What deep work time can tell you
Six hours with an editor open is not the same as six hours spent on one problem. Hina's Deep work number counts stretches of work-type activity of at least 10 minutes. Here is what it counts, what it leaves out, and how to use it without grading yourself.
How a block of deep work is built
Hina sorts every app into a category, and each category belongs to one of three groups. Coding, Writing, Design and Meetings are in the Work group. Browsing, Comms, Reading, Utilities and Uncategorized are Neutral, and Media is Leisure. Deep work is built only from the Work group. A continuous stretch of work activity forms a block, and a block counts as deep work once it holds at least 10 minutes of that activity.
Switching between work categories does not break a block, so Xcode, then Figma, then a meeting call and back is one stretch. Leaving work does break it, with two tolerances. A non-work detour shorter than 5 minutes, such as a chat reply, is absorbed and counted as an interruption instead. A gap of up to 15 minutes in which you were mostly without keyboard or mouse input, at least 80 percent of it, also bridges the block, so a phone call or a long build does not split a morning in two.
The unit is a block, not a clock, which makes the number easy to misread. A block can span 90 minutes on the clock and hold 70 minutes of work, with the rest spent on short detours. The report states the minutes of real work, and it names the longest block with its start time, its app and the project it matched.
What the number leaves out
Reading documentation in a browser, answering a message a colleague is waiting for and reviewing a design in a chat window are all real work, but Browsing, Comms and Reading are Neutral. They never add minutes to deep work, and once they last past the 5-minute tolerance they also end the block, so a day of research can show very little of it. The number measures stretches of work-related activity, not flow or outcomes.
It also cannot see what is on the screen. Hina reads which app was in front and whether there was input, but app time and input counts do not show what changed in the code or prove that you stayed on one problem. When a category does not fit the way you use an app, right-click its entry in the Trail, choose Classify as and pick another category. The same menu offers Restore automatic classification. Change a category so the record describes the activity correctly, not to lift the total.
Interruptions and short switches are different counts
Hina counts two kinds of leaving. An interruption is a break of under 5 minutes in the middle of work, credited to the app or website you went to, and the Interruptions card on the Week view lists the biggest sources. A short switch is leaving a work app and coming back to the same app within 3 minutes with no idle time in between. Handing a task to an AI agent and glancing at something else while it runs has exactly this shape. In Day and the report, short switches appear as a count of short switches with a median length, along with the most common switch from one app to another.
Neither count is a verdict. Moving to documentation may be the task, a chat window may hold the answer you are waiting for, and a day with many short switches is a normal shape when an agent runs in the background. On Day, the line labelled Worth a look names the biggest interruption source when a day has at least 5 interruptions. It raises fragmentation only when more than half of the day’s stretches were under a minute, deep work was under 30 percent of active time and at least 20 sessions were recorded, and it words that as an observation.
Using the numbers for a decision
The Week view has Golden hours, which groups deep work by the hour a block started and finds the hour with the most. Hina words it as a suggestion to plan important work then only when at least three days that week had a deep block starting in that hour. Otherwise it simply states what happened. The Month view adds a Deep work calendar and a pattern by weekday.
A test that works is to change one thing, such as moving messages to two fixed times or opening the documents and tests before a session, and compare the same weekday over two or three weeks, not yesterday against today. The Week view's Versus last week card compares only the days of this week that have already finished with the matching days of the previous week, and today stays out while it is still running, so a comparison read during the day never sets half a day against a full one.
A low number on a hard day is information about the day. An afternoon spent untangling one stuck question can hold 40 minutes of deep work and still be the day that mattered. Hina has no productivity score, and a short day in the record is not a judgment of you.
When the number looks wrong
Check the record first. Time when Hina was paused, not running or the Mac was asleep has no records, and hovering over the title of the Last 7 days chart shows a note that a gap means no recorded activity, not necessarily no work. After 10 minutes to 2 hours without input, with the Mac awake and not locked, Hina may ask once when you return whether you were in a meeting, reading or thinking, watching a video, or away. If you do not answer, the stretch stays away.
Then check the classification. A large share of Uncategorized time sits in the Neutral group and never reaches deep work, even if you were writing code in that app. The Trail shows the category of each entry, and the same right-click menu fixes it.