windwerfer

[windwerfer @windwerfer]

diggus maximus (my research buddy):
1) searched for EEG datasets (with permissive licenses) that are clearly labeled and can be used to train EEG foundation models to recognize those labels (eg: drowsy/hypnagogic)
2) he then trained / fine tune the REVE EEG foundation model on kaggel (a platform that allows people without gpu access to train ai - shout out to google for this possibility!) - whoever ever used a jupiter notebook will love grok bot doing the debugging for him!!
3) evaluated the results, and told me which labels actually will work with what foundation model goal is to develop a eeg feedback app, that can use ai to detect certain mind stages (meditation related ones).

the hard part is to search for labeled datasets, evaluate if they are valid for the device i want to support (electrode montage should match / able to convert) and then check if the result is better than pure chance. EEG feedback has a lot of limitations, and evaluating the result is more important than producing one.

grok bot did that in tandem with me, but it was more me slowing him down and wanting to understand what he does. amazing to just point it at something and see results coming in..

https://t.co/v082n71cl8

(I dont live in the US)

[Rupert Lowe MP @RupertLowe10]

The Rape Gang Inquiry Report.

https://t.co/EuKgGWBRhS

[windwerfer @windwerfer]

I have the premium sub, in the new app I can only select from the last 20 bookmark categories (I have over 30, so I can not access those when I save a bookmark)..

and I think in the new version, the videos that started playing don't stop downloading when I scroll further down. that makes new videos super slow to start (I have 0.5mbyte connection speed). probably adds loads of unnecessary bandwidth and makes tracking when people scrolled past a video harder..

otherwise, good new design, grok Integration rocks! super quick compared to the old one.

ps: on tablet/phone (tablet more annoying) in landscape mode, the images scale to 100% width. that's horrible for scrolling, because 1 portrait image takes 3-4 screen heights. meaning it takes a lot of time to scroll past one image+it's impossible to see the whole image at once (only 1/3 at a time). simply scaling it to 100%height completely fixes that -dead simple in the code :-)
#NewXAndroidFeedback

[james hawkins @james406]

110-year-old Turkish grandma shares her secret to a long life:

"i never once used Microsoft Teams"

[Elon Musk @elonmusk]

Good explanation of nihilist philosophy

[windwerfer @windwerfer]

Interface is super fluid and way quicker on my slower internet connection. I use the new interface on my phone (portrait), but on my tablet which is landscape (10" Lenovo, 2500x1600pixel), it's impossible for any posts containing video or images. They are just waaay too wide. If in landscape and the image is portrait, its twice the screen width just to scroll past 1 image.. otherwise: way better, especially that clicking on the grok button loads so much faster. As soon as the landscape issue is fixed, I switch :-)

The old x app has a good ratio. New design feels like getting beaten by the overwhelming image/video size. :-)

[windwerfer @windwerfer]

@AlanN69375 @andst7 @elonmusk @gualtierieurope @grok is it true that lithium is *this* damaging to the environment / live as the image suggest?

[Alex Phillips @ThatAlexWoman]

The ENTIRE WORLD needs to know what has happened in Britain

They need to know how politicians covered it up

They need to know how police covered it up

They need to know it is STILL happening

For the sake of those girls, and girls still being abused now, I plead PLEASE share this

Share with your Presindents, Prime Ministers, the UN

They are torturing and raping our girls.

Something has to happen

Please help

Because our Government is refusing to

🙏

[windwerfer @windwerfer]

@ArthurMacwaters Chatgpt chat not available (any more)

[windwerfer @windwerfer]

@swyx @altryne @ChatGPTapp I noticed that pressing the voice mode button where the prompt is entered, is different to circled 'voice mode' (when starting a new chat). Circled button = singing
(That's on Android)

[👉M-Û-R-Č-H👈 @TheEXECUTlONER_]

This woman is from Iran.

She is wondering how liberals and Democrats can support Iran when women there basically have no rights.

She details the extraordinary abuse against women in Iran.

Yet, the Democrats & Liberals are supporting Iran? How can liberal women be for this?

[Charlie Kirk @charliekirk11]

New research from the University of Texas shows that all the justifications for mutilating children were lies. They analyzed more than 100,000 patients with gender dysphoria, and looked at the outcomes for those who received surgery vs. those who didn't.

Depression, anxiety, suicidal feelings and drug abuse were all "significantly higher" for those who received surgery compared to those who didn't.

Let's repeat that: Treating gender dysphoria with sex change surgery was linked with drastically WORSE mental health outcomes.

Phony doctors belong in prison for what they did to our children.

https://t.co/n2ExvtWlES

[windwerfer @windwerfer]

@elonmusk Is it only me or is there no android version yet?

[windwerfer @windwerfer]

Community Polls (regional)

There is a strange kind of unreliability to statistical data. Sometimes the sample size, sometimes the organisation that does the poll, sometimes there is just no data available. X has the reach most poll companies can only dream of.

How about a community poll request option to leverage the awesome cross section of opinions and views, spread of people and speed that this platform offers?

It could truly reflect the broad opinion of a large sample size. "Your help would be appreciated. Please share your view with your fellow humans"

I think there should be :
-a maximum of 1 every view days
-The people that are polled should be chosen randomly (smaller bot likelyness)
-"never bother me again with polls" opt out option

I think I would trust recent poll data from x quite a bit. It could even develop into a "polls over time" feature.

I also think that there should be automatic annotations about the probable validity (eg. based on sample size)