Same task. Same models. One has your context. This runs on a demo engineer's Context Key — an authored example self — so you can watch the membrane work before you write a single atom of your own.
Pick a task above — watch every model answer twice: once cold, once with the self's context.
Point a Context Key at your own self and every tool — Claude, Cursor, ChatGPT, your own agents — answers like the right column, everywhere.
npx @ctrlai/cli init