Compare an AI side project using time, receipts and repeatability

Use the same measures for every project

Thanks to Tigerless Labs for sponsoring this video.

The video compares KDP books, a faceless Psalms channel, a small SaaS and YouTube. The reported earnings and time belong to Luke’s examples at that date.

1. For each project, record setup time, time to first revenue, cumulative revenue and ongoing work.
2. Distinguish revenue from profit and include services or subscriptions needed to operate it.
3. Keep estimated agent hours separate from your own working time.
4. Mark missing figures as unknown rather than manufacturing a comparable total.

Look at the work behind automation

Luke describes a copyright-related rework problem, a fully automated publishing process and a SaaS with one paying customer. Automation did not remove every operational task.

1. List the source material and permission checks your own project depends on.
2. Record review, repairs and maintenance alongside generation time.
3. For a SaaS, distinguish signups, paying customers and continued use.
4. For a content project, separate production automation from audience growth and commercial relationships.

Choose a test you can evaluate

The sponsor segment introduces AutoHarness, which turns corrections into reusable Skills. That is a workflow aid, not evidence that a side project will earn money.

1. Write the specific user or audience and the task your next experiment serves.
2. Choose a limited test with a visible outcome and a review date.
3. Keep actual results even when they are small or disappointing.
4. Compare the experiment with your original criteria before adding more production machinery.

The takeaway

This is a framework for collecting evidence from your own work. The recorded earnings and revenue split do not predict another creator’s results.