Hey Product Hunt, I built SpecDD because AI kept forgetting what we were building. Not literally forgetting the chat. More like forgetting the shape of the product. It would write code that looked fine, but missed the point. It would ignore a boundary I had already explained. Add a dependency I did not want. Solve the ticket while quietly breaking the idea behind it. After enough of that, I stopped thinking the problem was “the model is not smart enough.” The problem was that the context had nowhere useful to live. And yes, I know the obvious answer is “just use docs.” But be real. A folder full of Markdown works about as well as dumping Confluence on developers. Which is to say: not very well. Nobody wants to read hundreds of pages of documentation before making a change. AI has the same problem. It does not need more walls of text. It needs the right context, close to the thing it is changing, in a format it can actually use. That is what SpecDD is for. It is an open-source framework for specification-driven development with AI. Small specs live close to the code and describe what the code is supposed to do, what it should not do, and what “done” means. Before, during, and after implementation. The aim is not to make AI write more code. It already does that. The aim is to make it write code that is closer to what you meant. I am launching it here because I think a lot of people building with AI are running into the same thing: fast output, but too much re-explaining, correcting, and cleaning up afterward. Would love to hear where this breaks in your workflow.

SpecDD

Build better software with spec-driven AI

Build better software with spec-driven AI