mspisces8

#MayTheForceBeWithYou #StarWarsDay

@ZackKorman @m0bilej0n @kmcnam1

@m0bilej0n @kmcnam1 I also exist not on this app

Awww... my friends Daniel Card and Zack Korman making international AI news this week <3

Please RT for reach. If anyone knows @I_Am_Jakoby IRL, can they please reach out? I don't know them or I would, but they seem to be in distress. @i_am_jakoby/video/7666589898442280206" target="_blank" rel="nofollow noopener">tiktok.com/@i_am_jakoby/v…

🤣🤣

Have to admit this is very satisfying to see

@kmcnam1 @ZackKorman

I’m having ice cream for breakfast because… Well because… I can!!! 😜

Jensen Huang just broke 𝕏. 2,000 followers per minute, 5M views per hour. His first post is a banger. He breaks down his thesis on why open-weight models are the future. This letter is signed by NVIDIA, YC, Meta, Microsoft, Palantir, a16z, and more. The core argument: Open-source software didn't just cut costs in the 80s. It built the foundation the entire internet runs on today, including systems used by the US military itself. The letter argues AI is at the same inflection point right now. Why open weights matter, according to the letter: → Startups, hospitals, universities, and small businesses can build on advanced models without training one from scratch or paying frontier prices → Competition keeps AI's gains distributed instead of concentrated in a handful of companies → Organizations get to own their data, their customizations, and their accumulated knowledge → Open models let a broad community find and fix vulnerabilities faster Kimi K3 was a huge shift for open-weight intelligence. Huang's own position: he called fears of hidden backdoors in Chinese models "a misconception," and argues that broader AI adoption, open or closed, ultimately increases demand for Nvidia's chips either way. The letter draws a clear distinction: distillation (learning from another model's outputs) is a normal part of how AI improves. Unlawfully extracting value from closed models is a separate issue that deserves separate legal treatment, not a blanket restriction on open models. Absolutely goated first post.

Here is the DEFCON 34 track talks for those wondering the topics. There is a good mix of subjects and technologies. Which talks are you most interested in? defcon.org/html/defcon-34…

AI safety people will literally run an unmonitored cyber eval for days instead of going to therapy.

Pre-blackhat reminder: edit the QR code on your blackhat badge so that its email goes to your least favorite person and have EVERY vendor scan it.

Another new sticker for hacker summer camp. 👀

Crazy thought. Why doesn't the U.S. make legit frontier level open models themselves? 🤔

date idea: i take a nap on you