Stylist AI
A concept for a personal AI stylist — it learns the wardrobe you already own and dresses you from it. In design, not yet built.
Vision
LLM
Python
Concept
February 20, 2026
Overview
Status: concept. Stylist AI is a design I've sketched out but not yet built. What follows is the intended product, not a shipped one.
Every styling app I've seen eventually sells you clothes. Stylist AI starts from the opposite premise: you already own enough. The interesting problem isn't what to buy, it's what to wear from what's already in the cupboard.
The intended design
- Photograph your wardrobe once. A vision pass catalogues each item — type, colour, cut, formality, season — so the system reasons about your actual clothes rather than a generic catalogue.
- Dress for the occasion, not the aesthetic. You'd say "client meeting, 38°C, outside half the day" and get outfits that respect the weather, the dress code, and what's clean.
- It learns from what you reject. The signal isn't the outfit you're shown, it's the one you actually wear. Preference gets inferred from behaviour rather than a questionnaire you fill in once and never revisit.
- Gaps, not upsells. If it repeatedly can't complete a look, it tells you which single item unlocks the most outfits — the one honest place a purchase suggestion earns its keep.
Why it interests me
It's a good test case for agents operating on the physical world through a messy, user-supplied dataset. Nothing about a real wardrobe is clean, labelled, or complete — which makes it a more honest benchmark than a curated demo.