There's a number from ActivTrak's 2026 workplace report that I haven't been able to put down. The average knowledge worker's focus session — an uninterrupted stretch of real work — is now 13 minutes and 7 seconds. Down 9% in three years. Focus efficiency, the share of the workday actually spent in that state, fell to 60%. A three-year low.

Same report, same year: 80% of employees now use AI tools, up from 53% two years ago. The average org runs seven of them, up from two. And here's the part that should stop you — most people in these surveys say AI made them more focused. A separate study tracking real output across thousands of workplaces found the actual productivity gain came to about 3%, while three-quarters of workers were sure AI was saving them serious time and sharpening their attention.

So we have a feeling — I'm more locked in than ever — pointing one direction, and the clock pointing the other. Thirteen minutes. That gap between how the day felt and how the day went is the most interesting thing happening to knowledge work right now, and it's exactly the gap Timex was built to close.

Why AI shreds attention while feeling like the opposite

I noticed this in my own days before I read any report, because I use Claude Code all day and I watch my own Timex log.

The shape of AI work is a trap for attention. You give the model a task — refactor this service, draft this migration — and then you have thirty to ninety seconds of nothing to do while it thinks. That gap is too short to start real work and exactly long enough to check something. Slack. The PR queue. A tab. By the time the model comes back you're three contexts deep and have to climb back into the problem you were supposedly working on.

Do that forty times a day and you've invented a new way to feel busy and productive — you shipped, the model did the thing — while your longest continuous stretch of actual thought was, yeah, about thirteen minutes. Researchers have started calling the end-of-day version of this "AI brain fry": the fog and slowness that builds up after a day of prompting and then double-checking what the machine produced. The early 2026 research on heavy AI users keeps landing on the same finding — measurably more mental fatigue and information overload than lighter users report. You're not lazy. You're context-switching at a rate no human was built for, and the tool that caused it is also the tool telling you it made you faster.

The feeling is real. It's just not evidence. And you cannot fix a problem you can only feel.

Feelings don't have timestamps

This is the whole reason I don't trust manual time tracking, and by extension why I don't trust my own sense of a day. Ask me at 6pm how I spent it and I'll reconstruct a flattering story: solid morning of coding, a couple of meetings, some email. Ask Timex and it hands me the actual timeline — second by second, because it samples once a second into that local SQLite file — and the story is different every single time. The "solid morning of coding" was forty minutes of coding, cut nine times, wrapped around a Slack thread I'd genuinely forgotten happened.

That's not a guilt trip. It's the raw material. You can't shorten the meeting you don't remember sitting in, and you can't protect the focus block you don't know keeps getting cut at minute eleven.

The specific thing that makes this work for AI-era days is that Timex knows the difference between the tools. It doesn't see "you were in a browser for two hours." It sees github.com and a pull request as focused development, and an autoplaying YouTube tab in the same browser window as something else entirely, because in 1.2.0 the categories follow the website, not just the app. Claude Code getting real work done and the Slack you fell into during the model's thinking gap land in different buckets. The log can finally tell your productive AI use from the fragmentation it dragged in with it — which is the exact distinction the feeling can't make.

A score you argue with, not one that grades you

Timex turns that log into a Focus Score — one number, 0 to 100, at the top of your day. But I need to be careful how I describe it, because the entire attention-economy problem is tools that hand you a verdict.

This isn't that. You write the rubric. Every category carries a weight you set, from deep focus to pure waste. If your best thinking happens in long Slack threads with a co-founder, push communication up and the score follows. If you believe an hour of aimless reading is where your ideas come from, count it. The score is arithmetic on numbers you chose. It's a mirror you ground yourself, not a grade someone else assigns — and there's no streak, no nag, no weekly "you failed Tuesday" email. If you set every weight to zero, Timex goes back to being a plain honest log, which was always the real product.

What the score gives you that a raw timeline doesn't is a single glanceable answer to was today the day I meant to have? — and, over a week, whether the thirteen-minute problem is yours too, or whether you're one of the people the average is hiding.

The fix isn't more software pointed at your attention

Here's where I'll disappoint anyone hoping for an AI feature. The answer to attention shredded by tools is not another tool that watches your attention harder. It's two boring things: seeing where the time actually went, and taking real breaks so the fog doesn't compound.

That second one is why Timex has a break timer at all, and why it's the part I refuse to charge for. Not Pomodoro's rigid slice-your-focus-into-25-minute-blocks — I never got along with that; it chopped up the good stretches to prevent the bad ones. A break that arrives when you've actually been heads-down for a while, offers a 30-to-90-second reset — water, look 20 feet away for 20 seconds, roll your neck, stand — and then leaves. The thing "AI brain fry" research keeps landing on is that the fatigue is cumulative across the day. The cheapest intervention against cumulative fatigue is not letting it accumulate.

The AI made you feel like a machine that never needs to stop. The thirteen-minute number is what actually happened to the human running it. Timex just shows you both, honestly, on your own Mac, and lets you decide what to do about the difference.

The trial is 100 hours of tracking — no account, no card — and the break timer stays free forever after that. Download it, work a normal week, and look at your real focus sessions. I'd bet money the number surprises you. It surprised me.

— DK