home tags events about login
one honk maybe more

benjojo posted 23 Nov 2025 15:06 +0000

Interesting, it seems that Discord no longer includes the entire history of their gender guessing model output, here is my most recent "discord gender assignment" (75% male)

{
  "user_id": "282657081457115136",
  "day_pt": "2025-07-08T00:00:00",
  "gender": "male",
  "predicted_gender": "male",
  "probability": 0.75259220600128174,
  "prob_male": 0.75259220600128174,
  "prob_female": 0.15768958628177643,
  "prob_non_binary_gender_expansive": 0.089718155562877655,
  "prediction_type": "fixed",
  "model_version": "2025-06-29T00:00:00.000000Z"
}

Given that last time they put this number at 80%+ Male, I don't quite know what to think of this (other than that this is a remarkably stupid endeavor)

On the flip side, the age vibe/guess has become more modest:

{
  "user_id": "282657081457115136",
  "day_pt": "2025-07-08T00:00:00",
  "age": "18-24",
  "predicted_age": "18-24",
  "probability": 0.48142609000205994,
  "prob_13_17": 0.0231526680290699,
  "prob_18_24": 0.48142609000205994,
  "prob_25_34": 0.33853498101234436,
  "prob_35_over": 0.15688623487949371,
  "prediction_type": "fixed",
  "model_version": "2025-06-29T00:00:00.000000Z"
}

18-24? Aww thank you soul-less model, you flatter me

You can find your own numbers if you do a GPDR/Privacy/Data export (that will take ~12 hours hours in my experiance) and it should be in the (surprisingly massive) file Activity/analytics/events-2025-00000-of-00001.json, You will likely need to just grep for "predicted_gender"

benjojo replied 23 Nov 2025 15:16 +0000
in reply to: https://benjojo.co.uk/u/benjojo/h/SRy3jxdRc5rL15w3BX

I think discord must be the most invasive piece of software that I run on a daily basis.

Although, I guess I haven't look into what firefox/thunderbird/steam/vscode is reporting

The sheer volume of data in bytes that comes out of the privacy export is staggering, rivaling quite close to my mbox inbox file in bytes alone.

Most of the stuff in the files is full of ancient data points, so ancient in fact that I can currently track my last 4 major OS upgrades on my desktop, and all 200+ Linux kernel versions I've run

benjojo replied 23 Nov 2025 15:21 +0000
in reply to: https://benjojo.co.uk/u/benjojo/h/ny723kq8gl8SXzD3R5

I suppose I actually don't need to be running the desktop client anymore.

I used to run it back when I cared about things like push to talk voice, but I didn't really do that anymore (voice calls) she doesn't make sense to be appending to some giant database table constantly with stuff (because there's no obvious gain running the desktop client vs just a browser tab)

also if I move to just using it as a browser tab I won't have to deal with the " must be your lucky day! (to manually download a deb file) because we can't run an apt repository " every week.