Notes on context windows you'll never fill.

Million-token windows arrived faster than the workloads to use them. A working theory of how attention budgets should be priced.

This essay's full text is forthcoming. The design system is in place; the words are still on the desk.

In the meantime, the rest of the archive is over here.

Written by the editor of Vercilo.

I write about the parts of computing that don't make it into the keynote. Past lives in systems engineering and ML research. New essays land roughly every two weeks; no recommendations or "what we're reading" — just the work.