01 · lens
a lens on any store page
The browser extension puts a score on every product card, opens the reasons on click and saves finds to your shortlist.
three surfaces · one taste model
the sale shelves of the stores you trust, filtered to what fits you and what you would actually wear.
Paste a link to a store and it joins the watch. A few times a day a collector reads the public catalogues of your stores, keeps items that exist in your size at your discount, and scores each one against your style profile. The score comes with its reasons, and every link opens the product with your size already selected.
02 · catalog
a quiet catalog of what passed
Take now, wardrobe gaps, everything in your size, signals and a shortlist. A running log tells what changed on the shelves.
03 · ledger
a ledger that holds the rules
The database behind both: every record, stores added by link and checked automatically, thresholds, sizes, the taste profile the lens pulls, the shortlist, every change logged. Private, sign-in only.
what it looks like · synthetic data
dense tables, short lines, numbers on the right: a report you can read in ten seconds. the same three surfaces, filled with made-up items and stores.
live model · synthetic items
six made-up items run through the same rules the lens uses. pick a profile, open a card, read why it scored what it scored.
how taste is formed
a demanding baseline, then evidence.
- baseline 24nothing is good by default.
- brand tier +3…+28compressed so a name alone never reaches take.
- colour −30…+22the strongest separator: the swatch beats the title.
- cut and function +7 · +10relaxed volume, membranes, ripstop, pockets with a job.
- return memory −30 and morewhat you sent back after a try-on outweighs any theory.
- size gatestyle score and purchase score stay separate: a great item in the wrong size stays visible, never as take.
- take ≥ 82 · consider 55–81shoes are their own class: take ≥ 78, consider 52–77. below that the item leaves the radar.
taste graph
your profile, drawn as what the lens adds and subtracts.
Eight branches: colour, silhouette, function, season on the left; brands, footwear, return memory and what you avoid on the right. Node size is the weight, red is a penalty, the number is how many items on the shelves carry it right now. In the ledger a click on a leaf opens those records. Below: a made-up profile.
wardrobe setup protocol
twenty minutes that turn a wardrobe into a profile.
- sizestop, bottom and shoe size in the systems stores use (EU, US, UK, waist), plus the neighbours that sometimes fit and the ones that never do.
- ten anchorslinks or photos of the ten things you wear most. colour, cut, fabric and brand are read from them.
- three returnswhat you sent back and why: size, colour, softness, a logo too loud. each becomes a penalty with a date.
- palette in percentfor example black 50, charcoal 25, bone 10, tan 5, one accent. the swatch outweighs the product title.
- four brand ringscore, close by, everyday base, accent in a dose, and a short list of what you skip.
- climatewhere you live and which layers wait for their season.
- stores by linkthree to ten stores you trust. each link is probed: readable catalogue, currency, region, sale collections, how many items fit you today.
The answers become a profile file, a taste graph and thresholds. The lens works from the profile alone, right in the browser, before any server is involved.
architecture
one collector, one ledger, two views, one lens.
collectora few times a day · public catalogue JSON of the stores in your ledger · a new link is probed first · polite pace, one request per 2.5 s
→
ledgerour server · deals, signals, stores, thresholds, taste profile, shortlist, change log · sign-in
→
catalog · lensthe catalog reads the ledger; the lens pulls the taste profile and writes to the same shortlist
Signals: new, cheaper or back in your size, inside the take and consider lanes. A run that finds any sends one quiet notification.
private beta
it runs for one wardrobe today, the next ones by invite.
The ledger is a private database on our server; you sign in with an invited identity. The closed beta starts with the setup protocol: we build your profile and graph together, you get the lens with it, then your stores join a ledger of your own. Write to ask for a place.
ask for a beta place ↗ledger · sign-in by invite
it reads, links back and never buys.
style os reads catalogues that stores publish, keeps a narrow slice in your size and sends you back to the store's own page. It never touches carts or checkout and keeps no tracking. Product photos on this page are drawings, not store images. Store and brand names identify sources; there is no partnership unless stated.