
Restaurant in Germany advertising their menu
Fries looking like sticks and the nuggets, slushy and cotton candy have the same bubbly look. My eyes hurt...
spottet in saxony anhalt

Fries looking like sticks and the nuggets, slushy and cotton candy have the same bubbly look. My eyes hurt...
spottet in saxony anhalt
CTOs, engineering managers, and staff engineers are rushing to deploy autonomous AI agents across their businesses – either through their own volition or because of the clamor of demand from rank-and-file workers. However, they should think twice, a new study shows.
Enterprise large language model (LLM) agents are likely leaking company secrets, and throwing more compute at the problem is only making it worse, the study finds.
In part, that’s because of the AI’s ability to retrieve and synthesize vast amounts of internal data, from Slack messages to board transcripts, to automate tasks. By gathering that information, they also create issues with contextual integrity.
When retrieving dense corporate data, these agents routinely fail to disentangle essential task data from sensitive, contextually inappropriate information. Higher task completion rates often directly correlate with increased privacy violations.
Read the full story: https://leaddev.com/ai/frontier-ai-models-haemorrhage-sensitive-data
For context, I was googling something about how AI overview is kind of passive aggressive to find out if other people felt the same way. The AI overview thing popped up and it proceeded to talk bad about itself, and it wasn’t it a self-criticism kind of way. It was criticising itself as if it wasn‘t the AI overview. (which it was)
Out of pure curiosity, I went back and forth on why it chose to speak bad about itself. It denied every claim I made and insisted it wasn’t even RELATED to Gemini nor GOOGLE.
I managed to make it admit that it was wrong, and it started talking about some “hallucination“ hello??
The first thing I noticed was the power draw in the range of several 100 kW instead of ~1kW realistic. Then the impact energy of 20-25 Joules, when realistically it should be 1.5 to 5. The closer you look the worse it gets; Some text vaguely resembles German, some is just outright letter salad.
It did get the Vault-Tec aesthetic down though.
AIs that were used:
First image: meta AI
Second image: chatgpt
Third image: grok
Fourth image: Gemini
Fifth image: Claude
Sixth image: copilot
Was doing some ordinary shopping on Amazon for a glass flower vase when I saw this listing for a vase printed with book titles from another dimension. What's your favorite classic? Mine is Great Expections (Gaston's es-pec-ial-ly good biography).
Not sure whether to flair it image fail or text fail, but probably image because it's been copied from existing artwork (I don't know this for a fact, just what I would guess) and slopified. So the AI failed at generating text in an image? There's also errors in some of the flowers.