A small experiment in honest measurement

After the demo.

The first draft is only part of the work.
Did AI actually save you time?

Your evidence, at a glance

The workbench

Net time saved

—

Your first measurement starts here

Quality pass rate

—

Set the same bar for both methods

Total tool cost USD

—

Incremental cost across your trials

No observations yet. Start with one task you can measure both ways.

Record a trial

Same task. Same scope. Same quality bar.

What did you work on?

Use the same name to track repeated attempts.

Without AI baseline

min

Time the equivalent completed task without AI. Prefer an observed baseline to a guess.

With AI include the whole attempt

Prompting

min

Generation / waiting

min

Review & verification

min

Corrections & rework

min

Use non-overlapping minutes. Include waiting only when it occupies or blocks your work; enter 0 for a stage you did not need.

Did the final result meet your quality bar?

Passed Not yet

Tool cost USD, this attempt only

Use incremental cost, or a consistent share of a subscription. Keep one method across trials.

Quality bar & observations optional

Saved only in this browser. No uploads. No model calls.

Where the minutes went

A fair comparison includes the invisible work.

Your baseline and full AI effort
will appear side by side.

Prompting
Waiting
Review
Rework

Review and rework are work. Keep them in the total.

What is that time worth?

Optional lens

Put a value on the time you measured. No annualized promises.

Your time value USD / hour

$/ hr

Measured time value − tool cost—

Add a trial and an hourly value to see this estimate.

Break-even is shown only when measured time saved is positive.

Repeated work reveals the pattern

Trial log 00

No grand claims. Just your first trial.

Record a task above, then try it again. Small samples are observations, not conclusions.

Exports contain your real trials only. All costs are USD.

A little rigor goes a long way

Measure the finish.
Not the first draft.

Baseline − (prompting + waiting + review + rework)
= net time saved

Define “done” first. Use equivalent scope and the same quality check for the baseline and assisted attempt. Choose a narrow, repeatable task.
Log the full attempt. Count each minute once. Include failed attempts and their cost. A negative result means AI took longer.
Keep quality beside speed. Net time saved includes every attempt, even quality failures. A faster failed result is not a usable deliverable.
Repeat before you generalize. Learning, task difficulty, and order affect results. Try multiple comparable tasks and alternate methods. This log is descriptive, not a causal study or a forecast.

How the optional value estimate works

Estimated value = (net minutes saved ÷ 60 × your hourly value) − tool cost. Break-even hourly value = tool cost ÷ hours saved, available only when hours saved are positive. Neither calculation adjusts for output quality, unpaid work, opportunity cost, or whether you can actually use the saved time. Both apply only to the trials recorded here.