From a business point of view, it's a terrible thing to just start coding on an idea only relying on gut instinct. From a personal point of view, it's incredibly relaxing to just pick a target product and start making it real.
I started building reflow.io as something different; automated unit test code generation. The idea was that by running code in a tightly controlled execution environment, you gain enough information to automatically write unit tests for that code.
Whilst building this was very cathartic, the interface to creating those tests was unpalatable. It was a slow, automated program that turned on/off breakpoints, gathering execution data and source maps. Once complete, it outputted a series of passing unit tests that mocked, passed in, and validated execution data as it was when the program was run. After being a little disheartened by this, I started working on reflow as the product-before-the-product.
Reflow is now something I'm proud of: it's a no-code UI record/replay test automation tool with visual regression support. It can be run entirely on-premise, allowing a development team to easily create and maintain resilient end-to-end tests against a locally (or behind a firewall) running web server. It is also available online with no dependencies, allowing anyone in an organization to contribute to the testing process too.
I've opened it up to public registrations with the hope it will help someone out.