Find the drift nobody decided on

Diff this year's output against last year's and see what quietly changed in your look while nobody was choosing it.

Brands rarely decide to change. They drift, one small compromise at a time, and notice eighteen months later when a rebrand is the only remaining option. A library with dates in it catches it while it is still a choice.

Why this will not look generic: It compares two real slices of your own output, so what it surfaces is change that actually happened rather than a trend someone read about in a roundup.

What you walk away with

A prompt for every asset in each period
What changed between them, with counts rather than impressions
A call on which changes to keep and which to reverse

Typical run: 30 to 100 in each slice. Run it quarterly once the corpus exists; the whole point is catching drift early rather than auditing it later.

This one happens in one conversation

You bring: Two folders of your own work, tagged by period. A year apart shows more than a quarter.

Open with: Compare the prompts tagged 2025 against the ones tagged 2026. What changed, and is any of it something we would have decided on purpose?

What the thread holds on to:

Both periods at once, which is the only way a gradual change is visible at all
Which shifts you have already said were deliberate, so it stops flagging them
The emerging account of what happened, which you can test against specific assets

Then you can just ask:

Which of these drifted because of a tool default rather than a decision?
Is the newer look actually better, or just newer?
Write the directive that locks in the half we want to keep

What the reverse actually gives you

Point it at: Two years of your own shipped work, tagged by period and read side by side.

What comes back (representative example): Three things moved and only one of them looks deliberate. Contrast has risen steadily: last year's set sits in a mid-tonal range with soft shadow rolloff, this year's clips shadows in roughly half the assets. Crops have tightened, with the subject occupying noticeably more of the frame and negative space largely gone, which was the single most distinctive thing about the earlier work. Palette is unchanged and still disciplined. The tightening tracks the shift toward square and vertical placements, so it has a cause, but nothing in the record suggests anyone chose it.

Things you would never have thought to type:

shadows clipping in roughly half of this year's assets
negative space, previously the distinctive thing, largely gone
the tightening tracks a placement change, so it has a cause
palette unchanged, which rules out a general loosening

What you change: Whichever half of the drift you decide you want. The output is a decision you are now in a position to make, rather than one that has already been made for you.

How it works

Reverse both periods (In your code) - Everything you shipped in each window, not a curated selection. Curating defeats the point.
File each period under its own tag - The tags are the time axis. Without them this is one undifferentiated pile.
Ask what moved (In Claude) - Press for counts. 'Feels warmer' is the thing you were already able to say without this.

Cost and time

This workflow runs at 2 reverses + about 1 minute. Your result carries automatically from each step into the next, so there is no re-uploading between tools. Reverses are billed from your reverse meter and generation steps from your AI Credits, both bundled in every pay-as-you-go credit pack.

Browse all guided workflows or all individual tools. Pay as you go, no subscription, AI Credits never expire.