I've been running Headroom on my coding-assistant work. On Nov 10 my dashboard showed 32.1M tokens saved and about $124 in estimated value.

I want to be straight about that number: it's Headroom's estimate of usage avoided, not a refund and not proof my monthly bill dropped. Your results will depend on how much repetitive tool output your assistant reads. If there's little to shorten, you'll save less.

What I like: it shortens bulky tool results before the AI reads them, the compression is reversible, and the dashboard makes the savings visible.

I explained it with one simple example, plus my own dashboard, in this video: https://youtu.be/OePHFrlJbXo