Notes from Building My Second AI Video Site

Source: Dev.to

Introduction

Soraviz is my second AI video project. When I started it, I thought I had learned enough from my first AI product to move faster and avoid obvious mistakes. In some ways that’s true, but building the second one reveals a different set of problems.

Current Phase

Right now I’m in a phase where I’m not adding much. Instead, I’m reviewing what already exists.

• Some features work, but feel rough.
• Some ideas sounded good early on, but don’t hold up in real use.
• Some issues aren’t about features at all — they’re about reliability, speed, and expectations.

Key Decisions

This part of building is quiet and a bit uncomfortable. There’s no clear “launch moment,” just a lot of small decisions:

• What’s worth keeping?
• What should be rebuilt?
• What doesn’t deserve more time?

Future Plans

I’m considering a proper upgrade of Soraviz — not to make it bigger, but to make it more solid. The goal isn’t to chase the next AI trend; it’s to build something that can survive longer than a few hype cycles and slowly become more useful.

Reflections

I don’t have strong conclusions yet, just notes, doubts, and a growing respect for how hard it is to make AI products reliable. If you’re also building in public, this phase might feel familiar. Sometimes progress looks like slowing down and cleaning up.