Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Gradient.horse

gradient.horse

Post reply on HN
Print view View original

41–50 of 64 posts

Re: Gradient.horse

#41

Post by echelon »

spython wrote: post #33
spython wrote: post #2

Hey, it's my project, nice to see it here! I wanted to do something that I liked in the early times of the web - something small, optimistic and whimsical. The AI-assisted drawing moderation is inspired by drawafish.com and works quite well at keeping the horse parade family friendly!

As a bit of an observation - the horses went viral on tumblr, bluesky and to a lesser extent facebook and instagram and smaller pages at different days, so I could observe how the drawing style and content changes: - Most beautiful horses and least penises were drawn by tumblr users. Maybe because they've drawn a lot of penises before and got it out of their system : ) - Most penises: HN (but I guess it's my fault fo…

Tumblr is still around? That's a good data point.

Re: Gradient.horse

#42

Post by maxbond »

echelon wrote: post #41
spython wrote: post #33

Earlier quoted context omitted.

As a bit of an observation - the horses went viral on tumblr, bluesky and to a lesser extent facebook and instagram and smaller pages at different days, so I could observe how the drawing style and content changes: - Most beautiful horses and least penises were drawn by tumblr users. Maybe because they've drawn a lot of penises before and got it out of their system : ) - Most penises: HN (but I guess it's my fault fo…

Tumblr is still around? That's a good data point.

They're well on their way to becoming the next PDF.

Re: Gradient.horse

#44

Post by Imustaskforhelp »

spython wrote: post #33
spython wrote: post #2

Hey, it's my project, nice to see it here! I wanted to do something that I liked in the early times of the web - something small, optimistic and whimsical. The AI-assisted drawing moderation is inspired by drawafish.com and works quite well at keeping the horse parade family friendly!

As a bit of an observation - the horses went viral on tumblr, bluesky and to a lesser extent facebook and instagram and smaller pages at different days, so I could observe how the drawing style and content changes: - Most beautiful horses and least penises were drawn by tumblr users. Maybe because they've drawn a lot of penises before and got it out of their system : ) - Most penises: HN (but I guess it's my fault fo…

Is there a thriving tumblr tech community, anything similar to HN on tumblr?

Most penises: HN (but I guess it's my fault for saying there is a filter) followed by b3ta.com and twitter

- Most swastikas: twitter and, somehow, HN.

I expected better from HN :/

Someone once told me that HN is like 4chan but for tech people & I defended a lot against it but now I feel like seeing an overlap between HN,twitter,4chan

> - Most unexpected source of traffic: substack

How much traffic did you get out of substack?

Re: Gradient.horse

#45

Post by asplake »

Nord warns: This website may be unsafe We detected that the website you’re about to visit could be a scam.

It was actually the first time I'd seen that message, so, erm, congrats?

Re: Gradient.horse

#46

Post by Tommix11 »

LeoPanthera wrote: post #8

MTBP (Mean Time Before Penis) is about 30 seconds. If you tab out and then back again it draws about 20 horses all on top of each other making some kind of eldritch horror.

I saw a xenomorph, who drew that one?

Re: Gradient.horse

#48

Post by llbbdd »

semitones wrote: post #40
LeoPanthera wrote: post #8

MTBP (Mean Time Before Penis) is about 30 seconds. If you tab out and then back again it draws about 20 horses all on top of each other making some kind of eldritch horror.

I only see my penis for some reason

Me too but I guess I should have been paying more attention to the screen

Re: Gradient.horse

#49

Post by spython »

Imustaskforhelp wrote: post #44
spython wrote: post #33

Earlier quoted context omitted.

As a bit of an observation - the horses went viral on tumblr, bluesky and to a lesser extent facebook and instagram and smaller pages at different days, so I could observe how the drawing style and content changes: - Most beautiful horses and least penises were drawn by tumblr users. Maybe because they've drawn a lot of penises before and got it out of their system : ) - Most penises: HN (but I guess it's my fault fo…

Is there a thriving tumblr tech community, anything similar to HN on tumblr? Most penises: HN (but I guess it's my fault for saying there is a filter) followed by b3ta.com and twitter - Most swastikas: twitter and, somehow, HN. I expected better from HN :/ Someone once told me that HN is like 4chan but for tech people & I defended a lot against it but now I feel like seeing an overlap between HN,twitter,4chan > - Mos…

Yeah, I too had probably an idealized image of HN - but I guess what makes the HN great is the moderation :)

And it's not like there are no beautiful horses being drawn - I observed quite a few ingenious horses and non-horses coming in in the last hours. But I'm surprised at the level of.. carelessness.

One way of dealing with it would be training a penis-drawing-recognition ai and only show other penis drawings to people who draw dicks. Though it would go against the whole idea of, well, having a moment together.

Substack traffic is hard to quantify, at least for me as I am only looking at HTTP referrers, and I had the feeling the referrer field was mostly filtered out. It is comparable to facebook and mastodon traffic, both in volume and in the sense that it is a slow burn - the link gets share in a small community, and then gets shared a few days later in a different one. In contrast to tumblr, HN and bluesky, where a spike in traffic comes sudden, a cultural moment is shared by many thousands of people, and then the attention mostly disappears the next day.

Re: Gradient.horse

#50

Post by dmit »

gus_massa wrote: post #3

I expected it to detect the legs automaicaly, I guess I'm spoiled by AI. In a similar project, a hard part was filtering penis and nazi symbols. Protip: add a few legs parts in the tail, so it moves too.

> In a similar project, a hard part was filtering penis and nazi symbols.

Check out how the precursor of this project did it! https://news.ycombinator.com/item?id=44719222

> I built a basic CNN trained against penises and swastikas, and then anything that doesn't hit the 63% confidence score gets sent to a mod queue, a vibe-coded fish-tinder (https://bsky.app/profile/bigass.bsky.social/post/3luvikxn3f2...).

Post reply on HN