This is the raw record behind every published short on the channel: each generation call, the model that made it, the prompt, the references, the credits, how long it took, and what was kept or thrown away. It is the part of the work most projects leave out.
I wanted to publish this as a museum of my most viral episodes. I cannot, because the numbers do not say that. Here is what the channel actually is, measured, not estimated.
| YouTube subscribers | 347 |
| Videos published | 4 |
| Views, last 30 days | 5,871 |
| Best single video | 1,200 views |
| Second best | 438 |
| Third | 24 |
| Two others | under 10 |
A thousand views is a start, not a success, and calling it a hit would be the same mistake as putting a role on a CV that was never held. So this page leads with the study and the record instead, which is the part that is genuinely unusual.
Four records joined by a single task id: what was requested, what came back, what I decided, and what the quality gate measured. If any of those are missing for a call, the row says so rather than showing a zero.
Every thumbnail below is the actual file that call produced, not a stand in. Tap any one to see the prompt that made it, what it cost, how long it took, and what the quality gate measured. A receipt with nothing attached is a claim. This is the evidence.
Sourced from the study's own append-only run log. Nothing here was summarised by hand. See the tooling · Back to the portfolio