Fable 5 GEOExtension Micro SaaS Prototype

I used Fable 5 to turn a domain I had been sitting on, geoextension.com, into a working micro SaaS prototype.

The idea is GEOExtension: a Chrome extension that helps small and mid-size businesses understand how visible they are inside LLM answers. If someone asks ChatGPT, Gemini, Claude, or Perplexity for a local recommendation, this tool should help a business understand whether it shows up and what to improve next.

This was not a perfect build. That is what made it useful. The prototype loaded, the landing page existed, the onboarding started, and then real testing exposed the next set of fixes.

Start With Plan Mode, Not Code

The most useful part of this Fable 5 micro SaaS build happened before code.

I did not start by saying, "Build me a Chrome extension." That kind of prompt is how you get random features, fake dashboards, and an app that looks done before it works.

Instead, I started in plan mode with three simple details:

• What I wanted to build
• Who it was for
• What outcome the tool needed to create

For GEOExtension, that meant a Chrome extension, landing page, and lightweight web app for small businesses that want to improve LLM visibility.

The outcome was specific: help businesses understand what they need to do to reach the top five mentions for their target keywords, industry, and location inside LLM answers.

That gave Fable a product direction before it made technical choices.

The GEOExtension V1 Scope

Version one stayed intentionally small.

The build included a Chrome extension, a simple landing page, and a lightweight web app. The extension would run on a business website or social profile, analyze the page, and help the owner understand their GEO readiness.

GEO here means generative engine optimization. In plain English: how visible your business is when buyers ask AI tools for recommendations.

The first version uses a bring-your-own-key model. That means users provide their own provider keys instead of GEOExtension handling all usage and billing from day one.

Fable planned support for:

• OpenAI
• Anthropic
• Gemini
• Perplexity

One constraint mattered a lot: the app should not break if a user only has one API key. Some users will only bring OpenAI. Some will only bring Gemini. Version one needs to work anyway.

Why Zero Demo Data Matters

I also gave Fable one hard rule: zero demo data.

When you're vibe coding, AI tools love to add fake users, fake charts, fake reports, and fake companies. That can make a prototype feel more finished, but it can also hide broken product logic.

For this build, I wanted to test from version one.

That meant no fake dashboards pretending everything worked. I wanted a prototype that I could load, run, and break with real inputs.

That choice paid off quickly.

What Fable 5 Got Right

Fable built the core pieces in a little over an hour, including time lost to model outages.

The extension had a simple logo, which matched the minimal direction I gave it. The landing page was clean. The web app gave the product another way to run reports outside the extension.

The plan also left room for the future business model. Version one has no billing, but it does not block a paid version later. That matters because GEOExtension can start as a bring-your-own-key beta and grow into monthly or annual passes.

It was not the flashiest thing Fable could have built. But it was practical, and that was the goal.

What Broke During Testing

The extension did not load correctly at first. Chrome said the manifest was missing or unreadable. After checking the project, I found the Chrome extension output inside a hidden .output folder.

Once I loaded the right folder, the extension worked enough to test.

I used a real local business search example: HVAC Las Vegas. I entered OpenAI and Anthropic keys, moved through onboarding, and hit the first real bug:

No active tab found.

That bug matters because the extension needs to understand the page the user is currently viewing. If a business owner opens GEOExtension on their website, the tool should pull context from that site so the owner does not have to manually fill everything out.

This is exactly why real testing matters. A screenshot would not reveal that problem. Using the extension did.

The Product Roadmap From Here

The next version needs to move from prototype toward production.

That means fixing the active tab issue, improving the landing page, adding a backend for saved reports, and deciding how the paid version should work.

There are two likely paths:

• Keep bring your own key and charge for the software layer
• Offer a paid plan where users do not need to manage provider keys

The landing page also needs a clearer CTA. Instead of pushing people into a manual free report first, the page should probably push users to download the extension so GEOExtension can analyze the page directly.

That is the real value: business owners should not need to know the perfect keywords or prompts. The tool should help find the right keywords based on their business, location, and services.

My Takeaway From This Fable 5 Build

The lesson is not that Fable 5 magically ships perfect products.

It does not.

The lesson is that AI coding gets much more useful when you guide the product before the build starts and test the real flow as soon as possible.

The working pattern is simple:

1. Start with the product purpose
2. Name the customer
3. Define the outcome
4. Use plan mode before code
5. Add real constraints
6. Avoid demo data
7. Test the actual flow
8. Fine-tune what breaks

That is how a parked domain becomes a real product direction.

The goal for Clearmud now is one new micro SaaS or product per month. GEOExtension is first, and the next step is shipping it to production.

Subscribe to the Clearmud newsletter for more build-in-public notes, and watch the full video for the complete Fable 5 planning and testing walkthrough.

🎥 Full video: https://youtube.com/watch?v=B9tMrTe1fmc