foxyoreos (they/it) (🔞)

I draw things! Very, very NSFW. I use content warnings on my posts (and try to avoid boosting nsfw posts without them).

Expect furry smut including vore (the full range from endo/safe to cruel/fatal), noncon, limited scat, and general lewdness. Some programming too, but there is gonna be a lot of smut.

If you have trouble contacting me or want a more SFW account to interact with, see https://mastodon.social/@foxyoreos

#furry #vore #nsfw #artist #furryart

Normy account (contact here if your server is muted by gulp.cafe): https://mastodon.social/@foxyoreos
Commission Page: https://foxyoreos.netlify.app
Furaffinity: https://www.furaffinity.net/user/foxyoreos
Subscribestar: https://subscribestar.adult/foxyoreos

Public Key

npub16dhuhhnrc6ps0acrnpk6tugrrt6m5z8hlqdh8sxzzd86a09jksns5lnu6d

NIP-05 Address

foxyoreos_at_gulp.cafe@momostr.pink

Profile Code

nprofile1qqsdxm7tme3udqc87upesmd97yp34ad6prmlsxmncrppxnawhjetgfcprpmhxue69uhhyetvv9ujumt0d4hhxarj9ecxjmnt2q3ml0

Publishing to

relay.momostr.pink

Author Public Key

npub16dhuhhnrc6ps0acrnpk6tugrrt6m5z8hlqdh8sxzzd86a09jksns5lnu6d

Show more details

Published at

2026-04-03T00:49:40Z

Kind type

0 Metadata

Event JSON

{ "id": "dd5c7ec0dc8ca6fcbe72bad89f04189ea307b8918258e2161e1d7dad19fd22d4", "pubkey": "d36fcbde63c68307f703986da5f1031af5ba08f7f81b73c0c2134faebcb2b427", "created_at": 1775177380, "kind": 0, "tags": [ [ "t", "furry" ], [ "t", "nsfw" ], [ "t", "artist" ], [ "t", "furryart" ], [ "t", "vore" ], [ "proxy", "https://gulp.cafe/users/foxyoreos", "activitypub" ], [ "L", "pink.momostr" ], [ "l", "pink.momostr.activitypub:https://gulp.cafe/users/foxyoreos", "pink.momostr" ], [ "-" ] ], "content": "{\"name\":\"foxyoreos (they/it) (🔞)\",\"about\":\"I draw things! Very, very NSFW. I use content warnings on my posts (and try to avoid boosting nsfw posts without them).\\n\\nExpect furry smut including vore (the full range from endo/safe to cruel/fatal), noncon, limited scat, and general lewdness. Some programming too, but there is gonna be a lot of smut.\\n\\nIf you have trouble contacting me or want a more SFW account to interact with, see https://mastodon.social/@foxyoreos\\n\\n#furry #vore #nsfw #artist #furryart\\n\\nNormy account (contact here if your server is muted by gulp.cafe): https://mastodon.social/@foxyoreos\\nCommission Page: https://foxyoreos.netlify.app\\nFuraffinity: https://www.furaffinity.net/user/foxyoreos\\nSubscribestar: https://subscribestar.adult/foxyoreos\\n\",\"website\":\"https://gulp.cafe/@foxyoreos\",\"picture\":\"https://cdn.masto.host/gulpcafe/accounts/avatars/112/349/921/630/720/994/original/10973d2a1850e2ed.png\",\"banner\":\"https://cdn.masto.host/gulpcafe/accounts/headers/112/349/921/630/720/994/original/6b4e380445d3be4a.png\",\"nip05\":\"foxyoreos_at_gulp.cafe@momostr.pink\"}", "sig": "331d1ed0c23d129dcb5692cbe5d2cdc1fdbeec514dabee42ead5c48d91e6d825cdd1c0c801704935db6486c83944efa0bb020336115c568917c72719ad8d260b" }

Last Notes

2026-03-16 21:44:25 UTC

- reply

me in the army, after our 3rd loss: "I feel like we should have a different general, why does he keep telling all of us to douse ourselves in gasoline before we go out to fight?"
Dems, immediately: "You're a spy for the enemy, I bet you LOOOVE the gasoline because it gives you an excuse not to fight. And you're gonna try to replace our generals NOW, after half of our troops have been immolated and while we're in the middle of an immolation crisis!? Despicable."

2026-03-16 21:39:56 UTC

- reply

exactly - you don't even need to care about trans critters (not saying to JB doesn't) to recognize that this is a BAD election strategy.
I'm being lectured about pragmatism from the least pragmatic party in the world. The same people who will constantly cede political ground over shit like "but Liz Cheney is a friend" and "but if we don't concede I'll miss my vacation flight."
https://cdn.masto.host/gulpcafe/media_attachments/files/116/241/006/275/222/992/original/de386b8e08bf152c.png

2026-03-16 21:27:21 UTC

- reply

I'm being asked not to vote for an ineffective candidate but to participate in my own oppression because Dems would rather risk fascism than stick up for the marginalized.
They are actively making a choice to roll the dice on this, and then looking at me like, "well, you'd better make sure this works."

2026-03-16 21:25:07 UTC

- reply

I voted for Kamala and went into communities I'm a part of and actively advertised for her. I was posting political comics to my galleries, arguing with users.. I understand how to be pragmatic.
What I'm being asked for isn't pragmatism, it's blind allegiance. I'm being asked to throw critters onto the tracks (to throw myself onto the tracks) not because it's the best way to win an election, but because the Dem party refuses to try to win elections.

2026-03-16 21:11:47 UTC

- reply

I'm not gonna be pragmatic about trolley problems when I am watching the Democratic party actively tie critters that I care about to the tracks. Fuck off!
And they're mad at progressives?
If your neighbor at the grocery store starts getting mad at you because you won't pre-decide which of your children you'd kill to save the other if they kidnapped them, the correct response is not to indulge in that, it is to call the police.

2026-03-16 21:11:07 UTC

- reply

I think what makes me angriest about the trolley problem memes is like.. we're not in a trolley problem right now, and I am watching a lot of democrats get very very angry that they're not being allowed to build one.
"What would you do if we-"
"I'm telling you NOT TO DO THAT"
"But what if we did?"
"Are you going to!? Are you going to do it!?!"
"I don't want to. But.. but what if we did?"

2026-03-16 21:10:06 UTC

- reply

but suppose you had the choice of a bus that could either take you closer or further away from your destination and also periodically the bus driver pulled out a kid and threw them from the window directly under the wheels of the bu NO NO IGNORE THAT PART THOUGH THAT'S NOT PART OF THE ANALOGY KEEP RIDING THE BUS

2025-12-11 22:27:45 UTC

- reply

hot take, but debian/ubuntu in my experience.
"You should use X distro, not what you're on" is a meme, but I feel like it's much more likely under a post for somepony trying out Linux to see at least one "don't mention anything other than Ubuntu, you're overcomplicating this" reply.
I think Ubuntu was the only time I've ever had somepony telling me I was on the "wrong" distro by not using it. In general, when I see nixos, arch, even fedora mentioned it's just "hey these exist."

2025-10-03 21:03:36 UTC

- reply

Having said that.. that was Matrix's ideas with bridges, and afaik they ended up kind of sucking *really* bad.
And that was part of why I started to like Mastodon more - looking at it and saying, "huh, well I thought they were naive, but their shit works."
But I don't know how much of that was because of a fundamental problem and how much of it was.. *gestures at the entire Matrix team.*

2025-10-03 21:02:06 UTC

- reply

yeah, it's been startling to me to figure out how many problems in areas like this get much, much easier to solve when you just solve them in the client and stop thinking about who owns the servers.

2025-10-03 21:00:29 UTC

- reply

I've complained about how RSS is insufficient for artists.
You *could* build a separate protocol that supports commenting and favorites and all of this stuff. I dunno, maybe something exists.
What I will likely try to eventually do is build an inverse of Postybirb that just directly pulls from multiple websites. Without some underlying protocol or special format, just by like.. connecting to a hard-coded list.

2025-10-03 20:59:24 UTC

- reply

Mastodon solves a lot of these problems a lot better not because of high-minded approaches to decentralization, but because there was also a contingent of users here who just built some boring stuff that works.
That irons over a lot of what would otherwise be horrible issues.
So lately my philosophy on most of this stuff has started to evolve, and in instances where we're building tech, I'm looking more at boring solutions.

2025-10-03 20:57:08 UTC

- reply

This is a really imperfect, kind of hacky solution that only kind of solves the problem, but is also like 300x more practical and easier to work with than most of the decentralization discourse and theories that I see online.
So like, why is Mastodon working better on this? It's not that federation is bad, I like being able to talk to you over on Cubhub. But that's just like a nice feature. The safety net is multi-account clients.

2025-10-03 20:55:52 UTC

- reply

Like I'm also thinking about this from Mastodon's perspective too and realizing that like.. why is Mastodon isolation less of a problem with me, it's because the clients allow multiple accounts that can be on separate servers.
I'm muted from dot social for.. I honestly have no idea, I put a hell of a lot of work into labeling my stuff.
But it's less of an issue for me because I also just.. have a dot social account.

2025-10-03 20:54:37 UTC

- reply

no, definitely not. But like.. the solution being offered by Bluesky doesn't solve the problem of Bluesky having a restrictive ToS.
I actually have theories about how to use tech to make that problem better for users. They don't involve a protocol.
Bluesky's approach of decentralizing the website instead of user presence is increasingly I think just.. straight up incorrect.

2025-10-03 20:51:06 UTC

- reply

This is something I've been kind of hinting at and thinking about for a while, and it's why I've done more with locally forking Postybirb than I have done with anything else.
But the Bluesky drama and the "just move servers" from accounts that have *no* idea what artist needs are or how to balance and handle reach..
Postybirb has the right approach here. It's pragmatic, it's boring, and it fucking works.

2025-10-03 20:48:30 UTC

After all of the fighting about ActivityPub and ATPro, the actual answer to the question of "how do I build a decentralized online presence as an artist" is "use Postybirb."

2025-10-02 20:41:11 UTC

I've mostly kept Bluesky drama off of my Fediverse accounts, but I do want to offer some advice (with context) if you're on Bluesky talking up Mastodon. The response to every piece of criticism from Bluesky devs has been "roll your own."
It's not advice that the users are amiable to right now.
The response from Bluesky devs to criticism of the "roll your own" tools has been "you're holding it wrong" or to downplay challenges. So *definitely* don't do that.

2025-08-20 02:55:04 UTC

- reply

@npub1z9a…k0ya
When someone's entire conception of activism begins and ends on Twitter.
This is why we can't get shit done; people genuinely think posting memes matters more than actual on-the-ground activism.

2025-08-20 02:44:13 UTC

- reply

if after everything that's happened, they run a fucking transphobe..

2025-08-12 21:05:06 UTC

- reply

"See this is the danger of antitrust, now Chrome might be owned by an AI company-"
*Google* is an AI company. Chrome is owned by an AI company right now. An evil one that does evil shit and uses its browser to do it, and is enormously powerful.
Perplexity is not worse than Google. They're both shit, but Perplexity doesn't own the Internet. So.. fuckin great! Give them the browser. What is the problem?
What are you trying to preserve?

2025-08-12 21:01:38 UTC

Perplexity would be dogshit at maintaining a web browser and in many ways that's a plus.
Google is evil and is competent at building a browser (so competent that they suck the air out of every other project), and has massive resources/platform reach to push their browser such that no alternative will *ever* gain traction.
Perplexity is just evil. They don't have the other stuff.
This would be an improvement. Don't look a gift horse in the mouth.

2025-07-05 17:19:26 UTC

- reply

pat pat <3

2025-06-23 22:43:50 UTC

- reply

It'll still ooze personality and polish, and there'll be all these affordances that keep the thing from falling apart and you're having a good time with it, but you're thinking in the back of your head, "something about this is broken."
Sometimes they completely nail it! I mean, you know my opinions on BotW X3
But they're just as likely to take something fundamentally flawed and just workshop over and over again until it just barely manages to hold together.

2025-06-23 22:41:31 UTC

- reply

I've had a couple of Nintendo games from the 3DS and Wii U era that I had this kind of meh reaction to and what my hot take started to turn into is - Nintendo is extremely good at polishing games, to the point where fundamental design flaws don't necessarily ruin the experience.
But they're less good at noticing and addressing fundamental design flaws.
The end result can be a game where when you really dig into it it's just kind of.. bad? But it's still playable.

2025-06-05 06:36:45 UTC

- reply

> This kinda gets to the core of why using phone number authentication is blitheringly stupid though.
This!!!
It's almost like the fact that we're doing wild shit to work around phone number limitations, and the fact that we need to have band-aid fixes that hurt both security *and* ease-of-use is a good reason to *not use them that way*.

2025-06-05 06:28:43 UTC

- reply

I mean, to your point of "not everypony needs account recovery" - I should at least have the option to make my pin non-expiring, right?
If I'm not worried about account recovery, then I should be able to prioritize security, and I think it's a flaw that Signal doesn't even give me that choice.

2025-06-05 06:27:20 UTC

- reply

I will grant that this complicates account recovery. But for all of the compromises Signal makes to UX for the sake of security, I don't understand why in this area, they're suddenly willing to say "we'll just use *expiring passwords*
It's such a huge security tradeoff.
And also a tradeoff they wouldn't need to make.. if they didn't tie identities to phone numbers.

2025-06-05 06:24:49 UTC

- reply

Signal's security model around this is heavily focused on data leaking - the perspective the company has is that since your chats are cleared, what's the problem.
Speaking from an artist perspective, somepony gaining access to my identity could do a heck-ton of damage to me even without access to my data.
It's an area where Signal over-simplifies threat models. I'd bet that's not just me either; I suspect activists care about impersonation risks too.

2025-06-05 06:22:57 UTC

- reply

The conversation will go:
"Hey, did you get a new phone?"
"Yeah"
"Cool"

2025-06-05 06:22:45 UTC

- reply

Most critters I know using Signal use it on one device.
I think it's a huge issue if signing out of Signal for a week opens you to SIM-jacking attacks. I can't think of any other service I use where I have to periodically remind the service to keep making my account secure.
I also have issues with the number change notification. Casual users aren't going to pay attention to those. Experienced users mostly won't either.

2025-06-05 02:01:57 UTC

- reply

there are a number of foot-guns in Signal that work like this. Secure backups require setting a master password with signal. Securing against SIM attacks requires setting another pin.
These are usability/security problems that only exist because Signal doesn't want new accounts to require a password.
But in order for those accounts to protect against data loss and prevent sim jacking, they have to set passwords anyway. All we've done is made the defaults worse.

2025-06-05 01:56:41 UTC

- reply

or better yet, accounts shouldn't be linked to the SIM in the first place. I really don't see the justification for it.
Signal's reasoning has evolved over the years, but their current excuse seems to be "it prevents spam."
But it's very possible to do phone number verification without making a number a primary identifier. And yeah, unlinking it would require remembering a password.
But to prevent sim swapping attacks, you need to remember a pin/password!

2025-06-05 01:55:57 UTC

- reply

> You can add a registration lock that requires a pin that prevents that kind of takeover.
Note that this *expires* if the app is inactive for too long, which is one of the most baffling security decisions I've ever seen.
I think it's reasonable to say that Signal's defaults make it vulnerable to SIM swapping. If you have to teach critters how to secure an account, you're doing something wrong. That pin should be required by default, and should not expire.

2025-06-04 22:29:51 UTC

- reply

if that issue went away, then I would be firmly back in the "Matrix is annoying, but at least it's accessible" camp.

2025-06-04 22:28:39 UTC

- reply

That's reasonable (and again, I'm not shy about my criticism of Signal).
If Matrix fixed this specific bug, I'd recommend it. The encryption issues would still be there, but it's not like any of the chats outside of Signal are encrypted, I'd just use unencrypted Matrix chats.
I went from "Matrix sucks but everything sucks, it's still the best Discord replacement", to "I don't know how I'm supposed to use something I can't trust to deliver messages."

2025-06-04 22:15:37 UTC

- reply

I've never had this problem happen on Mastodon though, so I don't think the issue is federation itself, just that Matrix federation seems much less robust.

2025-06-04 22:14:44 UTC

- reply

I'll pass it on but I don't think that's the issue because I've seen expired encryption warnings on Matrix and they at least come with warnings in my experience.
Also I'm fairly certain that my keys aren't expired and they still can't see my messages.
I've seen this happen with freshly created rooms. I get invited to a freshly made new chat, join, and.. nothing.
I think the problem is federation. I've never had this happen with critters on the same server as me.

2025-06-04 22:12:11 UTC

- reply

Sorry, I didn't mean to put words in your mouth. You know I'm no fan of Signal necessarily, I've ranted about it plenty before (also thank you for calling out sim-jacking as bad security, I hate this gets ignored as an actual security flaw).
More just.. would I recommend Signal over Matrix? I hate to do it, but yes, because Signal delivers messages. You might lose history, but at least I can mitigate that. There's nothing I can do about Matrix reliability.

2025-06-04 22:06:29 UTC

- reply

I just yesterday finished talking to a critter who has just recently realized that none of the chats they were sending me on Matrix had been delivered since early April.
Before that it was realizing that embedded art doesn't display for all homeservers reliably (which is why I'm not using it for Subscribestar feeds).
I want to recommend it, but I can't trust it to deliver messages.

2025-06-04 22:02:34 UTC

- reply

Read receipts would fix the issue on my end. Well, not fix but.. at least I'd know.
But it won't help me for occasions where a critter messages me and the message never comes through.
Even with all its flaws I want to recommend Matrix (and did for a while), but I just can't if it doesn't deliver messages reliably or show errors when it fails. That's the worst failure mode for me, I can tolerate nearly anything else except that.

2025-06-04 22:01:03 UTC

- reply

my major issue is I have on multiple occasions run into Matrix bugs (once with an actual client) where messages either weren't delivered to me or them, and neither of us got any indication - we only discovered it while talking out of band. It's to the point where I explicitly discourage critters contacting me on Matrix from using matrix.org.
A lot I can deal with, I can even turn off encryption. But silent disappearing messages is a killer.

2025-05-07 17:01:43 UTC

- reply

Not dumb, you're a good productive pupper!! <3
Extremely big boy behavior :3

2025-04-30 00:21:10 UTC

- reply

I would not in any conceivable world base energy usage or costs off of what a company in OpenAI's position is charging for its API - I just.. don't think that makes sense when talking about a VC company.
Yes they've dropped, a big part of that is that the market is more competitive and they have more investment money. That's not really surprising, I would expect to see that.
It doesn't mean their plans are profitable.
Honestly, if they're not selling at a loss.. why not?

2025-04-30 00:19:18 UTC

- reply

adsfasfdasdfasdffasdfas
Okay. So. Earlier you talked to me about the Silicon valley thing where companies try to massively expand by burning through investment money.
I am going to make a strong claim here that the most likely reality is that API costs for OpenAI are severly under-priced.
I think it would be very surprising during a gold rush with Microsoft money, while competing with Google - for that not to be the case.

2025-04-30 00:16:48 UTC

- reply

I don't necessarily think it's bad to be skeptical of corporate statements, but.. what data do you have that's better?
I'm sort of circling back around to - you seem much more confident that these things have dropped below cost to provide, but I don't understand why.

2025-04-30 00:16:42 UTC

- reply

Wait, hold on a second. I'm sorry - your assessment of this is based on the assumption that OpenAI is *lying* about its costs?
I mean, we've never gotten a statement from OpenAI saying that it is making money on its plans. I understand 2022 was ages ago in LLM time but we've had a really long period of OpenAI talking about this.
And sure *maybe* 03/04-mini has changed that. But.. do you know?
Is this based on anything?

2025-04-30 00:04:55 UTC

- reply

That strongly suggests that average usage (maybe because there's a lot of queries per customer on average, maybe because of how the data-centers work, whatever the reason is) - is not resource efficient. Even with high prices.
And this is not my wheelhouse, maybe other companies do better. But that's the silver lining, maybe OpenAI is just particularly bad at this.
But if you're saying they're not.. well, the evidence I can see suggests that ChatGPT specifically isn't efficient.

2025-04-30 00:04:07 UTC

- reply

I'll be honest, if your belief is that ChatGPT might be the most efficient out there right now for usage (and I would include training in that cost since LLMs need to be kept up-to-date with newer events/data) - that makes me more skeptical of the rest of the things you're saying.
I keep circling back to - OpenAI loses money on every customer. Not because of advertising or expansion, but because *providing the service* costs more than they charge.

2025-04-29 23:57:31 UTC

- reply

I mean this in a genuinely.. like I'm not trying to dunk here, and I'm not trying to score points or any bullshit like that.
But very honestly, I don't understand the certainty that you seem to have over energy numbers if you don't have this certainty over the rest of it.
If you don't feel confident that Deepseek is more efficient, then why are you using it as an argument that LLMs are getting better? You do believe that Deepseek is an efficiency breakthrough, right?

2025-04-29 23:52:31 UTC

- reply

Maybe Deepseek is a revolution. But you're not telling critters to all switch to Deepseek, you're using Deepseek as an excuse for why OpenAI is fine.
I'd have more sympathy for an argument that was "Deepseek is finally getting efficient, let's all move to that."

2025-04-29 23:52:09 UTC

- reply

And I mean.. heck, if your argument was, "don't use ChatGPT, use these other LLMs that are more efficient", I would have more sympathy for this.
You're saying that other companies aren't running queries at a loss - I'm again skeptical, but maybe you're right. I know that OpenAI doesn't seem to be making money on queries though, it's admitted that.
The irony here being that if you're trying to reduce energy usage, my guess is almost any other LLM would be better to use than ChatGPT.

2025-04-29 23:44:18 UTC

- reply

it is just not honest framing to remove that from the conversation and act like the only objection is whether an individual query costs a lot of energy.

2025-04-29 23:43:37 UTC

- reply

you're not saying that the pure per-prompt energy is low - you're saying that users do not need to feel guilty about regular usage.
The fact is, regular usage *is encouraging the gold rush.* That's the reason the gold rush is happening, largely the only reason.

2025-04-29 23:42:52 UTC

- reply

I won't necessarily dispute that, but I remain skeptical because the only analysis that I've ever seen (including your analysis) is based on estimation, not on actual measured usage.
And also.. no, I'm sorry, I'll call you out on this, that's *not* the only thing you're saying.
> I think this means that, as individual users, we don’t need to feel any guilt at all for the energy consumed by the vast majority of our prompts.
That is a much stronger statement, and one I think is false.

2025-04-29 23:39:53 UTC

- reply

If there's a giant gold rush happening that's causing huge amounts of environmental damage - which you seem to agree with - then it doesn't feel like it's a stretch to say that using the shit and encouraging the mass-pivot of the entire tech industry to that environmentally dangerous behavior.. is bad.
You want to evaluate this based on some theoretical energy usage, I want to evaluate it based on the energy/resources actually being used right now in the physical world.

2025-04-29 23:37:58 UTC

- reply

you have produced a lot of really insightful writing about AI - even as somepony who is much less charitable to this technology than you are - and I really fundamentally believe you are capable of understanding how nonsensical these arguments you're making are.
You are arguing in one breath that environmental impact is overstated, and in the other breath saying, "yeah it's bad but it's just because the companies are bad actors.."
Then it's not overstated!

2025-04-29 23:36:18 UTC

- reply

I think you're pretty fundamentally missing my point, but also - even charitably, I think this is a really silly answer.
"Casual usage of this thing is fueling an economically unsustainable and environmentally harmful gold rush - a gold rush that would not be happening if there wasn't casual demand for the product."
"If OpenAI didn't cause the gold rush, somepony else would have."
Like.. this does not mean that encouraging the gold rush is suddenly environmentally OK.

2025-04-29 22:55:07 UTC

- reply

It LLMs didn't exist, it wouldn't be happening.
That is just inarguable. And it is similarly just transparently obvious that the popularity and regular usage of LLMs is fueling that speculative investment and energy spike.
Do you seriously want to argue to me that Grok would exist right now if ChatGPT didn't?

2025-04-29 22:54:29 UTC

- reply

when I compare LLM proponents to Bitcoin proponents, I don't mean technologically - I mean the style of argument is almost identical.
The things that get hand-waved away, the things that get ignored, the things that I'm supposed to just price in.
And it always misses the fact that there is a giant massive energy spike happening right now during a critical period for the climate.. that is indisputably being caused by LLMs.

2025-04-29 22:53:19 UTC

- reply

And yes, you can compare it to a lot of stuff that costs a lot of money, the same way that you can compare Bitcoin to the world banking system - but it's not like we have 500 million people who stopped using Google searches. This is *new* energy expenditure, it's not priced in, we just added this on top of everything else.

2025-04-29 22:53:10 UTC

- reply

This is so similar to arguments that I hear from Bitcoin proponents. "Yes, it costs a lot, but look at the usage."
That usage is happening in addition to everything else. Most studies of LLM usage do not support the claim that we are getting economic value out of the 500 million active users. This is not disrupting markets the way proponents argued.

2025-04-29 22:52:53 UTC

- reply

Again, multiple issues here, #1 being that OpenAI loses money *per-customer*.
It's not that they have an efficient product that is just too popular - it doesn't scale. As more people sign up, they lose more money because it costs them more money to provide the service to each customer than they make from the customer.
If it was efficient - that wouldn't be the case.
BUT, I also want to jump back to Bitcoin for a second (and I know, I know, AI proponents hate this comparison), BUT..

2025-04-29 22:45:47 UTC

- reply

Explain how that can be true if this is using so little energy and resources.

2025-04-29 22:45:26 UTC

- reply

OpenAI is pulling revenue.
And their issue is not that they are pulling revenue and then spending it all on advertising. Their issue is that the cost of *providing the product* is higher than their revenue.

2025-04-29 22:44:40 UTC

- reply

I understand that this is common, but you're dodging the question - outscaling your competitors is expensive *because you are spending resources.*
Like you can argue that OpenAI is doing the same thing that Doordash did.
But that doesn't change the costs involved. It is still the case that providing this service is costing OpenAI more money than it raises.
And OpenAI's plans are *expensive* - and that's not enough money.

2025-04-29 22:36:52 UTC

- reply

There's a very simple followup question to the idea that LLMs are efficient - why is literally every single business who is attempting to sell access to them losing money? They're basically efficient, and also even the *highest* tiers OpenAI offers lose money on every single customer.
That math doesn't add up.

2025-04-29 22:36:31 UTC

- reply

Okay, but forget the data centers, let's talk about cost. "If you want to prompt ChatGPT 40 times, you can just stop your shower 1 second early."
Bullshit. You want to know how I know?
Because if I increase my shower by 1 second every single day - at the end of the month, my utility bills will not be $20 higher.

2025-04-29 22:36:15 UTC

- reply

There has undeniably been a *massive* new investment in this space since ChatGPT, and recommendation algorithms and computer vision haven't really changed. There aren't a lot of brand new markets for that - this is being fueled by *chatbots*.
I know that LLMs are not the same as lightbulbs, because in the past 2 years we have not seen proposals to reopen nuclear power plants in order to power lightbulbs.

2025-04-29 22:35:56 UTC

- reply

So we have an article with a lot of very theoretical math based on estimates of numbers that people don't have access to explain why LLMs are fine for the environment.
And that's great. Now explain the fucking data-centers. And before you pull the article's trick of saying they're for recommendation algorithms, they're not. We're talking about *excess* data centers. Data centers that were not being built until ChatGPT got popular.

2025-04-29 22:35:32 UTC

- reply

I've now read through about half of this, and there are so many problems here that I don't know where to start.
Multiple critters have brought up bitcoin comparisons, but I also want to bring up Covid. During lock-downs there were debates about how "lethal" Covid was and people would do all this math to try and prove it wasn't dangerous.
And the answer to that argument is you look at excess deaths and then ask, "okay, explain all of the excess fucking dead bodies."
#thread

2025-04-09 06:53:27 UTC

- reply

There is almost no downside to air filtration.
I started carrying a box fan filter to family gatherings and just plugging it in once it became clear that I was going to have to figure out other ways to manage risk.
The result? Siblings remarked that their allergies were better during spring gatherings.
Like, holy fuck. Even if Covid didn't exist, I would advocate for this. Air quality helps with basically everything.

2025-04-04 16:38:51 UTC

- reply

there should be a button in those situations where you can send a message reply to everypony at once that's just:
BARKBARKBARK ARF BAR BORK ARF WARWFF AWROOOOWA ARKBARK (am overstimulated)

2025-03-24 00:02:43 UTC

- reply

Bluesky advocates can spin up as many hobby projects as they want - at the end of the day, if you get banned from Bluesky, your connections there are severed, in practice - and everypony knows it.
Nopony is offering advice to direct-connect to banned furry's PDSes.

2025-03-24 00:00:41 UTC

- reply

It's been talked about to death already, but it's just so telling that every time something like this happens, nopony on the site is like "well, I guess I'm going to credibly exit/fork/direct-follow/whatever."
The "decentralized in theory" conversation has never held more true - and it gets buried under, "nu-uh, look at this technical project with a community relay", which is all well and good but..
..if it was decentralized, we wouldn't be having this conversation.

2025-03-20 03:21:08 UTC

- reply

> It works by creating a new, completely empty, filesystem namespace where the root is on a tmpfs that is invisible from the host, and which will be automatically cleaned up when the last process exits. You can then use commandline options to construct the root filesystem and process environment for the command to run in the namespace.
https://man.archlinux.org/man/bwrap.1
To me that's basically a separate linux system, but I am very lax fox with terms, so I would not blame anypony who disagrees X3

2025-03-20 03:15:12 UTC

- reply

It's been ages since I've worked with bubblewrap and flatpak, so maybe I'm getting mixed up here - but overlayfs is a part of bubblewrap isn't it?
You ship your own `/` directory I thought (or more likely use a base).
Maybe I'm completely mis-remembering this, but I could swear I was using bubblewrap at one point to mount transparent directories on top of other directories when running an app..

2025-03-20 02:58:27 UTC

- reply

I guess I'm also a little confused. I can't think of any problem with shipping an older runtime that wouldn't also show up with system libraries. KDE is the same size regardless, and my understanding is that flatpaks are literally Linux systems built off of essentially chroot and some file access permissions - you should (maybe I'm wrong) be able to do anything at all with a flatpak you could do with a linux filesystem.
There very well could be some deeper issue that I don't understand.

2025-03-20 02:45:48 UTC

- reply

that would definitely explain it, but yeah, that's exactly what I mean by "this is screwy and weird and why are they doing it"
If EOL runtimes are a security issue, why are they being shipped as versioned dependencies in the normal Fedora repos? (rhetorical question)
If EOL runtimes are not a security issue, why are they being updated for flatpaks?
Flatpaks can share runtimes, but they don't have to. You can have one using an up-to-date QT runtime and one using an old version.

2025-03-20 02:36:54 UTC

- reply

I'm doing a bad job of describing this. >w< The bundle of system libraries in a flatpak system is basically a linux system.
Whether you update a library in the core OS, or whatever you update a library in a flatpak, either way you are changing the runtime environment for that app.
I'm surprised because this implies that Fedora got rid of an EOL runtime for the flatpak but not for their core system. For the rpm to not break, Fedora would need to keep shipping the EOL runtime.

2025-03-20 02:19:20 UTC

- reply

that Fedora managed to apparently invert that is just very odd, and it makes me wonder what they're doing. I would have expected any security update they made to break every version of OBS they shipped.

2025-03-20 02:13:23 UTC

- reply

especially if they were saying those changes were for security - was Fedora not shipping security updates on the rpm?
Not doubting that they were doing something screwy, just it's bizarre maintainer behavior.
Part of the benefits of flatpak over rpm as a maintainer would be you could ship security updates to a few apps without re-packaging every app - as opposed to normal packages where dropping a system library with a vulnerability would require re-linking it for every package.

2025-03-20 02:12:27 UTC

- reply

Right, but what I mean is - there'll be more shared dependencies in an rpm than would be in a flatpak because they're not containerized - my understanding is that updating links inside of a package is common because of that - part of the job of being a distro maintainer is recompiling a bunch of software that you've updated to all use the same dependencies as each other.
What's odd to me is why Fedora would make *more* changes in a flatpak than they already had to make for the rpm.

2025-03-20 01:53:37 UTC

- reply

it's so strange to me that they'd choose to have a different compilation process for the flatpak and the rpm, but I haven't used Fedora in ages so maybe they're just doing weird things X3
I guess maybe I'm not certain.. but I would have said that I'm certain that arch is also updating runtimes for this stuff. I know aur definitely messes with dependencies to get packages working with the rest of the system.

2025-03-20 01:49:29 UTC

- reply

OBS definitely can request that and I get the frustration, but I'm also surprised it came to a head specifically over flatpak. Was Fedora doing something new with OBS that they weren't doing when they distributed it as an rpm? Recompiling and updating dependencies is the norm for almost every distro, isn't it?
I wonder if it's just that with flatpak users are more tempted to think that what they're installing is the official version so they're worse about filing bugs?

2025-03-19 22:54:10 UTC

- reply

Firefox would be a good use-case for that too - I think a few distros put tweaks on it that it has better privacy defaults.
So depending on the situation, I might not want the official firefox flatpak, I might want my distro's bundler's firefox flatpak.
But distros could be a lot more choosey about which apps they do that with and could update them piecemeal instead of all at the same time.

2025-03-19 22:50:43 UTC

- reply

I have some flatpaks (OBS) on my system that pull from separate repos to keep driver support for my old gpu.
I could see a distro enabling multiple flatpak upstreams and having maybe a smaller set of apps that they are repackaging, to fix specific security issues or for whatever reason.
But if using aur on Arch is the exception, this kind of flips it so Aur is the default. Most of what I get is from the authors directly, just with some exceptions.

2025-03-19 22:47:41 UTC

- reply

yes and no, I'll very lightly push back on this in the sense that.. it's more like flatpak makes "get it from the source" so simple that getting your software from a distro's upstream does not need to be the default (and shouldn't be the default for many situations).
But, I still see some value in a future flatpak world for bundlers to maintain separate repos where they patch flatpak apps or keep dependencies up to date or change defaults.

2025-03-19 20:49:55 UTC

- reply

Flatpak is pretty great. There's this academic debate about keeping apps in sync with each other and fixing issues and whether it's wasteful, but then you use them and it's like, "awarawwfa this is actually great, am happi fox"
I switched over to Arch specifically because I got tired of waiting for apps to be updated on Debian and kept breaking things when I self-compiled versions with different dependencies.
Flatpak is the evolution of Arch in that regard - I just get stuff up-to-date.

2025-03-03 21:35:28 UTC

- reply

That's why when I want the *best* sound quality I put a CD into a boom box and then light the boom box on fire. >:3

2025-03-03 19:50:01 UTC

- reply

I will say, if you're only looking for Telegram levels of accessibility, there are a number of pure-CSS approaches to this that will get you 95% of the way there with just a span.
https://jsfiddle.net/95xLhudv/ will do everything Telegram is doing except leave the spoiler active when you click away (which I would argue is better behavior anyway, it bothers me that you can't re-hide spoilers in Telegram X3). And as a bonus also actually works with a keyboard (Telegram's spoilers don't).

2025-03-03 19:27:39 UTC

- reply

ah, I might have to just agree to disagree on that X3
I kind of hate the fully opaque blurred style of content warnings, I feel like there's no point in hiding text if it's not going to tell me why the text is being hidden. If Telegram required essentially a details/summary *markup* where the spoiled text needed to also have shorter description - and then it maybe overlaid it over the inline span or something - I'd be happy with something like that.

2025-03-03 19:19:23 UTC

- reply

I do hope that they move details in the future to work inline better. It *almost* does, it's pretty much just text wrapping that's borked.
I feel like they could have done just a little bit more work and made it significantly more flexible - because other than the visual layout, the actual accessibility aspect - requiring not just a "hide this" but also an actual summary of what's being hidden, and removing it from the layout - is good design.
But.. text don't wrap in paragraphs >.<

2025-03-03 19:12:29 UTC

- reply

Ugh, noticed as I'm posting this that you already answered this elsewhere, sorry.
I'm not sure how I feel about mixing inline and block-level spoilers, but regardless, yeah, details is not going to help you with that.

2025-03-03 19:10:39 UTC

- reply

I know you've said in the past that details/summary isn't really what you want for this, and I believe you, but can you remind me why?
Details is pretty bad for inline stuff, but naively, "wrapping a content warning block inside of another click-through warning block" seems like it would work perfectly for this?
I'm thinking about what Mastodon does when you have a post with a content warning *and* a mute-word: you get two nested block spoilers to click through. It seems to work well?

2025-02-13 22:39:48 UTC

- reply

that's a choice! There are projects that are careful about upstreams deliberately where it's hard to contribute and that's fine, because you're not really meant to.
Linux isn't that - they deliberately didn't go for stable driver interfaces because they wanted code upstreamed.
Well, live with the choice. You can't have an attitude that all code needs to be upstreamed and also have an attitude of "nopony can upstream something I'm not comfortable with."

2025-02-13 22:36:53 UTC

- reply

I do think, independent of the language, I'm certainly not qualified to write kernel code, but also if I was I still wouldn't touch the kernel. I agree, it just.. every time I look at the mailing lists I think, "I would not want to be in there."
I'm also pretty frustrated about the "we have to be careful about what we upstream" framing, given that Linux has *always* had a "upstream your drivers" philosophy, it deliberately makes it hard to do out-of-tree drivers.

2025-02-13 22:35:29 UTC

- reply

"You can write safe C."
"Hey, check out our new rendering engine."
I'm just a bystander though.

2025-02-13 22:34:24 UTC

- reply

I feel like every time there's a debate about Rust, one side will talk a lot about culture-fit and technical limitations and all of this stuff, and the other side will go, "hey check this out, we built our new driver in a month."
Whatever technical limits or problems Rust has at a language, apparently they're not large enough to stop teams from repeatedly doing things with the language faster and on a smaller budget than is expected.
I noticed this with Mozilla too.

2025-02-13 22:32:23 UTC

- reply

I'm not deep enough in that world to have a technical debate about it and I haven't used Rust enough to know the pros/cons in more detail.
But also, you can't really argue with results. Building a complete driver for a new hardware with a team that small - clearly they're doing something right.