iConnect

It records the conversation next to the contact: who you met, what it was about, when you last spoke. Five kinds of interaction – meeting, call, email, event, coffee – each with a note and a date, dictated by voice if typing is not an option right after a conversation.

On top of that the things you need when a name comes back weeks later: search across notes as well as details, follow-up reminders at a fixed time of day, colour-coded tags, a photo.

The standard advice is to sharpen your skills on a problem you have yourself. Mine was this one. At catch-ups that only come round every few months I kept losing the thread of what we had actually discussed – possibly a comment on my memory as much as on the tooling.

So the idea was to call myself after a conversation: say the two or three things that mattered out loud, and have them land on the right contact. Before the next meeting, read the summary and carry on where the last conversation stopped. That is the whole app, and it is why there is no server behind it. Notes about the people you meet are close to the most personal thing a working life produces – they stay on the phone they were spoken into. No account, no cloud, nothing tracked. It costs sync between devices and a web version, and that is the trade.

Not in the product. Dictation runs on the phone's own speech recognition; there is no model, no service, no server, and therefore nothing that sends a note about a person anywhere. I put it plainly in my own post about the launch: this is not a fancy AI tool.

In the build, everywhere. iConnect was written with Claude Code and Codex, and that is the only reason an app like this gets finished alongside three other projects and a degree. It is the same skill the other work runs on – here it stays a tool and never becomes a promise on the box.