Getting started with spec-driven development · Start here

This is the start-here primer for a series on building software with AI coding agents. Parts 1 to 3 build real features and already use words like nondeterministic, context, drift, and skill. This video defines those words first, so a beginner can follow the rest without getting lost. It begins with what a coding agent actually is: an LLM that runs in a loop and can read your files and run commands on your machine. From there it covers why the same prompt gives a different answer each time, why the context you give the agent matters more than the exact wording, how the context window fills up and starts to drift, and what a skill is. There's no live build in this one. It's the groundwork, the handful of terms the rest of the series leans on. The process works in any coding harness, and I happen to use Claude Code. The full write-up, with every link, is on merc.camp for members. Come build with us. ▸ Part 1, superpowers: https://youtu.be/gxYYrzst1so ▸ Part 2, project management: https://youtu.be/qZPFxi9fmSE ▸ Part 3, the dev-suite: https://youtu.be/odzkQgdT3ok ▸ Full write-up on merc.camp: https://merc.camp/experiments/fundamentals ▸ Twitch (I stream the dev): https://www.twitch.tv/n3rkmind ▸ Discord: https://discord.gg/zFZE7mfGke ▸ merc.camp: https://merc.camp Chapters 0:00 Start here 1:05 What a coding agent is 3:04 Why it's nondeterministic 4:12 Context is king 5:32 The context window and drift 6:53 What a skill is 7:45 What's next If you're building something and want a team on it, that's what we do at Warden Labs: https://wardenlabs.ca/en