Using AI to Turn Data Into Insight You Read at a Glance—I Had Claude Code Analyze My Own Browsing History

Since Chinese New Year, most of my time has gone into Claude Code—working with AI, getting things done through it—and I rarely open a browser to look at web pages anymore. So I got curious: could I use AI to run the numbers and see how much less time I’m actually spending online?

Letting Claude Code go through my browsing history

The idea was simple: hand the job to Claude Code and point it at my Chrome history. My original plan was to make Chinese New Year (February) the dividing line and compare which sites I visited before and after, and how much time each one ate. Then count it all up and lay it out visually as a web page, in the PPVI dark theme.

A snag partway through

It spotted the problem the moment it looked: Chrome only keeps the last 90 days of history, so my Chinese New Year data was simply gone. Fine—adjust on the fly. We switched to comparing the last 30 days against 30 to 60 days ago, which is last month against the month before.

The result: time online nearly cut in half

It came back fast, and it matched my own sense of it—way down:

• Total active time: from about 150 hours a month down to 83 hours, nearly half
• Daily average: from 5.1 hours down to 2.8 hours
• Visit counts dropped noticeably too

One Sankey diagram, and you see where the time went

The clearest part was the Sankey diagram—you can see exactly which sites the time drained out of and which sites it flowed into.

What I clawed back was mostly time spent browsing pages and watching shows (Netflix). Another chunk came from a drop in visits to my own product platform. That project is built now, so it’s in daily use rather than development—I’m not in there every day, and the team produces with it instead.

The new time went into content work: reading and building the blog, YouTube, Xiaohongshu (sites I use when making short videos), and some tools for running company business. Not much else changed, but the overall drop was real and steep—steeper than I expected. Set against the hours I pour into Claude Code every month, my time online is now a rounding error.

The real point: any dataset can turn into insight

What we looked at today is just an analysis of browsing behavior. But the approach travels: give AI a reasonably complete set of data and it will hand you back a read you can take in at a glance, and even go deeper—into what might be causing the pattern. It works just as well on business analysis.

Data on its own is dry. What AI is worth is turning it into something you understand at a glance and can act on.

Coming up: how to do topic selection and monitoring for a creator business, how to write with AI, and how to do SEO and GEO with AI. I’ll be sharing these one by one—stay tuned.

Source: EP0044_audio.mp3 · ASR model gemini-2.5-pro (chunked parallel) · full text of the original recording

[00:00] So I thought of something interesting. I noticed that ever since Chinese New Year, I’ve been spending huge amounts of time in Claude Code, talking to AI, using it—and hardly ever opening a browser to look at web pages. So I wondered whether there’s a way to see exactly how much less time I’ve been spending in the browser. Let’s use Claude Code to do it. So we open up Claude Code. Right. Please pull up my Chrome browsing history for me. Let’s use this year’s Chinese New Year, use Feb—

[00:25] ruary as the dividing line. So the browsing history before February—that’s January, December and November, three months—versus the browsing history after February, that’s February, March, April. What I want you to compare is: which sites did I browse, how much time did I spend. I need real numbers on it, laid out as a web page with a browsing chart, so I can see how much my total browsing time dropped, and where that browsing time went—

[00:50] which sites it flowed from and which sites it flowed to. Please use PPVI’s dark style—we’ll go dark for this page. Save it to the notes folder on my D drive. And once you’ve saved it, open it so we can take a look. Let’s speed this part up. In a moment we’ll see—let’s see what kind of result it gives us. Uh-oh, uh-oh, this might go sideways. After checking, it found that Chrome only keeps the last 90 days of history, so it can’t find my Chinese New Year—

[01:16] records. That’s fine, let’s have it adjust. Here’s how I’d put it: okay, got it, I understand the situation. Let’s drop the date-range comparison I asked for. Change it to the last 30 days versus 30 to 60 days ago—so last month’s browsing data against the month before that. We should be able to do that, right? I can see there’s detailed data. Alright, let’s wait for it. Feels like this is going to go sideways.

[01:42] Okay, it’s done now. And it matches my own sense of it exactly—way less. Let’s take a look. It compared April 19 to May 19 against May 19 to June 18. Very clear. It leads with some conclusions. First, total active time went from 150 hours a month to—nearly cut in half—83 hours a month. Daily average went from 5.1 to 2.8 hours.

[02:07] Visit counts dropped too. Let’s look—let’s look at this Sankey flow diagram. You can see the overall drop in time, it’s this much, close to half. And then where the new time went. You can see the new time went into xntj.ai—that’s browsing and building my own blog. Then YouTube, then YouTube, Xiaohongshu—those are sites I end up scrolling while I’m making short videos.

[02:32] And where the freed-up traffic came from: xntj.tv. Under the xntj.tv domain there’s an e-commerce platform and an AI comic drama platform. Visits there dropped because development on that project is finished—we’re just using it now. I don’t use those tools myself; the team’s the one using them to actually produce stuff. And Netflix is down in here too, right? Less TV and movies. And then Feishu, and Mercury—because last month I actually opened a—

[02:58] Mercury company card, so I spent some time on there getting the paperwork through. Nothing else changed much. But overall it’s still down a lot. I honestly didn’t—didn’t—didn’t expect it to drop this much. Compared with the time I put into Claude Code every month, it’s nothing, absolutely nothing. It also summarized how my time shifted across different categories of activity—which ones dropped the most, which ones grew the most. And then a full site-by-site—

[03:23] breakdown of time spent. What we looked at today is an analysis of browsing behavior. But this approach works in a lot of places. As long as we have reasonably complete data, we can get AI to hand us an analysis we can read at a glance—and even go deeper for some real insight, dig into what the causes might be. It works really well for business analysis too. That’s it for today. Tonight I’ve actually got an in-person—

[03:48] three-hour talk, and I’ve prepared a lot of real substance for it. So starting tomorrow, on the content-creator side of things: how to pick topics and keep an eye on them, how to use AI to help with drafting, how to use AI for SEO and GEO—I’ll be covering those topics one after another. Stay tuned. See you next episode, bye-bye.