The Science of Scent

The hard part of formulating around cognition and emotion was never the chemistry — it's the literature. Evidence on aroma compounds and the brain is spread across chemosensory neuroscience, psychopharmacology, and sensory marketing journals, each using different doses, different methods, different populations. Reading all of it, consistently, for every candidate molecule, isn't something one formulator does reliably.

So we built an AI model to do it. Give it a molecule — or a full accord, broken down to its molecular composition — and it cross-references the compound against published research linking that structure, or close analogues, to measured effects on cognition and emotion. It surfaces what's actually been tested, and how strong the effect was, before we decide whether a material earns a place in a formula.

It doesn't invent effects or predict anything the literature hasn't shown. Linalool's anxiolytic effect, tied to GABA-A receptor activity; the correlation between absorbed 1,8-cineole and cognitive performance after rosemary aroma exposure; the mood shifts linked to citrus compounds, though even there the research is split on whether it's the molecule or just how much someone likes the smell — these are documented findings, not outputs the AI invents. What it does is make that evidence usable at the pace formulation actually happens, instead of leaving it buried across a hundred papers nobody has time to reread for every batch.

That changes where we start, not what we promise. An evidence-ranked shortlist isn't a guarantee any more than a single molecule was in our last post — the effect is still modest, still individual, and "documented in a study" isn't the same as "you will feel this." But it means the materials that make it into a bottle got there because something real pointed at them, not because they tested well on a mood board.

The AI reads the research. We still have to decide what to do with it.