Observe the work
What did the workflow
actually take?
Record attempts, review the outputs, and compare the human effort behind accepted work.
Nothing saves automatically. Save on this device stores your notes in this browser profile, where others using it can read them. There is no account, cloud sync, AI call, or automatic publication. Keep names, contact details, credentials, and confidential material out of the notes. Download a backup before clearing browser data.
One task. One quality standard.
Each entry is one attempt at the same-sized output. Use comparable inputs, not an easier AI task. Record failed attempts too. Count setup once where it happened, plus briefing, production, review, and corrections. Track elapsed waiting time separately from your own effort.
Decide what acceptable means before timing. Try different run orders and more than one example to reduce practice effects. A single pair of runs is a starting observation.
Recorded attempts
What the records show
Human minutes per accepted output = all logged effort for that method ÷ its accepted outputs. The ratio divides manual effort per accepted output by assisted effort per accepted output. All logged setup and failed attempts are included. Missing effort, pending reviews, or no accepted output prevent a comparison.
This is your recorded sample, not an independently verified result, elapsed speed, or earnings promise. The productivity calculator is separate for exploring future scenarios.