AI-assisted development is fast, but it can leave you with a codebase that feels foreign a few months later. The problem isn’t the code, it’s the cognitive debt. Here are some lightweight documentation habits that hold up in an AI-assisted workflow.
A walkthrough of building TestGenAI, a Ruby CLI gem that identifies testing gaps and backfills them.
Cognitive debt is the mental model that never got built, the reasoning that never happened, the understanding that never got built. LLMs are great at generating it, and it’s disturbingly easy to accumulate. Here’s how to recognize it and practical ways to keep it in check.
Another weekend MVP story about building and publishing JoyConf, a live emoji reaction app for presentations. The why, the architecture, the implementation details, and the bugs. If you’re curious about building real-time interactive web apps with Elixir this is a fun case study.
A weekend MVP story about building and publishing MockOpenAI, a Ruby gem for testing OpenAI-compatible APIs locally. From idea to published package in four days, including the moment of doubt about whether it was solving a real problem.
When to use traditional search, basic RAG, agentic RAG, or a full conversational agent for your Rails app’s search feature
