What sits above the label on each piece

One decision, in the voice you picked, shown on real pieces from the library. Every number on this page was read from the database on 2026-08-25; the queries are on the file that builds it. Read it top to bottom and answer in chat with one line: CONFIRM, CHANGE: … or LEAVE OPEN: … (the box below spells it out).

The label on the piece is the only thing stored. Your twelve shelves are for browsing; your five worlds are saved views.

The idea, walked through

Right now every piece we save — a Nike brand film, an OpenAI tutorial, a press release — carries one label that says what kind of thing it is; we call that its type. That part is settled: you approved the 64 labels name by name. The open question is what sits above those labels.

Today the database groups its labels under ten family names, and that grouping was never yours — it came from an AI suggestion pasted forward from your BrandTrackers 5.0 planning document. It also misfiles things. The family called “Social-native & community” has no label you approved, yet it is the biggest one stored: 3,461 pieces sit under it — 2,518 social posts and 943 social clips — because “social” was written down as what a piece is, when it is really where the piece ran.

On 16 August you leaned toward the recommendation (“I want to take your recommendation, but a few call-outs” — the call-outs were your shelf names, three labels off the list and the name of the identity shelf, now “Brand & Identity”; all applied since). Here it is. Keep the one label on the piece. Treat the family names as nothing more than the shelves you browse the label list by: your twelve. And treat your five worlds — Brand Identity, Campaigns, Content, Digital, Physical — as saved views the product works out from facts it already has, such as which campaign a piece belongs to.

So the Nike commercial “Eventually everyone needs a professional. The Nike Book 2. Launching January 2nd.” stays a commercial. Browse the Campaign shelf and it is there; open the Content world and it is there too, because it belongs to no campaign yet; nothing extra was ever written on it. The table below does the same for nine real pieces.

This is what it is — nine real pieces

The pieces themselves, as they sit in the library today (the image is the piece's own thumbnail; text pieces show their opening lines):

OpenAI and Dell Technologies are collaborating to help more enterprises deploy Codex in the environments where their most important data, systems, and workflows already live. Codex is becoming one of OpenAI’s fastest-growing enterprise products…
text only — no image on this piece

At Boston Children's Hospital, AI is improving how care is delivered, from reducing administrative burden…
text only — no image on this piece

Video on the studio's site — open it.

GPT-5.5 is here! Have been daily driving this model for a few weeks and it's speed and capability…
text only — no image on this piece

Before and after, on the same nine

Read the two middle columns first: they are identical on purpose. That is the whole decision — nothing new is written on a piece; only where you find it changes. The grey code under each title is the piece's id in the database; long titles are shortened.

Which world would each appear in? Eight in Content, because they belong to no campaign — which is typical: only 312 of the 13,938 pieces do. The one exception is “DevDay 2025 — sizzle film (Studio Dumbar)”, which is linked to the campaign “OpenAI DevDay 2025” — so it appears in Campaigns as well, and that is exactly how a world is computed: from a link that already exists, never from a word written on the piece. Two similar names, two different things: the Campaign shelf is a group of labels (commercials, brand films, sizzle reels…), so a Nike commercial sits on it whether or not it belongs to any campaign; the Campaigns world is every piece linked to a campaign, whatever its label.

What else is written on these nine — and what this call does to it

A piece carries far more than its type. The database holds 44 tables that hang facts off a piece; these are the ones that matter for browsing. Here is what is written on the same nine today (a “none” means nothing has been written there yet — the DevDay sizzle film, for instance, has not been analysed at all):

This call changes none of it. Across the whole library, every one of these relationships stays exactly as it is — the decision writes one thing (which shelf a label sits under) into the label list, and nothing into any piece or any link:

Two of those rows are the argument for the decision. “The job it does” — awareness, proof, education, explanation — is already written on 9,032 pieces; the ten family names (Social proof, Education & support, Corporate communications…) mostly restate that same fact under different words, which is why storing a family on each piece would store one thing twice. “The campaign it belongs to” is the link the Campaigns world reads; because it is a link and not a word on the piece, the world updates itself when a piece joins or leaves a campaign.

What the probe found on the way — none of it is this decision's, all of it goes on the pipeline's fix list, and it is here so you see the real state of the rows behind this page:

This is what it changes

Where the pieces land. 3,324 pieces carry a label that maps straight onto one of your shelves. This is where they would sit on all twelve, worked out from the grouping you gave on 16 August, not typed (four shelves hold only new labels, so they are empty until the pipeline writes them):

The biggest single move is the 1,106 ad-library images: today they sit under “Advertising & promotion”, a family you dissolved on 16 August; under your grouping they land on Campaign. Every one of the 1,106 came from the LinkedIn and Meta ad libraries. The other visible move: 149 case studies go from “Proof & advocacy” to Users & Customers.

What this call does NOT move. 4,221 pieces carry one of 15 stored labels that are not on the list you approved. Each is handled in the leftovers section of the recommendation page, with its own proposed fate; this decision leaves all of them exactly where they are:

What the worlds would show today. A world is a saved view over facts already stored. The Campaigns world reads the campaign link — and only 312 of the 13,938 pieces have one (48 campaigns; the biggest: Customer Stories (30) · Codex for All Workflows (27) · Codex: Time to Fly (25) · OpenAI DevDay 2025 (19) · Just Do It: No Hesitation (17)). The Content world, your always-on world, is everything with no campaign: 13,626 pieces. That lopsidedness is real, and it is a pipeline gap — the step that works out which campaign a piece belongs to has not been built yet — not a flaw in the decision. A world stored on each piece would not fix it; it would freeze it.

The rules that switch on. The pipeline has a second sorting step that is switched off today: 12 rules, every one named after an old family — announcement · brand campaign · brand identity · company news · customer story · entertainment · event activation · feature campaign · partnership campaign · product campaign · social content · support education. The 16 rules that are on only look at where a piece came from. Once the shelves are yours, those 12 rules have real names to sort by; that work is engineering, later, not this call.

Browse “Campaign” tomorrow — the ten newest things you'd see

1. DevDay 2025 — sizzle film (Studio Dumbar) — sizzle reel, 2026-07-20, 6443fd99
2. Improving health intelligence in ChatGPT — brand film, 2026-07-06, ba3e5090
3. Plan with ChatGPT — commercial, 2026-07-06, 7ecc8d4a
4. Meet the ChatGPT Futures, Class of 2026 — brand film, 2026-07-06, 41f23142
5. Score with ChatGPT — commercial, 2026-07-06, 3f941a61
6. NikeSKIMS Studio Stretch | Nike — brand film, 2026-07-06, dd91f61a
7. Setup with ChatGPT — commercial, 2026-07-06, a83f6265
8. The Art of K-Pop Dance #GoogleDoodle — brand film, 2026-07-06, a224fdc3
9. “Deflated” is a companion piece to “air head”; @shy kids first experiment… — brand film, 2026-06-17, d3362525
10. Florp’s Solar Vacation was created as part of OpenAI’s Sora Showcase… — brand film, 2026-06-17, b0b59acb

If you confirm, and the alternative

If you confirm: nothing about any piece changes today. Later, as engineering work rather than by this call, the 64 labels get filed under your shelf names — some already exist in the database, the rest are new — and the switched-off sorting step gets its names.

The alternative is to store a world on every piece. It is easier to explain, but it doubles up a fact the campaign link already holds, and it drifts the moment a piece moves campaigns. The industry does it the first way: the type stored, the family as grouping, the rich views computed.

For scale, not for the decision: 13,938 pieces in the library (13,944 rows counting six deleted); 6,393 carry no label yet — that gap is a pipeline job, not this call. Your call, on one thing: is the label on the piece the only thing stored, with your shelves and your worlds as ways of looking at it? Answer in chat with one line — CONFIRM · CHANGE: what you would change · LEAVE OPEN: what you still need to see.