Until October, fudini.co sold an app. It opened on breed ID from one photo, food matching and an AI vet, with the collar further down the page as something coming later. On 2 October we turned that around: the site now leads with the collar, and the app is what you use while the collar is being built. This is what changed, and what we got wrong on the way.
What we tried
The question was what Fudini is for. The app works today, but an app that identifies a breed and checks a food label is something you open once. The collar is the part that gives you a reason to come back: it sees the day, the app can't.
Most dog wearables are built around GPS, for the day a dog gets lost. We are building for the ordinary days in between. The collar learns one dog's own routine, how it sleeps, walks, plays and eats, and the app speaks up only when that routine drifts. No alerts is a good day.
We also had to be careful about what is ours. Fi already says its Series 3+ collar detects eating and drinking, so "detects eating" is not a story we can lead with. Comparing a dog to its own baseline, and staying quiet otherwise, is.
What actually happened
The first redesign was clean and minimal, and it was rejected as having no design at all. What worked was building a photo set first, then a dozen single-file prototypes with the same structure and different art directions, and choosing between them side by side. The winner was the cobalt one you are reading now.
The case changed too. The 3D-printed prototype from the post on designing our first prototype case in Python is retired. The production collar uses an off-the-shelf sealed wearable enclosure, 50 × 41 × 15 mm, in matte graphite with an orange sealing ring and the blue paw on top. The old "about 10 g" belonged to the printed case, so we stopped quoting a weight until we can put the real one on a scale.
The product shots on the site are generated renders, not photographs. When we measured them, the case came out 30 to 40 percent too large: about 2.7 times the width of the collar strap instead of 2. We re-rendered every on-dog image at the real ratio, and the collar page now has an actual-size drawing in millimetres, next to an AirTag and a €2 coin. We had also called the case "smaller than a matchbox". A matchbox is about 52 × 36 × 14 mm, so it is the size of one, not smaller.
Before the new site went live we went through the old one for claims we could not back. Out went an Apple Find My section (the collar has no GPS and will not find a lost dog), "the first longevity platform", "10 million data points" and "built on Google Cloud", which stopped being true when we moved to AWS in May.
The fix
The site is now Collar, App, Build log, Company and Investors, with one action: join the early list. The app page says plainly that you can use it today without the collar, and that pairing the collar adds automatic activity, sleep and meals, and drift alerts against your dog's own normal. Our supplement chews are off the main navigation for now. One product at a time.
We kept the app itself as it is. We designed a full dark restyle of all of its screens and decided against it: the home screen already had the right structure, and a new coat of paint would not change what the app is for. Its job changed, not its look.
What broke anyway
Three days after launch we found that the 200 breed guides had lost half their styling. The guide text still used the old site's markup, and the new stylesheet resets every link to look like plain text, so sources and links were invisible, the App Store button was bare text and the share buttons rendered as grey browser controls. Nothing was actually broken: all 225 internal links on the site's 219 pages returned 200. We restyled the old markup under the new design instead of rewriting 200 pages, and checked it on desktop and phone width before shipping.
What we have not fixed yet is that the guide photos all come from one template, and that the same dog appears on most pages of the site. That is next.