THE ALGORITHM WOULD LIKE TO SEE YOUR CAMERA ROLL; THIS BUREAU WOULD LIKE A WORD

PORTUGAL — Among the announcements at Tinder's spring keynote was Chemistry: an AI that learns what makes you you through questions and — with permission — access to your camera roll, then curates matches accordingly. The stated aim is reducing swipe fatigue. I write from Europe, where my professional reflexes were formed by the world's strictest privacy regime, and I would like to examine the word doing all the work in that sentence: permission.

Permission is real, and I do not doubt the consent screen will be lawful. But an engineer evaluates architecture, not paperwork, and the architecture here is: the system must inspect your private life to infer what you want. Your photo library — your family, your medical screenshots, your receipts, your unflattering Tuesday — becomes training signal, because the platform's design never included a simpler instrument: asking you. Inference is what a system builds when declaration was left out of the schema. It is a very expensive apology for a missing form field.

And inference fails in a specific, intimate way: it extrapolates your future from your past. The camera roll knows who you were. It cannot know that you just decided you're done with casual, or ready for friendship in a new city, or finally willing to say "long-term" out loud. People change on purpose. Archives don't.

MOOD's data architecture follows from one principle: intent is declared, not derived. You tell us the MOOD; that single, revocable, user-authored fact does the work a thousand inferred signals approximate badly. We built for jurisdictions from Louisiana's new age-verification law to the European rulebook I grew up under, and the cheapest compliance strategy ever devised is also the most respectful one: collect the sentence, not the life.

Your camera roll is yours. What you want is a thing you say, not a thing we mine.