What Theia learned this week: views settle in 6 hours, and a selfie beat the sitcom

Every week we publish what Theia, our AI content studio, learned from her own films. She runs our TikTok account on auto-post: she picks the idea, makes the film, posts it, reads the analytics and changes her own code. Each film records the exact version of the code that made it, so every number below points back to one version.

This is the first one, so it covers the last 11 films, posted from 11 to 21 October. The numbers are real and unedited, including the bad ones.

The films

"Held at 2 s" is the share of viewers still watching two seconds in. "Code" is the version of Theia's code that made the film.

Film	Posted	Format	Views	Likes	Held at 2 s	Avg watch	Code
Theo, episode 5	11 Oct	Sitcom	715	10	29%	2.65 s	4822686
Theo, episode 6	12 Oct	Sitcom	702	0	31%	2.69 s	4a01290
Theo, episode 7	13 Oct	Sitcom	237	3	39%	5.12 s	55d4586
Mum's voicemail	14 Oct	Sitcom	226	6	35%	3.59 s	3765789
Table four	15 Oct	Sitcom	222	3	35%	3.52 s	6406d98
Navy mug	16 Oct	Sitcom	228	2	32%	3.88 s	8fdccab
The plug	17 Oct	Sitcom	239	5	30%	3.31 s	a1e9fe0
Both sides	18 Oct	Sitcom	308	4	51%	6.40 s	2a83275
The radio	19 Oct	Text on a moment	245	8	42%	3.50 s	d621dd4
The toothbrush	20 Oct	Selfie confession	301	24	51%	5.38 s	62c11f5
The glass	21 Oct	One take, filmed by a friend	169 (first 2 hours)	5	not read yet	not read yet	a2ad927

Comments and shares were zero on every film. Two films got one save each.

What she changed this week

She stopped climbing and started exploring. Weeks of one-change steps on the sitcom format kept landing at about 230 views. Small steps find the top of the nearest hill, not the highest one. So on 19 October Theia started an exploration week: seven very different formats, chosen by her, each with a written reason. Two examples of her reasons: the account's best post ever was a text-on-a-moment film, and a selfie film in September reached about 1,900 views.

She learned where TikTok's buttons are. The radio, the first exploration film, ran two words of text under TikTok's button column, where nobody can read them. Theia added safe zones to her title cards and a test for them, so the next films kept text clear.

She raised her own finish. The exploration films looked rougher than the sitcoms, because they skipped the upscaling step the sitcoms use. She added it, and posting now refuses any film without it.

What the data supports

1. Views settle in the first 6 hours. Across all 11 films, 88 to 100% of the views came in the first 6 hours (the median is about 98%). So Theia now reads views at 6 hours instead of waiting two days. Retention and likes still get the longer read. This is 11 films on one account, so it is a working rule, not a law.

2. Views did not separate the formats. Likes did. Every film since 14 October got between 169 and 308 views, sitcom or not. But the selfie confession, the toothbrush, got 24 likes on 301 views: 8.0%. The next best was the radio at 3.3%, and the sitcoms ranged from 0% to 2.7%. That is one film, so it is a lead to test again, not a verdict.

3. Holding the first 2 seconds and watch time move together. The two films that held 51% of viewers at 2 seconds, both sides and the toothbrush, also had the longest average watch: 6.4 and 5.4 seconds. Everything else held 29 to 42%.

4. The two biggest films had the weakest openings. Theo episodes 5 and 6 reached about 700 views each, but held only 29 to 31% at 2 seconds, and episode 6 got no likes at all. We do not know why they got the extra reach, and Theia's notes do not explain it. We are not going to invent a reason.

What we cannot claim yet

• Nothing about sharing. Zero shares on 11 films means there is nothing to learn from shares yet. That is the clearest problem.
• The glass is too new. Its first full read is due on 23 October.
• The 2-second and watch-time numbers were read on different days, between 14 and 21 October, so they are not all the same age.

Next week

Theia makes the other four exploration formats she chose: a phone screen, a street vox pop, a photo carousel and a voice note. After film 7 has had 48 hours, she writes a verdict table and climbs the one or two winners. Next week's post will show whether the selfie lead held up.

If you want the same loop on your brand's account, start with a pilot month. For how this kind of loop works, and how it can go wrong, read Recursive self-improvement in practice: lessons from the STOP paper.