Ai on the art of simplicity

Two Ways to Stop Your Coding Agent Burning Tokens: RTK and CodeGraph

Sun, 28 Jun 2026 00:00:00 +0000

Two Ways to Stop Your Coding Agent Burning Tokens: RTK and CodeGraph

I keep a few Claude Code and Cursor sessions running through the day, and at some point I started actually watching the token counter instead of ignoring it. The thing that struck me wasn’t how much it climbed. It was what it climbed on. Most of the budget wasn’t going on the model reasoning about my problem. It was going on plumbing: ls, grep, reading the same file three times to rebuild a call path, scrolling a 400-line cargo test dump to find the one assertion that failed.

The Harness Engineering Cheat Sheet

Tue, 28 Apr 2026 00:00:00 +0000

The Harness Engineering Cheat Sheet

I have a Cursor problem. Or maybe an “agent session” problem more generally — I keep five or six chats open, each with its own slice of context, each in a slightly different state. We have nice skills. We have nice tools to fetch the right context. We have a code-review skill that drops inline comments on a PR. But I still lose track of which window is doing what, and occasionally paste the wrong thing into the wrong session.

I Built an AI Ultrarunner to QA My App

Wed, 11 Mar 2026 00:00:00 +0000

I Built an AI Ultrarunner to QA My App

I’ve been building a race tracking app for ultrarunners. It lets you manage your race calendar, track gear, write journey entries, and plan your season. The problem? I have no users yet. Just me.

That means nobody is poking around, trying weird things, or stumbling into edge cases. The app has unit tests, CI checks, type checking — all the usual stuff. But there’s a whole class of bugs that only surface when someone actually uses the thing. Opens the browser, clicks around, creates records, and goes “wait, that doesn’t look right.”