Problem
The problem is not taking notes. It is fragmented thinking. Useful information arrives as a link in the middle of research, an idea on a commute, a PDF before an exam, a screenshot, a decision lost in chat, a voice note made while your hands are busy.
Most tools respond by asking people to become librarians: choose a location, name the page, add tags, build a database, and remember how it was filed. So people either don't save the important thing at all, or they save it into a pile they can never use.
How I fixed it
I designed Meld around one lifecycle that every piece of information follows: Capture, Understand, Connect, Act. The user's only job is to capture.
- Capture is instant. Saving returns immediately; the intelligence runs in the background, so capturing never interrupts your flow.
- Organization is derived, never filed. One capture becomes one note. The AI gives it a home and extracts topics and people, so there is no folder tree to maintain.
- The user approves actions. Meld can propose a next step, but it never silently changes your tasks or plans.
- Every state is visible. A capture is processing, ready, or failed. Nothing is lost quietly.
Solution
Meld is a web-first personal intelligence system with a friendly, honest character. Its mascot, a little stack of saved cards, carries the brand through capture, connection, empty states, and errors, so an AI product feels approachable instead of mysterious.
Who it's for
Connect
Results
Meld is in private beta. The launch promise is deliberately small and concrete: save anything, ask for it later. Save a link, thought, note, or file in seconds; see that it is processing or ready; ask for it in normal language; get a grounded answer with the original source.
Designing and building Meld end to end is also how I work with clients on AI products now: I know where AI genuinely helps, and where the interface has to keep the human in control.
