Case studies: real numbers from our own accounts, including the weak ones

The case studies

• Dumpling Does Nothing: a five-episode AI pet comedy, and what it really cost

  A capybara who tries to do nothing. Seven paid renders for five episodes, and almost no audience.

  Outcome: Production worked and cost was predictable. Distribution did not: single and double digit views on most platforms.

  Run 2026-10-23, numbers read up to 2026-10-28

• The Last Subscriber: an AI founder-drama series, from 112 views to almost none

  A first-person drama with a recurring heroine. One short test burst of views, then a flat line.

  Outcome: Episode 1 got a single test burst of views. Later episodes got single or double digits. It was not a proven format.

  Run 2026-10-23, numbers read up to 2026-10-28

• Pip Has One Job: three stickman episodes for 40 credits each

  The cheapest run we did: still-image stickman episodes. Early reads only, and mostly small.

  Outcome: Cheap and fast to make. Reads are only hours old and small, except one Facebook reel. No verdict yet.

  Run 2026-10-27, numbers read up to 2026-10-28

• The cat-fact Short that got 111 views: what that number does and does not mean

  A plain explainer video. 111 views is our best YouTube result for that channel, and it is still a test burst.

  Outcome: Best of its channel, still small: roughly one test batch of views, most viewers left in seconds.

  Run 2026-10-25, numbers read up to 2026-10-28

• What 24,000 Shorts taught us about virality: no formula we could trust (our study)

  We measured 24,273 Shorts. The honest result: no title or hook rule was robust, and structure does not predict outliers.

  Outcome: No structural rule survived our own checks. The strongest hint is small, and it is a hint.

  Run 2026-10-28, numbers read up to 2026-10-28

In progress: the 12-video formula experiment

We are testing a short checklist from the study below on our own accounts, as paired videos: one made with the checklist, one in our current default style, posted on the same day. Results come after the 72-hour reads. No numbers are published yet, and nothing from it will be written up as a promise of views.

How to read these pages

• Each number has an "as of" date. YouTube, Facebook and Instagram numbers change for days after posting.
• Our own views and comment visits are not subtracted unless a page says so. Treat small numbers as signal, not proof.
• Credits are what the product charged. USD figures from our provider costs are measurements, not invoices.
• Prices and what a prompt costs: what things cost.