replace the user, slug, and auth_token text as necessary
Create an automation in shortcuts that will run fairly often and set it to run the above shortcut
This could be a time scheduled run or linked to opening an app you use semi-frequently (beeminder!, safari, reddit, etc)
Create your ‘do-less’ beeminder goal
Change to timey goal
mark as api to prevent cheating
disable PPR
change to custom goal
change aggday to ‘last’ - you’ll have a ton of data coming in depending on how you automated it.
uncheck plot all - again, ton of data
TADA🎉
Jury is still out on how well this will work long term. Shortcuts can be finicky and apple is persnickity about exporting usage data like this (understandably). I still haven’t found a way to export per-app/per-site/per-content data either, which is my next goal. Might need to do some sort of break and inspect proxy/vpn for that. Anywho hope you enjoy : )
(I tried this to check vision capacities of an LLM we’re hosting (at my job). It works pretty well if you ask the model to respond with structured output.)
I considered that, but my plan for all of my goals is to have them more or less automated. Going from screenshot>save>upload to $LLM>chat>copy to Beeminder is just too much. You could definitely automate pretty much every bit of that except the screenshot>share part out though. Hallucination could be a problem but that’s not nearly as bad with simple text OCR like this in my experience.
The biggest blocker is that there’s no way to tell good!reddit from bad!reddit with just a single duration. Balancing the useful parts of Reddit against the tendency to scroll is hard, and frankly I’m failing horribly at it.
Your approach is correct and it makes total sense.
The biggest blocker is that there’s no way to tell good!reddit from bad!reddit
The interesting question is why do we want to reduce screen time, what’s the goal here and what will the side effects be. In my opinion, it’s just important to decrease the usage when it is a parasite on quality time slots (for focused work, for family). I put the longer explanation and solution idea in the post below.
But I’m in a weird state. The shortcut workflow works and it’s creating the datapoint on Beeminder. But for whatever reason, Jomo shortcut is outputting a completely different wrong screen time that is showing in their own app or Apple official app… Opal also returns a wrong number. Did you face this issue too?
I also already checked “Share screentime across devices” is not enabled.
I checked the shortcut too. Initially, I got an error. But now it works fine.
Things I adjusted:
for some languages, time is separated with a comma instead of a dot (Dutch for example)
make sure you just have one app connected to screentime settings
reboot the device and then refresh both normal and estimated screentime (dots right top - more)
I’m still checking if it excludes the right apps, so that I won’t accidentally max out because I practiced choir - but so far it seems to be working.
Edit: just noticed what might be the issue. Apple tracks websites too. And by that I don’t just mean the obvious browsing, but like Duolingo ads (from some google page) or visit an url in an app too. I just added the whole “other” category to excluded apps, to make sure the numbers make sense.
There’s a setting in jomo that determines what apps it tracks for total Screentime. Settings>screentime>estimated Screentime>improve estimate>select all apps