Too many tabs open? Sort them before closing any

The favicons have shrunk to slivers, the titles are gone, and finding the right tab now means hovering along the strip like a slow scan. The obvious advice is to close some. The reason that advice does not get taken is that every tab up there is holding something, and there is no way to tell from a two pixel icon which ones are holding something that still matters.

Closing at random produces a worse outcome than doing nothing, because the tabs that survive a random cull are the ones on the left, which are the oldest and least relevant. Sorting first costs about ten minutes and makes the closing step trivial, mostly because it turns out that only one of the four kinds of tab in the strip actually belongs there.

Every open tab is a decision that was postponed

A tab stays open for one of two reasons. Either something is unfinished in it, or closing it would require deciding whether it will be needed again. The second reason accounts for most of the strip.

This is why the count grows steadily rather than in bursts. Opening a tab costs nothing and takes no thought. Closing one requires evaluating whether the page can be found again, whether anything in it is unsaved, and whether the thing it was opened for is done. Three small judgments, multiplied by ninety tabs, is a genuine amount of work, and the browser offers no help with any of them.

The strip is also doing three different jobs at once without distinguishing between them. It is an application launcher for the handful of web apps that stay open all day. It is a filing cabinet for pages being referenced. It is a to-do list for things that were opened with the intention of reading them later. A single horizontal row of unlabeled icons is a poor interface for any one of those jobs and an actively bad one for all three simultaneously.

The fix is not fewer tabs. It is separating the jobs so each one gets an interface that suits it.

Four kinds of tab, sharing one row

Almost every tab in a crowded strip falls into one of these. Naming them is the whole trick, because each one needs a different action and applying the wrong action is what makes cleanup feel futile.

Applications

Mail, chat, calendar, the project tracker, the notes tool. These are not pages. They are software that happens to be delivered through a browser, they are tied to a specific signed-in account, and they will be open again tomorrow morning regardless of what happens today. Typically there are between four and eight of them.

Live references

Documents, tickets, and specs open because the current task needs them visible. They have a lifespan measured in hours or days. They are the only category where a large number is legitimate, and they are also the category that vanishes on its own once the task ends, provided anyone remembers to close them.

The queue

Articles, threads, and documentation opened with a real intention to read them, and not read. These accumulate faster than they clear, and they are the ones that generate guilt, which is precisely why they never get closed. The queue is a reading list wearing a tab strip costume.

Residue

Search result pages, intermediate steps, a page opened to check one fact, a link that opened in a new tab because the site was built that way. Residue has no owner and no future. In a strip of ninety tabs, this is frequently thirty of them.

The pass, in order

Do not start at the left edge. Open the tab list view instead, which presents the same tabs as a readable vertical list with full titles, and work from that. Chrome, Safari, and Edge all have one, and reading titles instead of icons is what makes the sorting possible at all.

Then go through in this sequence, because each step shrinks the work of the next.

1. Close all residue first. It is identifiable in under a second per tab and it is usually a third of the total. Nothing is lost, because residue by definition contains nothing that was not already used.
2. Move the queue out. Every browser has a reading list or a bookmarks folder. Put all of it there, without triaging what deserves to be read, because triaging is what stalls the pass. The queue's real problem is that it lives in a place that costs attention all day, and moving it solves that problem whether or not it ever gets read.
3. Group the live references by task. What remains should be small enough to see, and the grouping is what lets the whole set be closed at once when the task finishes.
4. Deal with the applications last, and separately, because they need a structural answer rather than a tidying one.

Two things reliably stall this pass. One is opening a tab to check whether it still matters, which turns a ten minute sort into a two hour reading session. The rule that prevents it is that the decision gets made from the title alone, and anything the title cannot settle goes to the reading list rather than staying open. The other is the fear that something unsaved is buried in there. That fear is usually justified for exactly one or two tabs, and the way to handle it is to scan for form pages and editors first, deal with those, and then stop worrying about the remaining eighty.

The pass ends with a strip that holds one task's worth of references plus the applications. That is the state worth returning to, and the entire question of maintenance comes down to whether the applications keep their own place.

Where each kind actually belongs

Kind	Lifespan	Right home	What goes wrong in a tab
Applications	Permanent	A fixed, dedicated window or container	Buried among references, reloaded, signed out
Live references	Hours to days	A named tab group in the working window	Survives the task and becomes residue
The queue	Indefinite	Reading list or a bookmarks folder	Occupies attention every hour of every day
Residue	Minutes	Closed	Multiplies silently, hides everything else

Reading the table by lifespan rather than by kind makes the underlying rule visible. Anything measured in minutes should be closed, anything measured in hours should be grouped, and anything measured in years should not be in a tab strip at all. The middle row is the only one the tab strip was designed for.

The row that causes the most trouble is the first one. An application tab and a residue tab are visually identical in the strip, and they sit next to each other, which means the six things needed all day are competing for space with thirty things needed for four seconds each.

Why the residue comes back within a week

Sorting once and watching the count climb again is the normal outcome, and it is not a failure of discipline. Three mechanisms refill the strip, and only one of them is under the reader's control.

The first is that most links in most web applications open in a new tab by default. Clicking through a chat thread with six shared links produces six tabs, none of which were consciously opened.

The second is that search is faster than retrieval. When a page is needed again, opening a new tab and searching takes four seconds, while finding the existing tab takes longer than that in a crowded strip. So duplicates accumulate, and a strip of ninety tabs commonly contains the same documentation page three times.

The third is notifications. A notification from a web app pulls attention into that app, and if the app is not already open in a known place, it opens somewhere new. Applications that live in a fixed location do not do this. Applications scattered through a tab strip do it constantly.

Only the third mechanism has a structural fix, and it is the one that matters most, because applications are the tabs that never close.

What the browser can and cannot do about it

It is worth being precise about what the built-in tools solve, because they are frequently recommended for problems they do not address.

Memory management features address memory. Chrome's Memory Saver lets inactive tabs release memory, and it maintains a published list of situations where it will leave a tab alone, including tabs playing audio or video, tabs with an active download, tabs with a partially filled form, and pinned tabs. That list is well designed for its purpose and irrelevant to the count, since a discarded tab still occupies a slot in the strip. A machine can run a hundred tabs smoothly and still be unusable for the person trying to find one of them.

Tab groups address retrieval, which is real progress. A collapsed group turns twelve tabs into one labeled item and makes the strip readable again. What groups do not do is impose a lifespan. A group named after a topic rather than a task is never finished and never gets closed, and eight permanent groups reproduce the original problem one level up.

Reading lists address the queue, and they work, on one condition: that the list is opened on a schedule. A reading list nobody opens is the same backlog stored somewhere quieter, which is still an improvement, because a backlog that is not visible is not costing attention.

The gap none of them close is the application problem. A tab strip has no concept of an item that is permanent, and so the six or eight things that will be open for the next two years are stored in the same structure as a search result from four minutes ago.

Giving the permanent things a permanent place

The alternative arrangement is to stop treating applications as tabs at all. Give each web app its own window with its own session, so that mail is always in one place, chat is always in another, and neither of them ever appears in the strip where documents are being read. The strip then holds only references, which are supposed to be temporary, and a temporary strip empties itself.

That arrangement is what an app aggregation browser is built for. Each app gets a fixed container rather than a position in a queue, and switching between them is a move to a known location instead of a search. The Features page describes how the containers are separated, and the Supported apps page lists which services fit the model directly. For anyone who has been running two accounts of the same service in different windows, the Workspaces page covers how those get arranged without a second browser.

What to change first

Run the four way sort once, with the residue closed first, and count how many tabs remain. If most of the survivors turn out to be applications rather than references, the problem was never the tab count and no amount of closing will fix it. Moving those applications out of the strip is the change that holds, and SpaceDeck is built around giving each of them its own window.