INITIALIZING
INITIALIZING
INITIALIZING
Claude Code. Codex. A million alternatives popping up every week. The harness changes exponentially. I keep catching myself getting attached to one, then watching it stop serving me. The harness is liquid. Whatever works today might not tomorrow.
Hot-swappable. Cheaper and more specific every day. The model is the brain—and I find myself swapping it out the moment a better one shows up. No model loyalty.
Code is nearly free and nearly instantaneous. AI generates it, modifies it, discards it. The files on disk are liquid—they flow, shift, disappear. I've stopped treating code as sacred. It's the output of intent, not the intent itself. The spec is what matters.
The runoff from specs. If the schema shifts, the docs flow to match it in the same breath. I've stopped treating docs as sacred artifacts I maintain by hand. Auto-generated. Ephemeral. Swappable.
They feel like craft because they fit the current model so well—until the next model leap makes half of them noise it already knows. As models improve, the walls evaporate—and so do the input tokens I spent building them.