⚙️ Your Parameters
1010K100K
$0.001$0.50$2.00
1%45%90%
5100200
Benchmark — From 99,167 real sessions
Waste % = steps classified as LOOP or STAGNATION / total steps
Defaults based on SWE-bench resolution rates (~22%) and typical GPT-4o API pricing.
Adjust to match your actual workload. Full methodology →
💸 Your Waste Estimate
Annual compute waste
$0
per year in wasted API spend
Daily sessions
1,000
Failed sessions/day
780
Wasted steps/day
—
Daily waste cost
—
Monthly waste cost
—
% of budget wasted
—
Detection confidence
Cohen's d = +1.099
Upload a trajectory file and see the exact waste breakdown for your actual agent sessions — with cryptographic proof.
Analyze My Trajectories →