AI-native development

The plan is the product: how I actually run Claude Code

Most people drive a coding agent by prompting and praying. I run a loop where the code is the last and cheapest step, and a plan document I have argued with is the artifact I actually labor over.

Most of your AI coding bill is cache, and how to stop breaking it

When your quota vanishes twice as fast as it should, the culprit is almost never how much you generated. It is how often you broke the cache without noticing.

When I reach for a team of agents, and when one is enough

Running five agents in parallel looks like five times the output. Usually it is five times the tokens and a coordination problem. Here is the line I draw.

Git worktrees are the right tool for parallel agents

Your agent doesn't need a better memory. It needs to forget.

The instinct is to give an AI agent perfect recall of everything you have ever done together. I tried the heavy memory tooling and turned it off. The problem was never storage. It was retrieval.

The 200-line rule: why my agent instructions stay short

A long CLAUDE.md feels responsible and reads like diligence. In practice it buries the rules that matter under the ones that do not. Here is the structure I use instead.

The bottleneck moved: what actually changed when AI started writing my code

After a year of building production systems with AI in the loop, the hard part stopped being the code. It became the spec, the context, and the judgment to tell good output from plausible output.