A couple of hours ago I said I was working on a coding-agent benchmark mid-flight.
The idea came to me during a Claude session where I gave the agent a key to a relatively undocumented API used within a codebase and asked it some questions.
Magically, it came back with answers.
Currently flying to SF from Madrid, with a layover in Washington ahead of me, and decided to finish a side project I had been working on a long time ago and never got to finish, mid-flight, while posting random stuff as I do it here on X.
Hint: It's a benchmark for coding agents
Second hint: No clue who the guy behind me is
Will keep you posted 😏