Another by-product of my AI experiments.
A while ago, unwilling to give up on the idea of AI transcription, I tried using ChatGPT’s voice input. The instruction was simple: keep my tone and opinions as much as possible, but remove the excessive pauses, filler words, and repetitions that come with speaking. I can see why many people find voice input useful. It can already recover eighty or ninety percent of what you mean.
But turning spoken output into a finished essay is another matter. The structure still has to be reshaped, and structure is usually the part I hesitate over most when I write—how much of that effort actually makes it into the finished piece, and how much a reader can feel, is a separate question.
When I pasted the result below, I still couldn’t resist fixing one of ChatGPT’s worst habits: putting every sentence on its own line. I manually merged a few paragraphs and added back a description about the scale that I had missed while speaking.
A few lines later, I forced myself to stop, partly to preserve the shape of the AI-produced text.
The speed is hard to deny. I spoke for five or six minutes, maybe. If the goal is to get something roughly usable out quickly, or to communicate in a low-stakes way, AI really does work.
The other problem is harder to ignore: once a shortcut exists, it becomes difficult to go back. I could rewrite this whole thing properly, but now that AI has already produced a more-or-less usable version, my own effort starts to look like unnecessary suffering. It is a little like what happened after I began using AI for programming: I have almost completely forgotten all kinds of JavaScript and Python syntax. I never built a solid foundation, yet I rushed onto the road and now cannot seem to stop.
The podcast mentioned below is Natural Selection. As for the article mentioned later, I can no longer remember where it was.
Why do I feel that I don’t need data as much as I thought?
I have a smartwatch that I have used for more than four years. Starting last year, its battery began to deteriorate noticeably. At first, I only needed to charge it once every ten days or two weeks. Later, it was almost down to one charge per day.
That was when I began to wonder, again and again, whether it was time to buy a new watch.
At the same time, there was another small annoyance. I always take the watch off when I shower, but if I want sleep data, I have to put it back on before bed. And every time I do, there is still that faint sense of having something strapped to my wrist. For a while, products like the Oura Ring appealed to me because they seemed to promise a form of tracking that could become almost unnoticeable.
But the Oura Ring had already crossed the line of what I considered a reasonable price. Then it introduced a subscription model: without paying, you could not fully view data that was, in theory, your own. That made the whole thing feel even more absurd.
Even then, I did not completely give up. I started looking for alternatives: Oura’s main competitors, Amazfit, some newer brands around the 150-dollar range, and even various Huaqiangbei-style dupes.
While I was stuck in that indecision, my watch battery finally reached the end of its life.
Then I discovered that replacement batteries for this model were available online. I bought one, replaced it myself, and the battery problem was solved.
Still, a small part of me continued to want a ring-shaped tracking device.
At the same time, another question became harder to avoid: why do I need to record my sleep data with such precision? Do I really need it at all?
I have always known that sleep tracking on consumer devices is not especially accurate. No consumer-grade product can truly determine sleep states with precision. When I look back at my own data, the problem is obvious. There were nights when I was clearly asleep, but the watch decided I was awake. Other times I had already woken up and was simply lying in bed, yet it continued to count that as sleep. As for the more detailed categories—deep sleep, light sleep, REM sleep—they are even harder to fully trust.
Later, I read a magazine piece about self-tracking. It raised a question: when data gradually turns into a form of surveillance directed at ourselves, do we really need it that much?
The piece itself did not give me much new information, but it made me reconsider something: how much data about myself do I actually need? How dependent have I become on these numbers to decide how I am doing?
Unexpectedly, only a few days later, a podcast I like, Natural Selection, released an episode about data.
The episode discussed many people’s fascination with data: the desire to digitize the self, to record everything over the long term, to keep tracking continuously. It also talked about how companies turn data into a business.
It mentioned that a once-famous genetic testing company had gone bankrupt. That surprised me a little, because the company had been extremely popular in its time.
All of this pushed me back to the same question: if the data is only for me, and not for proving something to other people, do I truly need it?
At work, data can indeed serve as a direct demonstration of ability. In many situations, you have to present metrics, because that is how modern society operates. In that context, I think data is a real and practical tool, a kind of supporting evidence.
But when I return to my personal life, do I actually want these things?
I have always strongly resisted the so-called Lyubishchev time-accounting method. I also resist dividing each day into strict blocks and then tallying what I completed, what I produced, and how well I performed. I dislike forcing my time and my condition into a set of labels.
I can accept data as a starting point.
For example, when I first decide to exercise, I might set a target for daily steps or minutes of activity. Those numbers can help me build a feeling of “I can do this.”
But once I have proved to myself that I can do it, if that requirement stretches into thirty days, six months, a year, or turns into something I must complete every single day, I start to resist it intensely.
Even if the activity itself is something I want to do and believe I should do, the moment a fixed framework appears, I begin to push back.
I am also the kind of person who has no obsession with streaks.
Unless there is a very concrete external reward—money for completing a certain number of consecutive days, for instance—something like a Duolingo streak eventually makes me ask: why does it have to be consecutive? What happens if I miss one day?
So I often break the framework on purpose.
For my personal life, long-term and highly detailed data tracking can very easily shift from being a tool of observation into a kind of panoramic surveillance. My instinct is to reject that surveillance.
One point from the podcast left a particularly strong impression on me: many concepts that now seem self-evident were actually packaged and created through commercial operations.
For example, the familiar idea of “10,000 steps a day” began as a marketing concept designed by a Japanese company to sell pedometers. Even today, they still sell that original kind of pedometer: extremely mechanical, almost without any smart features. Then there are scales and body-fat scales, especially the electronic ones rather than purely mechanical models. Because home-use devices inevitably produce fluctuations, some manufacturers choose a convenient workaround: if you step back on the scale again within a very short time, the device automatically shows the previous measurement instead of actually weighing you again. At that point, can there still be any trust between humans and machines?
This makes me more certain that data is only one indicator among many, not the whole picture.
I do not want to quantify everything in life. I do not want everything to become a chart.
I once seriously considered putting a heatmap on the homepage of my blog.
In the end, I gave up.
It could not truthfully reflect any real state of mine, and I did not want to observe myself in that way.
There are also certain data points that I delete once I realize they have become vanity metrics rather than tools for understanding myself. The writing statistics I used to place in the footer of my blog were like that. At some point, I noticed that what I cared about was no longer the writing itself, but the growing numbers. They were no longer observation; they had become vanity. So I removed them.
Looking back, I seem to have been doing this kind of thing all along. I keep trying to introduce a new framework, and then, usually quite quickly, I begin to resist that framework.
Unless it eventually becomes truly internalized as part of me, I will continue to reject standards imposed from the outside—whether they come from other people, commercial design, or my own vanity.
Perhaps this has always been one of my contradictions. I keep searching for an order that can help me, while constantly refusing any order that tries to define me.