Vertical 03
A system that publishes, then measures whether it should have.
Six pipelines that research, write, render, and publish short-form video without supervision — and a measurement loop built to answer the question most content systems cannot: which of these actually worked?
Proves the substrate runs unattended and measures its own output.
The channel the system publishes to. Every post on it was produced by a pipeline and cleared by a human at the publish gate.
Where the channel stands
156 followers. 95,900 views in thirty days.
Roughly 615 views for every follower the account has. Reach on this platform is not a function of audience size, which is precisely why an unattended publishing system is worth building — the constraint is how many good attempts you can make, not how many people already follow you.
Platform dashboard · 12 August 2026
The shape of it
One post did 125× what the weakest one did.
The same pipeline, the same configuration, and the same approval gate produced both. Averaging these twelve numbers would report roughly 4,000 views per post and describe none of them.
- 0131.1K
- 021,574
- 032,832
- 04394
- 051,996
- 061,616
- 07248
- 081,604
- 092,209
- 101,933
- 112,867
- 121,685
Table view — all twelve values
| Post | Views |
|---|---|
| 01 | 31.1K |
| 02 | 1,574 |
| 03 | 2,832 |
| 04 | 394 |
| 05 | 1,996 |
| 06 | 1,616 |
| 07 | 248 |
| 08 | 1,604 |
| 09 | 2,209 |
| 10 | 1,933 |
| 11 | 2,867 |
| 12 | 1,685 |
The finding
The average was the least useful number in the dataset.
The median post reached 494 people. The best reached 12,981 — and every one of the top five came from the same visual lane. That spread is not noise to be averaged away; it is the entire signal, and a system measuring raw lifetime totals would have hidden it completely.
Measured 6–20 July 2026 · own instrumentation
Why the loop is built this way
Each measurement run appends a snapshot rather than overwriting the last one, so the same post measured on day 1, day 2, and day 7 becomes a growth curve. That makes posts comparable at the same age instead of on raw lifetime views — a six-day-old post is no longer competing with one published this morning.
Pipelines
Six of them, one shared spine.
Each produces a different format, and all six share one config, one knowledge system, and one control app. A new pipeline inherits the machinery instead of recreating it.
Facts
Short-form fact posts with SHA-256 dedup against everything published before.
Places
Real-footage travel reels with narration in English or Hindi.
Anime
All-generated nostalgia reels — the lane the measurement loop identified as the outlier.
Jokes
Character-performed jokes from sourced material.
Long-form
Narration-spine video built from a research stage.
Written
Final copy, ready to publish without a human rewrite pass.
The gate
Built with