Product design
Kept: designing and building an AI-powered for saving links
Overview
Kept started from a personal annoyance: bookmarks scattered across browsers, folders, and devices, with no reliable way to remember where anything actually lived. The idea was simple, one place, one action, save from anywhere. I designed and built it myself, end to end, product, interface, and the code running it.
Approach
One click and that's it was the actual strategy, not just a UI shortcut. Every extra step, choosing a folder, writing a title, adding tags, competes with the one reason someone saves something in the first place: to stop thinking about it. So the person's part stops at paste. Everything after that, pulling the title, the image, generating tags, is the system's job. AI sorts it out so the user never has to.
Search is built to feel like a conversation, not a query. Nobody remembers "I filed that under Clothes," they remember "that blue sweater on sale." Tags and filters still exist underneath, but describing the thing in plain language is the primary path in.
And when the first result isn't quite it, the fallback isn't a broader search, it's a related one. Tapping a kept link surfaces a "Find Similar" option, pulling up other saved links sharing at least two tags with it. Instead of retyping a better query, you drift sideways from something close until you land on the one you actually meant.
Outcome
Kept is currently in a testing environment, not a public release. The goal right now is straightforward: surface and resolve usability issues before it's ready for anyone outside a small test group. One-tap capture, AI-generated titles and tags, conversational search, and Find Similar are all built and functioning, this isn't a prototype, it's a working product being stress-tested against real use. Publishing publicly by end of year is the goal, not a promise, that timeline depends on what testing turns up between now and then.
