On Chaos and Turning Inward
Musings on uncertainty and the virtues of control, transparency, and small communities.
Read the postRNG.md posts
Archive
Maybe It’s Time To Start Looking at Local Coding Models
Speed, cost, and predictability are starting to matter more to me than top-end reasoning.
Use yes to reproduce flaky tests
I use yes to saturate CPUs and reproduce flaky tests failures. Pair it with a small loop helper to rerun the test until it fails.
The Waymo Rule for AI-Generated Code
AI-generated code doesn't have to be perfect; it just has to be better than you.
Code Quality Gates for Vibe-Coded Projects
A CLAUDE.md for vibe-coded projects. Defines agent-enforced quality gates for tests, lint, complexity limits, security checks, and performance regressions.
Open Source Projects Are Cached Agent Output
Open source projects should become customizable kernels that save users time and tokens.
Something Is Happening
Building things sparks joy. Everyone I talk to is shaking with excitement. I'm not sure where this goes, but it sure is fun.
Simplifying AI-Generated Code With Lizard
I use lizard.py with a wiggum loop to clean up AI-generated code. The post shows the exact prompt and the Lizard targets I use in wiggle.
Migrating SlateDB Bindings To UniFFI
After landing SlateDB bindings, several people asked about my experience with UniFFI. It's young and uneven, but still the right bet. Particularly so if you need async across runtimes.
Wiggum Loop
I show the zsh Wiggum Loop I use to rerun Codex against PROMPT.md and the prompt that built uniffi-bindgen-node-js in 79 commits. It works for code when the prompt enforces TODOs, tests, and commits; I have not found it useful for docs.
No matching posts.