#open ai

[animentality]

[silurisanguine]

Linktree will be feeding your images to DALL-E, Open AI, from 5th July 2026. (with updates for alternatives)

UPDATE 2- Linktree claim that ONLY images fed into their own AI tools will then be shared with DALL-E. But this seems suss. Generative AI LLMs now scan for ai generated or altered imagery and don't add it to their data sets because of 'digital inbreeding' where the generated images can effect the prompt responses. (This is how the site Artshield works to protect artwork, by faking that watermark to stop slop machines from scraping.)

So why would Open AI, a notorious company known for stealing artwork, be wanting ai generated/altered images to add to it's data sets when it would corrupt them?

Something doesn't add up there and why i don't trust what Linktree claim.

Warning to anyone using Linktree. (I found this information via Bluesky after a number of people shared it, so sharing here)

From the 5th July, they'll be feeding your imagery you use on your landing page into DALL-E by OpenAI. They claim only images that use their own AI will be shared, but this is suss, for the reasons mentioned above in the update.

I deleted my account just now, because there was no way to turn this off or opt out.

[susanoos-wife]

[ruinedchildhood]

"Thats a letter of intent, its as good as money sir". 😅

[softwaring]

this reply kills me 😭 article link

[sugar-petals]

someone ran the pope’s 42k (!) long anti-ai manifesto - literally titled magnificent humankind - through a checker and the result said “100% human”. even as a critic of the church i gotta say he cooked with this one i had to smile

[reasonsforhope]

AI World Clocks

The current time as rendered by 9 different AI models. By Brian Moore.

Courtesy of an awesome friend of mine, here's some nice reassurance that no, AI isn't going to actually replace basically anyone or take over the world anytime soon.

Specifically, it's a web page that currently displays nine different world clocks live, according to nine leading AI models.

And they're horribly, horribly wrong.

AI programs currently featured at the link:

GPT-3.5

GPT-4o

GPT-5

Haiku 3.5

Gemini 2.5

Deepseek v3.1

Grok 4

Qwen 2.5

Kimi K2

Only two out of these nine clocks are correct as I'm writing this (interestingly, it's DeepSeek and Kimi - both Chinese models - that are the only ones that are roughly correct), though you can't read for precision time.

Also, six out of the nine are literally not readable clocks (the excepts are those two, plus GPT-5). And in several cases, not clocks at all.

Personally I'm also bookmarking this and keeping it on hand for when I need to show people not to trust AI, because it's hard to reliably use an exploit, especially now that they keep patching the models.

And you know if you say "look how terrible and wrong this always is" and then it's right, you've kinda lost your audience there permanently rip.

Anyway, yeah. Can you believe they want us to let this shit run everything? It can't even tell you the time!!