What Did That Cost?
Drive AI agents through a million lines of code you did not write. Recon, surgical navigation, codemods, characterization tests, and a toolkit of four reusable skills you keep.
I keep saying token use is top of mind, and you have watched me hit /clear over and over. So let us look at the meter. What did all that mapping actually cost?
/usage, current session. My session started at 11am and I have burned 5% of it building all of this. Current week across all models: 3%. Well under the cap. Fable is at zero, since I have been on Opus.
Fair disclosure, I have small side projects running that feed into the weekly number. The session number is the one to watch.
Then /context. The Copilot map session has used almost nothing. I zoom back to yesterday's git history session and check that one too. Also low.
That is the target. I want to live inside the first bar. If I am on a roll I might let it run to halfway through the second. That is a personal rule, not a law.
I know I have a million token context window. I know Opus is good. But every token I throw at it costs money and raises the odds of a hallucination. Tight, short, controlled sessions.
The whole point of doing the mapping up front is that the real work should now be cheaper.
Time to find out. Next I create a branch and start ripping.
Drive AI agents through a million lines of code you did not write. Recon, surgical navigation, codemods, characterization tests, and a toolkit of four reusable skills you keep.