Brainwidth is brain power’s neglected sibling. Power is how deep you can go on one problem. Brainwidth is how many problems you can keep alive at once without dropping one.
People reach for “bandwidth” to mean this, but bandwidth is just total capacity. Brainwidth is the width of it: the number of open threads you can hold and switch between cleanly. And it’s smaller than you think. Working memory holds about four things, and you don’t run them in parallel. You switch, and every switch costs a reload. Multitasking is the myth; brainwidth is the honest version of the same wish.
It used to be a niche skill. AI made it the bottleneck. When a model executes in seconds, one person ends up steering five workstreams at once. The constraint stops being whether you can do the work and becomes how many live contexts you can hold before one quietly goes off the rails. The failure isn’t running out of power. It is running out of width and not noticing the thread you left open.
Like any capability, it is trainable, and like any capability, it has a limit you would be wise to respect.