Zero-Cost Public Roadmap

Replacing a deprecated vendor tool in under a day using Claude Code.

Complete Claude Code GitHub Pages Google Sheets 0→1

Stack: Claude Code · Google Sheets · HTML/CSS/JS · GitHub Pages
Timeline: Idea to production in <1 day


Screenshot of the Blackthorn public roadmap

The Problem

The product team was changing our customer feedback platform. The new solution did not include public roadmap functionality. The company had always maintained a public roadmap that was heavily used as a reference by sales, customer success, and support teams.

The previous solution had multiple problems:


The Requirements

Any replacement needed to meet all of the following:

The obvious options were to procure a new SaaS tool (cost + procurement time) or request an internal engineering build (time + capacity). Both had tradeoffs. I saw an opportunity to solve it myself.


What I Did

I took the initiative to build a solution myself using Claude Code, leveraging existing tools and processes to reduce friction for the product team.

Total time from idea to deployed solution: less than one day.


The Result

The final solution was a success praised by leadership:


What I Learned

AI-assisted coding is a product tool, not just a dev tool. The value here wasn't writing code faster. It was eliminating the procurement process entirely. A PM who can vibe-code a solution in hours changes the decision calculus for the team: instead of "should we buy or build?", the question becomes "can we validate this ourselves before committing resources?"

The best solution isn't always the most technical one. A Google Sheet + static HTML + GitHub Pages is not a sophisticated stack. But it met every requirement, cost nothing, and the product team could maintain it without any engineering support. Simplicity was the feature.

Know when vibe-coding is the right tool. This worked because the scope was narrow, the requirements were clear, and the solution didn't need to scale beyond a single page. For a multi-page app with auth and database needs, the approach would be different. The judgment call is knowing which problems fit.