Put phone away box

[skorytnicki]

This weekend I’m planning to tackle a simple project to reduce my screen time. I dedicated some thinking to this topic and maybe my case will be interesting to you.

Analysing my usage patterns, basically there’s time where screen time is justified (waiting in a queue, commute in public transportation) and there’s time where it very isn’t - for example when kids are home or when I could spend time with my wife. It’s important to “weigh” the time, or classify the time and put proper attention to it. Then, you can also start lifting time from one class to another; goal is to generate high quality time slots and not waste them.

To fit my needs, I have to develop a tool that will measure my off-the-phone time in some slots. I can define them upfront (eg 17:30 - 22:30 from Monday to Friday), or just delete datapoints in beeminder and derail eventually.

The best tool for me would be a “put phone away box” that would send the reports to beeminder. I wasn’t the first person to invent that, but I don’t want to pay subscription for something that I can’t beemind. Also, I happen to have a raspberry pi that is perfect for this use case.

Raspberry is a small computer you can hide in your pocket. Very small, very flexible. You don’t need raspberry or arduino, it will work the same way with your laptop.

The idea is to put raspberry in some obscure part of the house, connect a USB charging cable for my phone and compute time it’s been connected with this cable. Datapoint will contain amount of minutes it’s been there + time range in a comment (so that I know when it was).

I have to find an SD card for the raspberry, put the system online and generate simple script with ChatGPT. Script will run on system boot and it will periodically scan ports to find a device and start measurements.

The goal would be to keep the phone as much as possible inside the raspberry charging station - eventually total screen time will decrease as I move more time into quality time and remove my phone from quality time slots. I think I should note down weekly the screen time from Apple anyway, just to check if total amount of time is decreasing. It is not necessary, but interesting to see. raspberry can live on battery too. I’ll start simple, I don’t have the battery.

I think this is the closest analogy to “GPS location in the gym“ or scanning QR code, but for the screen time. Also, I can put some rules in the code to ignore obvious cheats: plugging phone at night or plugging it for <10 minutes. I think I will report everything for the first week and then fine tune code and commitment. For reference, I spend about 20h a week with my phone. I think I should be at 10h.

Sounds good? I’ll share some code and photos once it’s done.

[poisson]

This sounds great, good luck with it!

[skorytnicki]

I discarded some other options, for example reporting time manually each week or each day.

Another discarded idea, that’s free to implement: you can achieve the same result with just Shortcuts app on iOS. One tap would mark a start date, second would be an end date. On tapping end, a request would be made to some simple service to calculate the minutes between them and sent to beeminder. This has some downsides; the physical element is removed and it requires you to interact with a phone to get your score. It has to work the other way. I can’t be in a state of “I can’t wait to interact with a phone to increase my score on Beeminder”.

@adamwolf

Most folks here are motivated by brrr number go up (and realize it!)

As I mentioned in the first post, the same usb-checking script can run on your laptop - but then you have to have your laptop with even bigger screen turned on. That’s what makes Raspberry a good option.

Nonetheless, don’t feel discouraged by technical details as I’m going to share numbers (baseline + result) as well.

[skorytnicki]

#!/usr/bin/env python3
import subprocess
import time
import requests
from datetime import datetime

LOGFILE = "/var/log/phone_time.log"
STATEFILE = "/tmp/phone_connected_since"
BEE_USERNAME = "YOUR_USERNAME"
BEE_GOAL = "YOUR_GOAL_SLUG"
BEE_TOKEN = "YOUR_API_KEY"

CHECK_INTERVAL = 60  # seconds

def log(msg):
    with open(LOGFILE, "a") as f:
        f.write(f"{datetime.now():%Y-%m-%d %H:%M:%S} - {msg}\n")

def is_phone_connected():
    """Detect Apple or Android phone connected via USB."""
    try:
        lsusb = subprocess.check_output(["lsusb"], text=True)
        # Apple devices: idVendor 05ac
        # Common Android: 18d1 (Google), 2e17 (Samsung), 12d1 (Huawei)
        return any(v in lsusb for v in ["05ac", "18d1", "2e17", "12d1"])
    except subprocess.CalledProcessError:
        return False

def send_to_beeminder(minutes):
    url = f"https://www.beeminder.com/api/v1/users/{BEE_USERNAME}/goals/{BEE_GOAL}/datapoints.json"
    comment = f"{minutes} minutes phone connected (ended at {datetime.now():%H:%M})"
    data = {"auth_token": BEE_TOKEN, "value": minutes, "comment": comment}
    try:
        r = requests.post(url, data=data, timeout=10)
        if r.status_code == 200:
            log(f"Beeminder update OK: {r.text.strip()}")
        else:
            log(f"Beeminder update FAILED ({r.status_code}): {r.text.strip()}")
    except Exception as e:
        log(f"Beeminder update ERROR: {e}")

def main():
    connected = is_phone_connected()
    if connected:
        log("Initial state: phone connected.")
    else:
        log("Initial state: no phone connected.")

    start_time = None

    while True:
        new_state = is_phone_connected()

        if new_state and not connected:
            start_time = time.time()
            log("Phone connected.")
        elif not new_state and connected:
            if start_time:
                diff_minutes = int((time.time() - start_time) / 60)
                if diff_minutes > 0:
                    log(f"Phone disconnected after {diff_minutes} minutes.")
                    send_to_beeminder(diff_minutes)
                else:
                    log("Phone disconnected after <1 minute (ignored).")
            start_time = None

        connected = new_state
        time.sleep(CHECK_INTERVAL)

if __name__ == "__main__":
    log("=== phone_tracker started ===")
    main()

with the following config:

sudo chmod +x /usr/local/bin/phone_tracker.py

# /etc/systemd/system/phone-tracker.service
[Unit]
Description=Phone USB Tracker for Beeminder
After=network-online.target

[Service]
ExecStart=/usr/bin/python3 /usr/local/bin/phone_tracker.py
Restart=always

[Install]
WantedBy=multi-user.target

# and then run:
sudo systemctl daemon-reload
sudo systemctl enable --now phone-tracker.service

And finally you’ll be able to inspect it:

sudo tail -n 50 /var/log/phone_time.log

[skorytnicki]

Some notes regarding future steps.

This code successfully puts a number on beeminder. It has many drawbacks. Doesn’t work offline, doesn’t work outside of the house (I won’t carry a raspberry with me etc). I think I will productise it a bit later, once I confirm it works.

We can imagine a portable device just a bit bigger than your phone, that can fit it and can be easily carried around, sort of pocket or small wooden box.

Winter is coming, I will spend a lot of time in the house, so I just want to anticipate some problems you might want to solve; I will check in a few months. It’s enough for me.

I will still log time spent weekly, just to understand if it works. Also, I will try to understand what are the leeches and what’s the good phone use. If it makes me happier, I wouldn’t mind to spend even more time in front of the screen.

I set the rate to 50 mins a day, effectively it should be in the afternoons during week days, so 1h from Monday to Friday works for me.

[aad]

I’m following with interest, thanks for sharing your progress!

I had to think of a box like Box breathing box - macwright.com as I read this, at least for the home use cases figured I’d drop a link for inspiration.

[skorytnicki]

Thank you.

On box breathing box - I appreciate Tom for taking time to prepare quality images on dark background that corresponds with his website’s. Side note: I would really like to have this sixth sense of a librarian for my own miniprojects: Projects - macwright.com

I ordered some books from the local library to kill night time scrolling and replace it with books.

I’ll report in a week with some measurements.

Nothing else comes to my mind right now.

[skorytnicki]

Let’s calibrate some assumptions.

Past weekly usage with projected future from GPT.

Out of ~26h each week, 15h is dedicated to Chrome and an hour or two to YouTube. Another 2h is focused tasks (Anki, Duolingo, Beeminder) and another 2h is chores or work (text messages, maps, spotify). Then add another 5 on random noise (banking app, more YouTube or checking emails).

I checked the data, and actually with 3h 30min each day I’m a perfectly typical Millennial (this is how I self-identify).

Overall average: The global average for all screen time is about 6 hours and 40 minutes, with a significant portion on phones.

American average: In the U.S., the average is approximately 5 hours and 16 minutes per day.

Gen Z: This group spends the most time, averaging around 7 hours and 18 minutes daily.

Millennials: They spend an average of about 3.4 hours (205 minutes) per day

I think we can realistically get to 15h, that would be great. I think that would mean that I don’t use my phone for entertainment too much. At 10h, I’d be a monk that is using phone just for work or self-improvement.

[philip]

@skorytnicki

On tapping end, a request would be made to some simple service to calculate the minutes between them and sent to beeminder.

I’m pretty sure that there’s an aggregation method that calculates this for you automatically, by using pairs of datapoints, which can indeed have time-like values such as ‘6:49’.

Found it:

clocky Sum of differences of pairs
(so-called because you can use this method to simulate
a timer, by entering start time & finish time as consecutive datapoints)

[skorytnicki]

Oh, this is great indeed, because it eliminates the man in the middle in my design and keeps everything on my device + beeminder.

This, plus auto-trim safety buffer is good value proposition for a premium plan.

So far this week my screen time is 16h, there’s 1 day left this week, so I’ll be 8-10 hours less in front of the small screen. The most important change is that indeed I managed to reduce the screen time when it mattered to me (after work, but before the evening routine). I finished my first book this year and I’m half way through the other.

My goal total is 471 minutes, that makes it almost 8h - it means that 80% of the screen time reduction can be attributed to the time I have the device plugged in. This is surprising, but it might be incorrect. One day I left it plugged in the morning, there’s some noise from initial testing, nonetheless - it certainly contributed to at least 50% of the reduction. However, looking at the data, I think it might be actually 80%, because look, I plugged it more or less right after work and before evening routine of my kids:

Next week, I have to fight screen time when I wake up and immediately before I fall asleep. This is the time when the willpower muscle is the weakest. I think it starts in the evening.

I have to drop the phone far from the bed before I go to sleep. I’ll just plug it to the device, but I think it should log at most 2h (2h 30?) with a comment. This will reduce time before bed.

Let’s start with evenings to fight sleep disruption. To be honest my sleep is not disrupted and I wake up well rested, but maybe it will get 15% better? And I’ll live 2 years longer, like Bryan Johnson? I don’t know.

Important notes:

• just keep phone away
• understand your time. Identify time slots and fight them one by one (before sleep, after wake up, after work - each requires a bit different approach imo)
• take care about environment - prepare books, prepare morning/evening routine to support your goal.

Later next week, I’ll put it all on the visualisation - past weeks, forecast, time spent, total time plugged in etc.

[skorytnicki]

In the logs, I see that it failed to send about 130 minutes in total, probably due to internet connection. This is even more interesting - phone was plugged in longer than the total decrease in time. I think it means that it isn’t 100% effective - not because other things are effective, but rather it doesn’t cover all the cases (screen time before sleep etc).

[skorytnicki]

Now you’ve got better picture.
Total time, result time (total minus useful apps but Chrome), forecast (my imaginary progress goals; 25 => 20 => 15 => 15 => 15) and time logged to beeminder. Yesterday I logged 4h (two very long blocks with max 2h).

The policy to keep phone in the box at night will skew time in the box results even further next week. Maybe I’ll cap the max block duration at 60 minutes instead of 120? I don’t know.

[skorytnicki]

Week 2 will probably go fine, but I expect week 3 and beyond (under 15h) to be difficult. I see small issues:

• I added phone pick ups as a secondary metric to note down every week. It might be even better metric to track, as an answer to the question “do I use my phone like an addict?“. Note it down once a week. Very beemindery thing to track.
• Total screen time across devices should also be tracked. I’m not doing this, too much noise from work. I noticed I am sometimes replacing a phone with a laptop - which is still better, but not ideal.

[poisson]

Personally, I don’t worry at all about the laptop usage. I guess that does depend on the way you use it, but I guess in principle I think a lot of the things I want to spend my time on are still on the computer. But I can see why for some people it might be different.

I agree strongly that “phone pick ups” is a good metric.

[skorytnicki]

I noticed that datapoints stopped registering and it looks like raspberry couldn’t connect to the internet. I can recover datapoints from the log, but I decided to count time spent charging the phone instead with iOS shortcuts.

Dealing with external device is more mess than I expected.

I have two automations: on plug in, store data in a file, on plug out, get data and compute diff in minutes, add result to beeminder. Let’s see.

[skorytnicki]

Week 2 Summary

I easily could’ve made it <15, but I was lazy during the weekend. Total time, minus useful applications was 14h 30 minutes. As you can see, total time decreased by 3h, but total phone pick ups didn’t follow exactly that amount.

For the next 3 weeks we’ll keep tracking to make sure I stay under 15h.

I can’t tell you what changed in my life. I started to read books. I don’t see noticeable sleep quality increase or reduced anxiety or any other magical detox - I didn’t expect that anyway.

Personally, I don’t worry at all about the laptop usage. I guess that does depend on the way you use it, but I guess in principle I think a lot of the things I want to spend my time on are still on the computer

Correct. Same for me.

Maybe we should aim at <10h - that would mean probably no short video content or other social media slop?

[skorytnicki]

@skorytnicki

on plug in, store data in a file, on plug out, get data and compute diff in minutes, add result to beeminder

Store if it’s bigger than 30 minutes and max of 120 - avoid problems with charging all night.

[skorytnicki]

Week 3 summary

Slightly missed the target in total time, but total time never was the goal. It includes almost all the time I drive my car, because CarPlay counts as Spotify / Google Maps usage. I don’t drive a lot and I try to avoid CarPlay now.

The biggest offender was lazy Saturday with 4h. It destroyed my beautiful chart. It could’ve been 13h total and 11h after deductions:

@skorytnicki

I think we can realistically get to 15h, that would be great. I think that would mean that I don’t use my phone for entertainment too much. At 10h, I’d be a monk that is using phone just for work or self-improvement.

I still do a lot of scrolling at 15h, but the gist was correct - we’re entering diminishing returns sphere.

I think there has to be stricter policy to drop phone in the box let’s say at 18:00 and get it back in the morning. I’ll start using smartwatch again to not miss important phone calls.

I’m setting 90 minutes daily goal (single session can be max 120 minutes) and ratcheting safe days:

I think I mostly achieved my goal, now I have to stick to it and build long lasting habit. I need to actively track it for at least 1 month from now. I would like to decrease phone pick ups. I don’t know how Apple counts it, so the goal remains to be 15h / week screen time.

Remember: watching short form video content or scrolling Twitter has absolutely no value. It’s better to stare at the wall than do it.

[skorytnicki]

Try to frame the “iPhone moment” as the event that centralised all your devices into one and made useless content available literally at your fingertips. What should happen now for me is to unbundle the iPhone again.

Following the “time slots” logic, the biggest issue now is mornings. I spend about 20-30 minutes after waking up on my phone. That’s now up to 30% of my screen time daily, also it’s useless screen time. There’s natural urge to check if business is in order, if I missed some messages, this is healthy, but it’s too much. I think it starts with the alarm clock. I have to buy myself a physical (digital) clock.

I don’t know much about habit and addiction patterns, but I know that replacement often works fine. I use books for the evenings, but there’s unfortunately no morning time replacement for the phone. At least nothing comes to my mind. Reading a book at 6:30 AM for 20 minutes wouldn’t work for me. There could be a complementary goal like “breakfast before 7:00“? I have to think about it and I’m open to your feedback.

[poisson]

I am unsure what the best thing to do in the morning is, but I just wanted to say that I am really happy about my decision to buy a standalone alarm clock, it definitely helps to avoid getting sucked in in the morning.