Before I wrote much code on my current project, I wrote documents.
The idea started after some frustration with social media. I spent time comparing different platforms and noticed how similar they all felt in practice. Most of them optimize for quick reactions and passive engagement. I kept wondering what kind of interaction would actually feel satisfying to respond to, where the creator might genuinely care about the reply, and where responding felt better than just clicking a like.
I gave the idea space by discussing it back and forth with different AI models. Over time this turned into a large, unstructured dataset of thoughts, constraints, and possibilities. Instead of trying to code from that directly, I treated it as raw material that needed processing.
Before doing that processing, I focused on orientation. As a programmer, what felt important to me was defining a Table of Contents in a README, writing a CONTRIBUTING guide, and creating an Agent.md file as an explicit entry point for prompting and discussion. These weren’t meant to lock decisions, but to give the project a shared mental baseline, even if that baseline was only for myself.
I wrote down best practices, architectural boundaries, coding rules, and style decisions as early constraints. I then refined those documents iteratively, again using AI models, before moving deeper into implementation.
These were the first steps I took before I could really start building.
I’m curious how others approach this phase. Do you do something similar, or are there other practices you’ve found useful early on?
Top comments (0)