital Lift and a marketer whose background includes eBay and PayPal's marketing office, focused on audience, resonance and revenue.

New to vibe coding? Start with our earlier episode, Vibe Coding: The Owner's Unfair Advantage. Next episode: Codex and ChatGPT.

Questions, or want help applying this in your business? Email mayandjab@maycontainjab.com or find May Zhan and Jab Siwela on LinkedIn.

Is Claude Fable Worth the Hype? Guardrails, Sub-Agents & the LLM Wiki Every Owner Needs

Is Claude Fable worth the hype? It's the question every AI-curious business owner is asking, so May put it to the test. She spent the week building real projects with Anthropic's Claude Fable, from website back ends to a personal LLM Wiki, and in Episode 9 of May Contain Jab: The Owner's Edge she and Jab break down what actually happened. Here are the key points in writing. For the full story, including May's live LLM Wiki demo, listen to the episode.

Never miss an episode.

Join the list for new episodes, behind-the-scenes notes, and early access.

No spam. Unsubscribe anytime.