65 Years Old x Pixel x Voice Journaling

Since March, I purchased a Pixel, and I have been using it for about four months. I turned 65 on June 30th. I have rounded the "final corner of life" and joined the ranks of the "elderly." I think it arrived just in time for this next chapter.

During this time, my work and life have changed completely. It even feels as if "myself" has been updated.Never before has a single device changed things this much. The impact was on a different level than when I first got my hands on a computer or smartphone.

This is because I can use three apps—Google Recorder, NotebookLM, and Gemini—in seamless coordination. For me, the benefits are so great that it is no exaggeration to call them my "three sacred treasures."

Here, I would like to take voice journaling as the prime example of this, and convey the impression I received from the Pixel along with my gratitude.

1 intro | The Challenge of a Fair-Weather Diarist

(1) Why I Can't Keep a Diary

I don't know how many times I've thought about writing a diary until now. How many notebooks and diaries have I bought for that purpose? It's not just handwriting, either. I've tried several diary apps. However, none of them lasted. I quit them all immediately.

I write on note every day. So why can't I write a diary?I asked ChatGPT, an AI that seems to have a strength in analyzing user personality. After discussing various things as if I were receiving counseling, it told me that I probably struggle with diaries because I have a personality like ① to ⑤.

In addition, it recommended that "journaling" might suit me better. Journaling? What's that? I looked it up. It was a form of self-therapy called "writing therapy" developed in the 1980s by American social psychologist James W. Pennebaker.

To briefly summarize the difference from a diary, a diary focuses on recording facts such as events, emotions, and thoughts. In contrast, journaling focuses on descriptions that dig into subjectivity and the inner self, such as recognizing facts, how one felt, and why one thinks that way. I see. I certainly felt that it suited me.

(2) The Inspiration of Voice Journaling

However, if I were to "write," given my personality of "seeking perfection in writing," I might get hung up on structure and expression, and my progress might stall. In that case, I thought, why not just use "dictation (voice)" instead of "writing"?

As a teacher, I often speak in front of people, but I don't really get better at speaking. So, since getting the Pixel, I have been recording my talks. I use AI to analyze and evaluate the text transcribed from the recording, and I use that feedback for my next speech.

Until now, I never imagined that someone would analyze, let alone evaluate, my speech. Being a small-minded person, I don't think my pride could handle being criticized by people. However, the Pixel is a machine. It's ridiculous to harbor such fears.Unless I specifically instructed it to "criticize harshly," I was able to accept the AI's analysis and evaluation honestly.

Moreover, the repetition of approaching a speech with a task, analyzing and evaluating it afterward, and then applying it to the next one is rewarding in itself. The speech I used to hate is not so bad anymore. Since that's the case with the speech feedback I do occasionally, I thought it might be even more fun if I did voice journaling every day.

2 framework | QHJ for Timeline Description

(1) A "Running Partner" for Journaling

Journaling is called "writing meditation." There is meaning in "writing," so I wondered if "speaking" journaling was nonsense. So, I consulted Gemini this time, asking, "What do you think?"

This is because, while ChatGPT is good at understanding human emotions and verbalizing thoughts, I thought it was unsuitable for remembering and managing daily journaling. In that respect, Gemini, which can coordinate with NotebookLM—a management tool—is perfect as a running partner for journaling.

Gemini agreed, saying it was a great idea. As usual, it explained the reasons for this in various ways, but even after reading them, it didn't quite click for me. Regardless, I understood that "spoken" journaling is not entirely meaningless. So, I decided to give it a try.

(2) "Free speaking" where I can't speak freely

So, how should I do it? When I searched, I found there was no set method. Since there are many ways to do it, I felt I couldn't start without deciding on "my own way." I'm the type who can't attack without securing the outer defenses first.

I immediately begged Gemini for help, but it suggested I shouldn't overthink it and just try talking for about 20 minutes. Since I'm a chatterbox, I thought 20 minutes would be a piece of cake, so I tried it right away... but it was a total failure. I couldn't speak. It was pathetic.

When I actually tried to speak, I got nervous and my mind went blank. Even if I forced myself to squeeze words out, it ended immediately. When I looked at the recorder thinking I had talked for quite a while, not even a minute had passed. I gave up right away.I realized that "free speaking" journaling was impossible for me.

(3) QHJ Framework

So, I decided to create a list of topics to talk about during journaling.I instructed Gemini to provide a list of appropriate topics for journaling. I stared at it intently, selected what I needed, rearranged the order, and turned them into short questions that I could display on my smartphone.

20 minutes is also too long for a beginner to talk. I shortened it to 15 minutes and, while I was at it, thought of a name.How about Quarter-Hour Journaling (QHJ)?

The key point of the list is that it is structured with a chronological awareness of current status -> past -> present -> future.First, I check my current state. Next, I decided to go back to the past and gradually shift to analysis and reflection while organizing the facts. I thought that by doing this, I could naturally transition into the abstract thinking I enjoy, following the timeline. With this, the outer defenses were secured.

3 Input | Google Recorder: Placing goals outside

(1) Super-fast transcription recorder

Since buying a Pixel, Google Recorder (GR) might be the app I use the most. That's how valuable it is to me.GR transcribes what I am saying as I speak.Watching my own spoken words turn into text right before my eyes is even pleasurable.

Pixel has a large language model AI called Gemini Nano integrated into the OS that is always running. Therefore, it works with the speech recognition AI to organize and summarize the transcribed text without communicating with cloud servers. Because everything is processed on the device, it is super fast and privacy is protected.

However, the transcription makes quite a few mistakes. For example, in this morning's journaling, I spoke about 3,000 characters, but there were about 20 misinterpretations and such. Perhaps GR is complaining that my pronunciation is bad in the first place.

How do I deal with GR's misinterpretations?I certainly can't speak it all over again. A one-time, serious attempt is the real thrill of voice journaling. So, should I correct the misinterpretations or leave them? After thinking about it, I decided not to correct the misinterpretations in the main text (leave them as is) and instead manually correct the summary generated by the AI.

(2) Now, let's start QHJ!

After practicing a few times, I became able to talk for about 15 minutes while looking at the "QHJ Topic List." I thought it would be best to do it at a set time as much as possible to make it a habit.Since I set a "Focus for today," I decided to record around 6:00 every morning.

Current Android smartphones support split-screen. If you set a multi-app icon on the screen, you can display GR and the list (written in Google Docs) at the same time with one touch. Preparation is complete in an instant. Then, I slowly tap the large red button on GR to start recording.

After saying, 'It is [Month] [Day], [Day of the week], [Time], the weather is [Weather], starting QHJ,' I look at the list and speak the written questions, for example, 'How did I wake up this morning?' My opening is always the same, but I find it easier to speak that way.

Once I start talking, I continue as thoughts come to mind and as I feel inclined.When one topic hits a dead end, I move on to the next question. However, I do not intend to answer every question on the list. Out of the three questions in one category, I only answer one or two.

I don't worry about the order either. For example, while talking about past facts, it often turns into 'realizations or flashes of insight gained in life.' In those cases, I make sure to preface it by saying something like, 'This will be about my outlook for tomorrow from here on.' I do this to be conscious of the shift in my speaking mode, and I think it also helps the AI when it analyzes the content.

(3) The effect of speaking out loud is outstanding

I actually started working on QHJ on May 26th. July 8th is my 44th day. I accidentally forgot one day, so I have recorded for 43 days. I am currently updating my 'diary' consecutive record.Why does it continue? When I think about it, it's because I can feel the effects and it's fun.

First is the 'Focus of the Day.' Before this, I had never listed what I needed to do today, let alone thought about prioritizing those tasks. I had a tendency to consciously avoid such troublesome things.

Now, I am happy to think about it just to fill those '15 minutes.' I don't just think about it; I use my throat and tongue to speak it out loud and listen with my own ears. In addition, if I read the text being generated at the same time, it gets fixed in my memory whether I like it or not, and I remember it many times throughout the day.

Furthermore, if a goal is not met, I have to speak about it and reflect on it in the next day's journaling. This is painful for a small person like me. Because of this, it has become clearly more common for me to achieve my daily work goals without missing any.The 'Focus of the Day' was the first harvest of my journaling.

4 process | NotebookLM that scoops up consciousness

(1) A super excellent librarian AI for my private library

There is nothing more embarrassing or boring than reading your own spoken words. So, I decided to summarize the main text of my journaling. For the summarization app, I chose NotebookLM (NLM) with future processing in mind. It is an AI that generates answers and the like by synthesizing only the materials added by the user.

NLM can store materials equivalent to a local government library. Moreover, it remembers every corner of the library. Even if I continue journaling from now on, I won't run out of capacity, and it remembers all the journals.

Using those resources, NLM summarizes my daily reports appropriately. Actually, the [Examples of misconversion by GR, etc.] mentioned earlier were part of what NLM showed me when summarizing. In other words,by summarizing with NLM, I am correcting some of the misconversions by GR to a certain extent.

Note that it is stated on the official website that NLM does not use added materials or the content of chats exchanged for training the AI model. Since it seems to handle everything while protecting privacy, I feel safe adding any content.

(2) Compressing the journal to 1/3

Once the recording is finished, I convert the transcribed text into a Google Document (GD) and save it to Google Drive.GR has a function to send transcriptions directly to NLM, but I intentionally don't use that function and take the extra step to make it a GD file. With this, the daily report text for journaling is finalized.

Next, I summarize the daily report text. I launch NLM, create an appropriate notebook, add the GD file to the source, and then, for example, write a prompt like the following.

'260701 Daily Report' is the filename of the GD file. (1) and (2) are instructions to make the summary easier for me to read. (3) is because I thought it would be easier for the AI to search if there were tags. When I press Enter, the summary is completed in a few seconds.

The length of the journaling varies by day, but the current average is about 3,000 characters. NLM summarizes that into about 1,000 characters by following the central context and dividing it by item so that it is easy to look back on later.

(3) The current 'self' brought into focus through metacognition

I copy the summary from the NLM chat. I return to the original GD file, paste it either before or after the main text, and then make final adjustments to the summary within the GD file.

While reading the summary, I sometimes find proper nouns that haven't been corrected yet. Also, perhaps because the way I spoke was poor, I sometimes find misunderstandings or a lack of nuance in the summary. There are also times when I feel like adding paragraph breaks.

If I find parts I want to fix, I manually correct them as I see fit, one after another.The important thing here is to correct only the summary. I leave the main text uncorrected. This way, the uncorrected main text and the corrected summary are contained within a single file.

It is only a few minutes of work, but a strange sensation arises once the manual corrections are finished. It feels as if my field of vision opens up, just like a camera lens snapping into focus.It might be a strange way to put it, but it also feels like I am 'giving up' on myself.

In my daily life, I spend my time vaguely without being very conscious of 'myself.' Yet, in the mere 1,000 characters or so in front of me, the consciousness surrounding my current 'self' is displayed. There are daily achievements spanning work and personal matters, as well as challenges; there are things I am satisfied with, and things I feel dissatisfied with.

Various thoughts are organized into solid words, along with the context and logic that precede and follow them. My current 'self' is neither more nor less than this. It is by no means impressive, but at the same time, it is not something I can immediately reflect on and improve.

Even so, it is the honest, authentic, raw version of myself. Realizing that is refreshing. It gives me a sense of clarity. Perhaps you could call it 'resignation.' I think the greatest benefit of QHJ is this feeling of resetting such a 'self'.

(4) Complementing the source through an ecosystem

After correcting the summary, there is one more procedure to perform by returning to the NLM. It is to tap the source in the NLM and synchronize the source with the Google Drive GD file.

Through this synchronization, the source, which now integrates the main text and the summary, is saved in the NLM. This is the biggest reason why I 'take the extra step to make it a GD file.' The AI becomes capable of analyzing using both the uncorrected main text and the corrected summary as sources.

That concludes today's journaling. Including the recording, it takes less than 30 minutes once you get used to it. You can record while walking, and you can do the summary, including corrections, even on the train. I did it this morning while sitting on a flower bed at school. That's because I forgot the key to the school building.

Moreover, the GD file remains open throughout the entire series of tasks. I can add an open file to the NLM and synchronize it. I was surprised at first because I was used to things like Word.The fact that I can do QHJ regardless of time and place is only possible because of the Google ecosystem, which is surprisingly light and easy to use.

5 output | Gemini threading the fragments

(1) An AI that understands the ups and downs of life

At first, I only thought about daily summaries. However, by the time I had been doing it for about 10 days, I gradually started to feel lonely. Nothing remained in my hands. All I had was digital data. After all, it is fun to look back at a 'diary' when it's a physical book.

So, I decided to print it out. However, printing the daily main text and summaries is uninspired. With about 3,000 characters of main text and about 1,000 characters of summary, that's about 4,000 characters a day. That's about 4 sheets of A4 paper. It becomes a huge volume in no time. It's a waste of paper, and it's a waste of time to read. Ingenuity is required.

Then, I'll make a weekly report and print that!It was just a whim. First, it started with the volume. I decided on double-sided printing on one sheet of A4 paper. In other words, about 2,000 characters. I compress 7 days' worth of about 30,000 characters into about 1/15th of that size.

Of course, if it were just about compressing, it would be possible with the NLM. But what could happen if I had the encyclopedic Gemini, which can access the world's knowledge, read a neatly organized source?I thought that with Gemini, I could summarize the week from various perspectives and have it analyze and evaluate me as an AI mentor.

(2) Creating a weekly report prompt

For that, the prompt is important. So, I created a separate notebook in NLM, added about 10 websites that explain "how to write effective prompts" as sources, interacted in the chat box, and created a [Weekly Report Prompt] while incorporating my own wishes.

The biggest point was that I wanted it to be "something that a third party could bear to read."I wanted the content to be such that even if the reader didn't 100% agree with my decisions or thoughts when they read it, they would have a certain level of understanding, thinking, "I see, that's how he was thinking."

However, because I expected a narrative quality, I ended up emphasizing the emotional side, and it seemed to be a bit too much for Gemini to handle. I ended up reflecting on this later.

(3) Gemini and NLM's |Gem| Collaboration

For the actual weekly report, I used a feature of Gemini called Gem.Gem is a feature that allows you to have Gemini memorize a single prompt to perform the same processing repeatedly. It is an extremely excellent feature, but it is extremely easy to use.

First, I pour the [Weekly Report Prompt] above into the Gem definition screen to create a working Gem. Next, I launch the working Gem and have it upload the notebook created with NotebookLM. Then, I just type and execute a very simple prompt like the following, for example.

Note that for personal Gemini (free and paid versions), the content of chats and uploaded data may be used for AI model training or viewed by humans in the process. Therefore, I set "Gemini App Activity" to "off," and I am careful not to include personal or confidential information, such as by redacting specific personal names of students or staff.

(4) Resilience Strengthened by Weekly Reports

In this way, I became able to hold and read a weekly report printed on both sides of an A4 sheet of paper every week. I became able to not only read the summaries of daily reports separately but also to overlook the transitions of the "self" over time.

After continuing for more than a month, naturally, various things happen. It is by no means only smooth days. Suddenly, my physical condition may deteriorate, or unexpected situations may erupt in the organization. Sometimes I get disappointed and depressed, and sometimes I get hot-headed and shaken.

Even at such times, I continued QHJ. That said, I was just maintaining my daily habit. I carved out time in the morning, and for the 15 minutes after tapping the red button, I focused only on what to talk about. Every morning, I spoke calmly, corrected the summary of the daily report, and at the end of the week, I followed the weekly report with my eyes.

Among the items in the weekly report, "2 Chronology" and "3 Focus" are particularly interesting.Even if you are living an ordinary daily life, one day you may suddenly fall into an unexpected situation. Until that day comes, the carefree self who knows nothing looks comical. After that day comes, you will re-experience the feelings that rise and fall as the situation progresses.

While repeating this process, I began to think that QHJ might have the power to restore the "self." In fact, in the "AI Profiling" of "4 Transitions," I was sometimes evaluated by Gemini as having "phenomenal resilience."

Reading the printed weekly report is to objectively view how one's feelings flow.Is it closer to the current "self" to see the "self" as a line rather than as a point? I don't know the mechanism well, but I feel like it encourages me to "turn over a new leaf" and keeps me moving forward.

However, on the other hand, "5 Reflections" and "7 Three Questions" are low-level. "5 Reflections" has a shallow understanding of emotions and thoughts, and I feel like I am reading a cheap novel. As for "7 Three Questions," I don't even understand what is being asked in the first place. I began to think that I must improve this.

Also, unfortunately, I realized that my own shortcomings are sometimes the cause of bad situations. I want to do something about this. For example, can I create a mechanism where an AI mentor watches my thoughts and actions and checks them in advance? If I could avoid even one or two bad situations by doing that, there would be nothing more wonderful.

6 Management | Will QHJ become a compass for the heart?

(1) Gemini's "Focus on Strengths"

While creating a [Weekly Report Prompt] with Gemini, I asked it: 'When reading about 30,000 characters of unstructured conversation, is it easy or difficult for Gemini to organize and summarize it into about 2,000 characters?'

The answer was, 'It is both easy and difficult.' For an AI that can easily summarize 100,000 characters of text, the volume is not an issue. However, it told me that 'weaving a story' is difficult for it.For Gemini, which excels at statistical prediction, it seems to be a burden to empathize with my emotions and depict the story behind the facts.

I asked, 'In that case, if I don't ask you to make it a story, can you provide a sharper analysis?' Gemini replied, 'If I don't have to worry about the story, it is easy, for example, to add a thorough analysis from the perspective of self-management, as Drucker says.'

This is exactly what I wanted. I am not a baseball team advisor, but I am a self-proclaimed |Druckerian| (a huge fan of Drucker). Taking Gemini's idea, I decided to improve QHJ from the perspective of a knowledge worker belonging to a non-profit organization called a 'school' practicing self-management.I decided to improve QHJ from the perspective of a knowledge worker belonging to a non-profit organization called a 'school' practicing self-management.

(2) Self-Management Prompt

I discussed daily and weekly reports with Gemini again. Gemini said that if I am going to introduce self-management, I should have an AI mentor evaluate and analyze it from the daily report stage. I hesitated. I don't want to do something so grand every day.

However, Gemini did not give in. It repeated that Drucker placed the most importance on daily self-management, so I should include the AI mentor's perspective in my daily reports as well. I cannot win against that logic. I decided to do as it said.

So, to set up the new Gem, I created new prompts for both daily and weekly reports. I will pick out the parts I thought about especially carefully.

I only started this self-management QHJ on July 6th. As for the weekly report, I have tried it with old data, but I haven't actually created one yet. If I actually try it, I might ruin the QHJ that was going so well.

However, for me, it is very worth working on. Building a self-management system based on my own inadequacies will undoubtedly be an epoch-making event in my working life.

7 outro | QHJ that made it to the final corner

(1) QHJ that expands new possibilities

'Use your digitized materials to the fullest.' I remember a senior teacher who taught me the basics of computers about 40 years ago saying that. If I compile my QHJ in an NLM notebook, I can truly 'use it to the fullest.'

For example, NLM has a feature called 'Studio.' It is a collection of automatic generation commands that process selected sources into various outputs. If you use commands like 'Mind Map' or 'Report,' you can literally create a 'map of your heart' or a tailored report with a single tap.

Furthermore, with Gem, it is even more as you wish. When I asked Gemini, it proudly said that it could create a Gautama Siddhartha Gem, a Marcus Aurelius Gem, or an Alfred Adler Gem.

Ten years ago, I first learned the concept of Life Logging in Kevin Kelly's book 'The Inevitable.' At that time, I didn't have a real sense of it, but after 10 years, AI has become familiar, and after trying QHJ, I feel like I finally understand what he meant in my own way.

Through QHJ, |Quantified Self| (self-tracking) that seamlessly objectifies both my work and private life has become possible. If I record that with AI, I can even achieve |Total Recall| (recalling everything in life). Furthermore, is it too optimistic to think that based on life logs, I will be able to choose a better way of life that I desire while consulting with AI?

(2) From Quarter-Hour to Quarter-History

I started QHJ with a light heart at first. However, through AI integration, I was able to build my own 'foundation.' Going forward, it might even contribute to my self-management. For me, QHJ has now become something indispensable.

Recently, I've gotten used to speaking, and I often go over the 15-minute mark. I have a feeling that soon, I won't be able to call it a 'Quarter-Hour' anymore. Should I change the name when that happens?

No, if I have just rounded the 'final corner,' I might still have about 20 years left to live. That means about a quarter of my life remains. In that case, I could call it a 'Quarter-History.' Please laugh and forgive an old man's ambition for using such an exaggerated word as 'History'.

All that remains is the straightaway to the finish line. The finish line for the last spurt that is about to begin is not yet in sight. Rather than looking that far ahead, I think I will continue my Quarter-History Journaling and race through the 'now'.