There is a species of work that never appears on a to-do list. Renaming a file so the recipient knows what it is. Converting a screenshot into whichever format a portal decided to accept this month. Shrinking a deck that bounced off an attachment limit. Re-exporting the same image three times because every tool wanted something slightly different. None of it is difficult. All of it is real work, and none of it gets counted anywhere.
We have started calling this the attention tax. It earned the name after a week of keeping a notebook open and jotting down every file chore touched in a working day. Not projects — handlings. Open, convert, resize, compress, re-upload, re-send, rename. By Friday the list filled two pages, and it had not even been a chaotic week.
One entry from that notebook still stands out. A screenshot went into a compression tool expecting the usual silent shrink. Instead it paused and said: converting this would make the file bigger — keeping the original format is the better call. A tool admitting it had nothing to add. That was the moment the tax started looking refundable, because a chore that can say "no, this is already fine" is a chore you can hand over entirely.
The arithmetic nobody runs
The research suggests the notebook week was not a special case. McKinsey's analysis of knowledge work found that roughly 19 percent of the average work week goes to searching for and collecting information — about 1.8 hours every day spent on getting hold of things rather than doing something with them. That figure usually gets quoted in debates about search, but search is the smaller half of the problem. The larger half is everything that happens after you find the thing: the converting, the shrinking, the re-attaching, the polite note to a colleague explaining why the file arrived in two pieces.
Being pulled away has its own price. Interruption research keeps circling the same uncomfortable number: once you are knocked out of a task, it takes about 23 minutes to get fully back into it. Pair that with the finding that people now hop between roughly 35 apps in a working day, and a picture emerges. Every file chore is a small interruption wearing the costume of a two-minute errand. Each one steps on the brake of whatever you were actually thinking about, and the restart afterwards is never free.
Try a rough count. Ten small file chores a day — the notebook week says that is conservative — each costing five minutes of handling plus a refocus tail. Even at a stingy estimate, that is over an hour of daily work that no role description mentions, no meeting covers, and no review rewards. It simply dissolves into the day, and everyone quietly absorbs it as the cost of having files at all.
What makes the tax worse than its size is its texture. A hard problem absorbs you; a micro-friction scatters you. Forty minutes on a difficult draft leaves you tired but intact. Ten two-minute chores leave you tired and fragmented, because each one carried a switch cost that the five minutes never paid for. The chore is not the bill. The chore is the tip, and context is what the restaurant actually charges.
That reframing changes how fixes should be evaluated. Stop asking "how fast is this tool" and start asking "how many switches does this tool remove." Speed matters less than it seems when each use still demands that you stop, open a thing, remember its menus, wait for a progress bar, and hunt down the output. A faster interruption is still an interruption.
What gets tried first, and why it falls short
The first instinct is usually the classic one: automate. Write the little scripts. Wire up the folder watchers. Build a rename convention so lovingly detailed that only its author can operate it. For about three weeks it is magnificent, and then a client sends files in a shape nobody anticipated, the watcher chokes, and the lesson lands: maintaining automation for chores is itself a chore — one that bills at senior rates and arrives without warning.
Batching helps more, honestly. Reserving Friday afternoons for the week's accumulated file drudgery collects the switch costs into one block, which is an improvement. But batching has a ceiling: half of these chores are not deferrable. The deck has to go out today. The portal closes at six. The portal always closes at six.
So the tax stays, roughly unchanged, for years. Which is why what comes next feels less like an upgrade and more like a change of kind.
When the chore becomes a sentence
The alternative that actually held up was handling files through an AI assistant the way one would ask an assistant in an office: in plain language, one request at a time. Shrink this deck so it clears the mail limit. Make that folder of scans sendable. The machinery behind it — SmartSlim, with a compress-mcp layer that lets the assistant drive it, described on the products page — takes the request and does the handling on the same machine, without anyone opening anything.
The difference is difficult to overstate precisely because it is boring. Nothing dramatic happens. The file gets smaller, and you stay where you were, thinking about whatever you were thinking about. The switch never occurs. A chore that used to cost a context break now costs a sentence, and sentences are cheap.
There is a second-order effect that nobody expects. When handling gets cheap, deferring stops. The deck gets shrunk before it gets sent, not after the bounce. Files arrive where they should arrive, at the size they should arrive, and the recipient's afternoon is spared too. Frictions are contagious in both directions — reducing yours reduces someone else's.
The day a tool said no
Here is the moment that earned permanent trust, and it was a refusal.
A batch of images went in with the usual request: convert these to another format. The tool looked at the first one and declined. Converting it, the answer explained, would make the file bigger, not smaller — this particular image would inflate in the target format, so the tool was not going to pretend to help. Would the original be kept as-is, or compressed in its current format instead?
It is worth being clear about why this mattered. Tools mislead constantly, in small, deniable ways. They report success because a process completed, not because an outcome helped. Everyone has walked away from a cheerful "done!" that left the file larger, the folder messier, and the problem untouched. A tool that stops and says "this specific operation would work against you" is behaving like a colleague rather than a vending machine.
It also changes how much the tool's yes is worth. When a no is possible, a yes means something. Had it converted the image anyway and reported success, the only lesson would have been to keep checking its work forever. Because it refused when refusal was right, a queue of two hundred files can be handed over without a backward glance. Trust compounds in the same direction as attention: spent once, saved forever after.
The quiet economics
The other half of the comfort is harder to photograph. The compression happens on your own machine. Client files, contracts, the folder of things nobody wants to describe — none of it gets uploaded anywhere to be shrunk. For people whose work involves other people's confidential material, that fact does more than any feature list. The tax gets paid, and the files never leave the room.
There is a money angle too, and it sneaks up on you. Feeding images into AI models is billed by tokens, and published figures put the drop in image token cost at about 85.9 percent after compression, measured on a GPT-4o basis. Small savings, repeated across every image in every document, over every workday, become the kind of number a finance person circles. The attention tax and the token bill turn out to be cousins: both punish careless handling, both reward making things smaller before they travel.
The budget worth guarding
Attention behaves like a budget with cruel fine print: interruptions do not have to be big to be expensive, they have to be frequent. The 23-minute refocus does not ask how important the interruption was. The 1.8 hours of daily searching and collecting does not care that each individual fetch was justified. The tax is levied on switches, and switches are what file chores are made of.
So the rule of thumb has become: anything that costs a switch should be sayable in a sentence. Not because conversation is a virtue in itself, but because a sentence is the smallest unit of delegation that still carries intent. Below a sentence, you are operating a tool. At a sentence, you are directing one.
The daily ledger changes, and the difference shows up somewhere unexpected: fewer reopenings. The measure of a good tool is not what it does while you watch it. It is how much of your mind is still in the room afterwards.
FAQ
Q: How is this different from just buying a faster compression app?
A: Speed was never the expensive part. A faster app still makes you stop what you were doing, open it, find the right menu, and fetch the output — the switch cost stays. Delegating the chore as a sentence removes the switch itself, which is where the attention tax actually lives.
Q: Isn't scripting and automation the proper long-term fix for file chores?
A: It works until reality deviates. Scripts and watchers only cover the shapes you predicted, and maintaining them becomes its own chore. Language-based handling absorbs the unexpected case without anyone maintaining anything, because the request adapts even when the files do not.
Q: How do I know the work was done correctly if I never opened a tool?
A: The results are reported in plain terms — what shrank, by how much, and what was deliberately left alone. Because the tool can refuse when an operation would backfire, its confirmations are trustworthy rather than automatic. You audit outcomes, not processes.
Q: What does the 85.9 percent token figure mean in practice?
A: After compression, images consume roughly 85.9 percent fewer tokens when processed by AI models, on a GPT-4o basis. If your documents regularly travel into AI workflows, smaller images translate directly into a lighter bill — the same principle as the attention tax, applied to money instead of focus.
Q: Where can I see more about this way of thinking about file work?
A: The blog at https://www.uglypear.com/en/blog/ collects essays on file handling, compression, and the cost of small frictions, alongside the product background on the main site.
Try it now
If the arithmetic above feels familiar, the refund is one configuration away. Start at https://www.uglypear.com/en/ to see the overall approach, browse the tools at https://www.uglypear.com/en/products/ — SmartSlim and its compress-mcp layer are what turn "shrink this deck" into a finished task — and read more essays like this one at https://www.uglypear.com/en/blog/. The chores will keep arriving either way; the only question is whether they keep costing switches.