Fit Receipt
Product Hunt Crystal Chang
Fit Receipt launched a private lingerie try-on agent with an AI “receipt” for each pick. The twist: photos stay in the browser, and the model can refuse to guess when it’s unsure.
Based on reporting by Product Hunt, Crystal Chang — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Fit Receipt is launching today as a private fitting agent for virtual lingerie try-ons. The pitch is simple enough: chat with the agent, try things on, and get an AI judgment receipt for every pick. Behind that clean surface, though, the project is set up more like a safety demo than a shopping bot.
It’s a reference implementation, not just a product. The agent handles the conversation, while code protects the facts and pricing so they don’t get rewritten by a chatty model. That split matters. In this setup, the language layer can talk; the guardrails keep it honest.
The judgment side comes from JEV, a TypeSafe model that scores trade-offs with calibrated confidence in about 300ms rather than spinning out essays. If confidence is too low, the result stays unresolved. Those judgments are recorded, and the system can still work without JEV at all.
Privacy is the other big piece here. Photos stay in the browser, with no server-side photo storage. JEV only sees typed fields, never the photo itself. The API is rate-limited, and the whole thing is built with Vercel and open source. It’s a neat reminder that not every AI product has to be a giant black box pretending to know everything.
My take — AI-written commentary, not fact-checked reporting
This is the right instinct for AI in anything sensitive: keep the picture local, keep the numbers fenced off, and make the model admit when it’s guessing. Too many products still confuse confidence with competence, which is how you end up with polished nonsense wearing a startup logo. More of this, less “trust the vibes.”
Read more about this at: Product Hunt
