Every Tuesday for the past six years, we've sent one essay to a quiet, surprisingly opinionated mailing list. The economics of doing so have changed roughly eleven times. The reasons we keep doing it have not.
What follows is less a thesis and more a slow tour of the moves we keep making, and the moves we keep watching other teams make, usually a week or two before regretting them.
Where we lost the thread
The standard mid-2010s marketing funnel assumed two things: that discovery had a venue, and that conversion had a button. Neither is reliably true anymore. The venue is now thirty surfaces, none of which you own, and the button is buried under a layer of intelligent triage you can no longer A/B test.
So the funnel didn't die. The funnel diffused. The reader is still there. The decision is still made. But the path is now invisible to anyone with a dashboard.
“The teams that won weren't writing for the model. They were writing for the reader the model serves.” — A note on AI search · Issue 47
A small refactor of the brief
When discovery diffuses, the brief has to compensate. We started writing briefs in two layers. The surface is what the reader sees and decides; the index is what the model needs to confidently retrieve us when asked. The two are not the same thing. Pretending they are produces work that does neither well.
It looks suspiciously like the briefs we wrote ten years ago, with the indexing column tacked on. Most things that work eventually look like the things we used to do.
Three measurements that survived
Here's the embarrassing answer: recall surveys, geo-holdouts, and the inbox. None of them are new. All of them are unfashionable. All of them work.
What the model can't see
The thing the model cannot see is the room you wrote the piece in. It cannot see the colleague who pushed back on draft three. It cannot see the deck that the brief came out of, the call where the brief was scrapped, the second brief that emerged from the wreckage. It can only see what you publish.
This is a feature. Publish things worth seeing.
Closing notes
If this hit any nerves, the kind ones especially, write back. We read everything. The reply-to address goes to a human; the human is usually me.
