Anthropic shipped Reflect on July 9, a built-in dashboard at Settings that shows how you actually use Claude
The monthly recap surfaces your top topics, your most active day, your peak hour, and observations about how you work
In beta on Free, Pro, and Max plans, on the web and Claude Desktop, so most people can open it today
I ran it on myself for a week and the gap between what I thought I used AI for and what I actually did was the real finding
A usage mirror, built in
On July 9 Anthropic introduced Reflect, a dashboard that lets you track and visualize how you use Claude. It lives at Settings and produces a monthly recap: the topics you spent time on, your most active day, your peak hour, and a set of observations about how you work with the tool. It is in beta on Free, Pro, and Max plans, on the web and in Claude Desktop.
The idea is simple and slightly uncomfortable. Most people carry a story about how they use AI, and the story is usually flattering and usually wrong. Reflect replaces the story with a record. It does not ask what you think you do. It shows what you did.
I like tools that turn a felt sense into a number. I keep a whole stack of them for my own work, which I wrote up in what I run as cron jobs in my Claude stack: 6 automations. Reflect is the same instinct pointed at the one process I never instrumented: my own habits.
It is worth naming the tension up front, because a sharp piece at TechCrunch did on the same day. A usage dashboard from the company selling you the usage is not a neutral mirror. It can just as easily nudge you toward more. I went in with that in mind, and I think the honest read is: useful, if you treat the numbers as data about yourself and not as a scoreboard to climb.
It helps to know what Reflect is not. It is not a productivity score, and it does not rank you against other people. There is no leaderboard, no streak to defend, no number that flips red when you slow down. It is closer to the screen-time summary on a phone than to a fitness band nagging you to move. That framing matters, because a tool that measured you against others would pull you toward performing for the dashboard. A tool that only shows you your own shape leaves the judgment where it belongs, with you. The whole value depends on that distinction holding, and for now it does.
What a week of my own data showed
I opened Reflect expecting confirmation and got correction. That is the whole value.
The first surprise was distribution. I would have told you my Claude time splits evenly across writing, coding, and planning. It did not. A large slab sat in one category I barely think about, the small back-and-forth debugging turns that never feel like real work but quietly eat the day. Seeing it as a block reframed it from invisible to addressable.
The second surprise was timing. My peak hour was not when I feel most productive. It was later, in a stretch I had mentally filed as low-value cleanup time. The recap said otherwise. Some of my best-shaped sessions were happening in a window I had been treating as leftovers. That is worth protecting now that I can see it.
The third was the most-active-day line. One day carried far more than the rest, and it was not the day I would have guessed. It lined up with a recurring commitment that pushes everything else into a pile, so the day after becomes a catch-up marathon. The data did not tell me to change that. It told me the pattern exists, which is the part I could not see from inside it.
None of these are dramatic. That is the point. Reflect is not going to hand you an epiphany. It hands you three or four small corrections to a story you have been telling yourself, and small corrections compound.
The mechanism behind why this works is worth a sentence. Memory keeps intentions, not actions. I remember deciding to spend the morning on deep writing, so that is the story I carry, even on a morning that actually dissolved into small fixes. Memory records the plan and quietly discards the drift. A record does the opposite: it keeps the drift and forgets the plan. Put the two side by side and the daylight between them is the most honest feedback you will get all month, precisely because no part of you chose it. That is the thing a mirror can do that reflection alone cannot, because reflection runs on the same faulty memory it is trying to check.
Using it without letting it use you
The critique is fair, so here is how I keep the tool honest rather than the other way around.
Read it as a mirror, not a target. The moment "peak hour" or "most active day" becomes a number you want to grow, you have inverted the tool. More usage is not the goal. Better-placed usage is. I look at the recap to move work into my strong windows, not to fill more windows.
Pair it with a subtraction question. Every time I read the recap I ask one thing: what should I be doing less of, not more. The debugging-churn slab is a candidate for automation, not for more manual grinding. If a topic is eating time, the win is usually to systematize it, the way I turned a daily writing task into how I auto-publish a blog article per day with GitHub Actions and Claude, not to spend more raw hours on it.
Keep the cadence monthly, not obsessive. This is a recap, and it earns its value across weeks, not by refreshing it daily and reacting to noise. I gave it a week to get a feel for it, but the real read is the monthly view, where the shape holds still long enough to mean something.
And treat the observations as prompts, not verdicts. The "how you work" notes are a conversation starter with yourself. Some will land, some will not. Take the one that stings a little, because that is usually the true one.
Guard against one failure mode in particular: turning a mirror into a mandate. It is tempting to read a recap and write yourself a rigid rule, only work on hard problems at your peak hour, never touch messages before noon. Rules like that break on the first unusual day and then you abandon the whole practice out of guilt. Hold the findings loosely. The recap is describing a tendency, not issuing a command, and the right response to a tendency is a gentle nudge you can keep, not a law you will resent by Wednesday. The point is to steer, not to police.
Who gets the most from it
If you are a solo operator, this is the highest-value group, and it is who I am writing for. When you are the whole company, your habits are the whole operation. You have no manager surfacing patterns and no teammate saying "you always do this on Fridays." Reflect is the closest thing to that outside view, and it is free to open.
If you lead a small team, the individual framing still helps you personally, though the tool is about your own usage rather than the team's. Use your own recap to model the habit, then talk about what you learned. The instinct to instrument yourself is contagious in the right direction.
If you are a heavy daily user, the recap will surprise you more, not less. The heavier the use, the further the felt story drifts from the record, because volume hides pattern. You are exactly the person whose mental model is most out of date.
There is one group I would tell to approach it warily: anyone prone to turning self-measurement into self-punishment. If a screen-time number makes you feel behind rather than informed, a usage recap will do the same, and the critique about a company nudging you toward more use lands hardest on you. The tool is only as healthy as the frame you bring to it. If your frame is curiosity, open it. If your frame is guilt, the honest move is to leave it closed and keep your attention on the work itself. No dashboard is worth trading your relationship with your own days.
And if you are skeptical of a usage dashboard from a usage company, good, stay skeptical. Open it once, read it as data about yourself, ask what to cut, and close it. Skepticism and usefulness are not in conflict here. They are the correct way to hold the tool.
Bottom Line
Reflect turns a felt sense into a record. It shows your topics, your most active day, your peak hour, and a few observations about how you work, in beta on Free, Pro, and Max, on the web and Desktop, so most people can open it right now.
The value is not the dashboard. It is the gap between the story you tell about your own habits and what the data says you actually did. Mine were off on distribution, timing, and load, in small ways that added up. Read it monthly, treat it as a mirror rather than a target, and follow every finding with a subtraction question. Used that way it makes you sharper. Used as a scoreboard it makes you busier. The choice is yours, and it is the whole game.
Top comments (0)