Hacker News Reader

Didn't Tumblr do a whole "alienate your user base" thing like Reddit did recently? They got rid of all the racey hot content and basically made it super boring no?

This always seems like a really bad idea.

They did, but that was a while ago. Things are still pretty bad -- if your blog gets flagged for racy content, you lose the ability to have a custom avatar, for example. (Even more, and you get banned.) They've also had quite transphobic moderation policies (really strange considering their userbase!) -- normal, non-sexual photos of trans people are "nsfw" now, which'll get your account flagged or banned. They've been making the app actively worse, redesigned the website to look like twitter, and attempted to cram "Tumblr live" down everyone's throats (only having a "hide for one week" button). Generally doing things nobody asked for, nobody wants, and continuing to break existing functionality while not listening to their users. This is an own-goal.

Tumblr users are rabidly against all the user-hostile stuff other large free sites do. I think they probably could have put the site behind a $1/mo paywall two years ago and people would have gone for it, if they changed nothing else. There's a general "we're fucked" atmosphere on the site because we all know they're trying to wring money out of their users and the enshittification has begun. It doesn't seem like anyone in charge understands their users.

> normal, non-sexual photos of trans people are "nsfw" now

FWIW, the algorithms that big tech companies use are often trained to recognize women’s underwear. So non-sexual photos of trans men wearing women’s underwear are often flagged, even when similar photos of women wearing boxers are not.

I’m not saying it’s right, just giving the actual reason this happens. (Albeit I don’t have Tumblr-specific knowledge.)

Cool. Now why does it consistently happen for images where no underwear is visible, and why doesn't it happen for cis women? Why are appeals not granted? And why didn't staff even acknowledge there's a problem?

No idea. I’ve never worked at any of these companies, I just learned this in an academic conference presentation on ethnographic research into OnlyFans users. OnlyFans doesn’t have its on recommendation system, so users need to find their customers on instagram, Twitter, tumblr, etc.

Hi, just to sate my own curiosity: can you please provide me a link to this presentation, if it's available online? Thanks...

Not online, sorry!

> OnlyFans doesn’t have its on recommendation system

This is by far the most mind-blowing thing about OnlyFans to me. I’m a very open minded person and I would have been so sure that a site like OnlyFans would absolutely require discovery. Like, so sure that if I was making it, I wouldn’t even question including it. The fact they just offloaded this to every other site, and then took all the monetization is amazing to me. Twitter, Reddit and Instagram can barely get money from these users, but they’re super popular, and they just link to their OnlyFans on their profile.

I seriously wonder what other opportunities like that are out in the wild.

They also centralize all the risk. Operating a “pay me $X/month” service is really complicated legally, ethically, etc.

In a way, OF is valuable to platforms like Twitter and Instagram, because it has monopolized a lot of the riskiest cash flow that those companies would prefer to avoid / explicitly disallow in their ToS for a “subscribe to this influencer” service.

If OF had their own discovery platform, they’d have to deal with the thorny editorial problem of ranking. In the current setup, they can totally sidestep it. It would probably cost them more to operate discovery than it’s worth.

How sensitive/stupid are these? Can the behavior be exploited to force innocent images get flagged as explicit? Asking for a friend… also would make an excellent shirt design if properly tweaked

Yes, there are some examples I’ve seen in the past year of images specifically generated to make image detection think it’s X when it’s really just a mess of shapes and pixels

thats fucking hilarious.

it surprises me how instagram and twitter can get away with it but tumblr cant

> it surprises me how instagram and twitter can get away with it but tumblr cant

It's a huge problem on Instagram too.

Twitter is different because it doesn't prohibit sexual content in the first place, unlike Facebook and Tumblr.

How is it an own goal? They were bleeding money, so tried to find a way to succeed through growth of users or revenue.

The path where they just left everything untouched is a path to certain failure.

My guess would be that they wouldn’t have to be bleeding money if they weren’t aiming for growth.

that was a long time ago, under a previous owner. Automattic has seemingly tried to be a good steward of the tumblr brand, but it seems like at this point, the only people left using tumblr are unmonetizable.

the bad idea was buying tumblr. but i think automattic knew that when they bought it, and just thought it would be fun to set some money on fire for a while.

They did that because a lot of the people creating and posting that Content were kids…

How did you quantify that?

Reddit-fuck-up is not even remotely comparable to Tumblr-fuck-up. What Tumblr did was more like if Reddit banned edgy memes.

> Didn't Tumblr do a whole "alienate your user base" thing like Reddit did recently?

No, they did that years ago (getting on for literally 10 years ago), and that's what ultimately resulted in selling it to its current owners for like 1/1000th of what Yahoo paid for it.

This isn't true at all, Blaze was a huge success, scrolling through Tumblr you see people with silly numbers of checks, lots of people pay to remove ads, it's just that that the successes aren't massively revenue generating in the way that ads are.

Trying to monetize user generated content is hard because both the users and creators believe that it's their content (which it is that's how community works) and so locking it away and taking a non-trivial cut is going to be met with insane vitriol and there's not much value to add over the base experience. Tumblr's patron thing you should of think like Hashicorp re-licensing Terraform.

And worse is that sometimes creating true value-adds naturally perverts incentives. Like for example a true value-add would be "bubble-as-a-service" because Tumblr has some really nasty subcultures which is why posters don't like when posts "break containment." It would be the easiest money ever to just pay and get a huge blocklist without having to go through the effort of making one yourself but then you kinda need those nasty communities to justify the value.

From the CEO’s blog today:

> You mention Blaze and Ad-free doing well, but their adoption is so small relative to the use of Tumblr their revenue couldn't support a fraction of the ~1,000 servers it takes to run Tumblr, much less any salaries.

I think we're saying the same thing, they're successful in the sense that Tumblr users like them but they're small potatoes. Tumblr users aren't necessarily allergic to monetization but the ones they've found don't pay the bills.

(person who doesn't really know what they're talking about when it comes to services of this scale asking) Why does Tumblr need 1000 servers to run? You can serve a lot of text and relatively well compressed image files with not a lot of hardware.

A lot of people use it to promote their art or to share screencaps/gifs/clips of movies, so I think people would get pretty cranky if the image files were overly compressed (unless it's the magic kind of compression that doesn't cost any fidelity. I don't really understand how compression works).

That's without getting into video content.

It's a social network. Tumblr serves a lot of traffic, and has a lot of data. The product data alone is over 1 trillion rows in relational databases by my estimation (former early eng there), and that's not accounting for replication / high availability.

1000 servers actually sounds surprisingly small. It was a fair bit higher than that a decade ago, and that was on bare metal then, vs (afaik) cloud VMs now. Then again, their traffic is a lot lower now, but still nothing to sneeze at.

For the YC audience: Tumblr has about 60B images using 8 petabytes of storage. 205TB of memory, 46,752 CPU threads, 250 kW of energy. About 100k requests a second, a petabyte a day of bandwidth.

So in short, the answer to my question is that the scale is way way larger than I imagined.

They don't seem to have tried the one thing that might have worked for them, namely, embracing the fact that they're first and foremost a porn site. Instead, they decided to be actively hostile to this reality, with predictable results.

Different owner

yes, and even though the newish owner has not banned all things naughty that would get flagged at fbk or similar, they still have banned and removed profiles that had erotic content.

It's fine to do what they want, but it's become a slightly naughtier pg-13 - I guess nc-17. So those who want naughtier are going elsewhere, and no one knows for sure what is okay and what could be okay today and get your account removed tomorrow - so why put any content into it?

The lesson:

"Don't remove porn when the userbase is left-leaning and embraces it."

Reddit and Twitter do it the right way: hiding adult-rated content behind the curtains.

The lesson is that user generated content shouldn't be turned into an algorithmic amusement park for advertisers and data brokers.

Tumblr wasn't making money with or without porn. It was just shifting between owners who thought they'd be better at squeezing money out of it.

Aren’t left leaning people all about “ethical porn” and power dynamics now?

It looks like all the left-leaning porn posters have moved to BlueSky (to the point the main feed was unusable, last I checked).

Things are pretty grim when you're asking your own staff to purchase merch because "every little bit helps".

This was live-streamed publicly and advertised to users. I didn’t make that very clear.

> the monetization strategies they’ve tried ... crabs

Like Bored Crabs NFTs or restaurant supplier or health insurance or...?

A virtual pet that lives on your blog and occasionally tries to eat your mouse cursor.

There is very simple monetization strategy: ads, cut bloat, move content moderation staff to India!

Tumblr has an army of content creator that work for free. It is a site that host text and images. How they even spend 30 million per year? Maybe on organizing "stuff convention", SF real estate and so on... Cut the bloat!

They don't spend 30 million per year, they spend 30 million over what they generate.

I don't think they'd generate peanuts so as to not even cover hardware costs.

> they spend 30 million over what they generate

That is my point. I do not understand how they total costs can be over 30M. It is mostly static content that begs for CDN cache. It is also not mission critical.

Probably they do a lot of content moderation, which takes human beings at least at some stage.