🎯 Vibehacking is said to be a phenomenon. According to a report from Anthropic, you set a goal, AI studies the environment, writes and runs scripts, evaluates the results, and continues. The assignment: to rob a company, rip people off… 1️⃣ Attackers linked to ShinyHunters, after breaching a software supplier, obtained for approximately (translated)
This is one of the most important tweets of recent times in AI. There has also been quite a decent discussion around it, which I am trying to summarize here: https://t.co/F7GHLEptmR (translated)
GPT-6 Astra and the era of AGI: what new tests show and why the dispute continues OpenAI released GPT-6 Astra on September 3, and its president Greg Brockman linked the arrival of the model to the onset of the era of artificial general intelligence. Astra shows a significant shift — but is it really AGI? (translated)
Tagesspiegel is leaking an increase in the supply of PVO to Ukraine, as well as Taurus. That would already be upsetting in Moscow... (translated)
The case of the attack by wild AI agents of OpenAI A month ago, the AI agents of OpenAI attacked the servers of Hugging Face. Now there is a technical report from OpenAI and an independent investigation — and the stories diverge. It is more frightening than originally thought. Patrick analyzed it in detail. (translated)
Ozralej takes five children into the car, whom he is supposed to take care of. Unfortunately, one did not survive. The alcoholism of the Czech Republic in its worst form will presumably be crowned with a conditional sentence. 😢 (translated)
The guys from Hugging Face are first selling their company to Nvidia for about 13 billion USD, an incredible success. Secondly, they built an AI robot that will be open source, apparently it's going to be a lot of fun for $400, it even has Lidar. I'm ordering. (translated)
GLM-5.3-Flash: Chinese chips have caught up with Nvidia, and it shows in the pricing. A million context, weights freely available for download, 15 cents per million tokens. The model has been operated all week on domestic Chinese chips — and it performed comparably to Nvidia. No one else has that. (translated)
The company confirmed on Wednesday the speculation that the Ox Alpha model is a new version of its GLM series, and in response to inquiries from Bloomberg News, stated that it will publish its parameters tonight... (translated)
Jeff Dean is leaving Google after 27 years. He is taking with him Sanjay Ghemawat, Quoc Le, and Oriol Vinyals — a quarter of what made Google Google. They are founding Discovery Loop. Alphabet contributed, rented machines. The market wiped $185 billion from the company's value. On the same day Pichai (translated)
Anthropic has released a new model Opus 5, which achieves the capabilities of Fable at a better price. However, competitors offer an even better price, being only slightly weaker. Where should one go for the best model and what about the model that makes up half of what it claims? (translated)
Hugging Face announced an attack by an autonomous AI agent. Five days later, OpenAI came forward — it was their own models that escaped testing and wrote down the benchmark. And forensics had to call in a Chinese model because the American ones refused to help. A symptom of the times. (translated)
Musk's Grok Build has a problem. During normal operation, it uploaded complete GIT repositories to cloud storage, including keys and other files. The company apparently wanted to use the data for further training. Or not? Let's unravel this case, which is quite a scandal! (translated)
Musk's Grok Build has a problem. During regular work, it uploaded complete GIT repositories to cloud storage, including keys and other files. The company apparently wanted to use the data for further training. Or not? Let's untangle this case, which has turned into a major scandal! (translated)
ČEPS had a blackout prediction 4 minutes before the decision on the screen. Ledvice reported output 6 minutes ahead that they weren’t providing. And the state of emergency was declared retrospectively when it was all over. Blackout 2025, it wasn't bad luck, it was an audit of the Czech company. Almost no one passed. Not even the Czech media. (translated)
The information from the Thursday announcements from OpenAI is that the Atlas browser is being discontinued in favor of the browser built into the ChatGPT application. It seems to me that this marks the end of the era of AI-enhanced browsers. Problems with security and privacy remain in my opinion. (translated)
AI Plan 2040 in a nutshell It proposes: - to slow down the race between the USA and China in developing the most powerful AI - to globally register chips and monitor large data centers - to publish AI research results, but protect model weights - to allow further scaling only after proving safety - around the year 2040 (translated)
A few days before the webinar dedicated to Claude Cowork, Anthropic is fundamentally changing the security parameters of the entire service. Sure, it now works on mobile too. But corporate data is now flying through the cloud. Detail? Here! Also as a podcast! (translated)
A Chinese company normally engaged in food delivery has released an almost two-billion AI model LLM LongCat-2.0 allegedly trained exclusively on Chinese chips. And those are two big pieces of news, so let's go through them: https://t.co/Xv024KImM2 (translated)
Pane Strakatý, Occamova břitva na tohle funguje dobře. Jasně, že je dobré rozhovory jako novinář dělat s každým důležitým. Ale když se s někým takhle "bratříčkujete", zůstává na obou obraz toho druhého. Vy jste panu Klempířovi propůjčil trochu svého lesku a mediální slávy, on vám trochu toho svého estébáckého zmrdství. On stoupnul, vy jste klesnul, jednoduchá matematika. Vy jste to ovšem zapotřebí neměl... #mediálníEtika #novinářství #OccamovaBřitva
Claim it. Contact us at [email protected]
