Context switching: do hundreds of window switches mean distraction?
By evening you open Hina’s Trail, and Safari, Terminal, Claude Code, Finder, and Notes appear back and forth. A day’s switch count can reach the hundreds, which is easy to read as being unable to sit still today. The count records which window came to the foreground. Distraction depends on whether the previous task was put down. Frequent window changes do not necessarily mean the task changed with them.
Attention residue is about the task
A large switch count in the evening may bring to mind attention residue, from Sophie Leroy’s 2009 research. It describes moving from one task to the next before the previous one is put down, so attention stays on the earlier task. You might leave an export halfway through to check email, with the line that reported an error still in your head. What stays is that task, not the Safari and Terminal windows themselves.
You edit an export from the blog repo in Terminal, switch to Safari to look up this API’s path, then return to Terminal to see which line the command stopped on. The foreground changed, and the task is still that export. You put the export down to write weekly meeting notes, or to revise a quote that has not been sent. The previous task is still hanging there. A while later you return to the export and have to find that line again. That is the kind of switch where attention stays behind.
Leave one instruction, then look elsewhere
When you use Claude Code or Codex on the blog repo, a common rhythm is to leave one instruction in Terminal, leave that window, and let it keep running. You check the API docs in Safari, look in Finder to see whether the generated file is there, then come back to see which section it has written. You do not spend that hour watching Terminal print line after line. Once the instruction is given, you do your checking alongside it. While you check, the foreground jumps among the editor, Terminal, and Finder. You can step out many times in one hour, sometimes hundreds of times. The Trail shows small pieces, while the same export is still in the original Terminal.
A real change of task is when you do not return to that instruction and turn to the client site repo to change the navigation and footer. The export stops halfway. When you sit down again, you first have to remember which part you asked it to change. That is not the same as checking a path and returning to the same window.
That cost of coming back has been measured. Gloria Mark, Victor Gonzalez and Justin Harris observed 24 information workers for a 2005 CHI paper and found that 57 percent of their working spheres were interrupted; interrupted work that was resumed on the same day came back after 25 minutes 26 seconds on average, with about two other working spheres in between. This is where the often quoted 23 minutes to refocus comes from, but the paper measured how long until people returned to the interrupted work, not how long it took to regain focus, and the figure 23 minutes 15 seconds does not appear in it.
Quick round trips do not cut a stretch
Hina does not cut a stretch of work for the quick round trips, the ones where you come right back. Moving between work apps, such as from a document to Terminal, keeps the stretch going, and a trip of under 5 minutes into another kind of app does not end it either. Such a trip is counted once as an interruption, and the Week view lists where those interruptions came from, without calling them distraction. You switch from a document to Terminal for a glance at whether the compile has stopped, then return to the original document. You can step out like that many times inside the same stretch. Leaving many times in the middle does not mean you were mind-wandering during that stretch.
Deep work records stretches of staying continuously in apps classed as work. When the editor and Terminal stay together on the blog repo’s export, with only a few switches that come right back between them, the stretch still counts as continuous. It describes how that stretch stayed connected, not a productivity score.
See what interrupts a continuous stretch
A day runs from 4 a.m. to 4 a.m. the next day. If you keep writing until 1 a.m., that still counts as the same day, and switches accumulate on that same boundary. What is worth looking at is not this total, but what interrupted a continuous stretch of work. The interruption might be a switch to the client site repo to edit another page, a relatively long stay in Away, or only one quick round trip counted as an interruption. These three do not correspond to the same count. You can scroll that day’s Trail and look at the three separately.
Halfway through the export, the Trail first shows a stretch of the editor and Terminal, with a few quick round trips in between, then a longer pause, and what opens next is the client site repo. The earlier stretch is still the export. The later stretch is already another task. With window titles on, you can more accurately separate switches inside the same task from a change of task.