A small garden of ramblings, notes, and half-built experiments. Tended carelessly and often.
If AI agents are going to do real work, the prompt is only one part of the setup. The interface around the AI matters more.
GitHub's Spec Kit gives AI agents structure. The Ralph Wiggum loop gives them persistence. Together, with the right prompt, they solve the two biggest problems in agentic coding.
LLMs can write code, but they're terrible at design. Here's why they default to purple gradients and three-box layouts, and what you can actually do about it.