Build log

July 3, 2026

I read every complaint about every tab manager before writing a line of code

I had about fifty tabs open across five unrelated jobs and no idea what half of them were for. That's not an unusual state. What is unusual is that there are dozens of tools built to fix it, most of them free, and people keep abandoning them anyway.

So before building anything I went looking for why. Extension store reviews, forum threads, anywhere people talk about the tab manager they used to use. The same three failures come up over and over.

They lose your data. This is the top complaint by a wide margin and it's not close. Tabs vanishing after an update, sessions corrupted after a crash. One review put it about as well as it can be put: a tab manager has one job, and if it loses your tabs it failed at that job. Nobody churns because a tool lacked a feature. They churn because it ate their work.

They fill up and never empty. Users of these tools describe their workspaces becoming link graveyards. Saved tab groups pile up with no practical way to clear them in bulk — one writer spent days across forums trying to remove months of accumulated groups that kept syncing back. Every one of these tools is a filing cabinet that only grows, which means they're really competing with bookmarks, and losing.

They change the deal. More than one popular tool has cut its free tier retroactively — unlimited saving capped, workspace counts halved, with the tool choosing which ones you keep. The reviews are brutal: people who'd organised hundreds of tabs over years suddenly couldn't reach them, and those ratings never recovered.

There's also academic backing for the underlying problem. A university study on tab overload found people keep tabs open as reminders and out of fear of not finding things again — and that about a quarter of participants in one part of the study had their browser or computer crash from having too many open.

One more finding worth its own line: AI tab organisers group by topic, but people work by project. That mismatch is why the automatic groups don't survive contact with real work. The AI can see content; it can't see what you're trying to finish.

Those four findings became the rules I'm building against. Never lose a page. Empty by design. Never take away what was free. And let the person say what the task is, rather than guessing at it.