mothasa.hashnode.devResearchers Told Four AI Models to Hack Nine Others. They Succeeded 97% of the Time.Jailbreaking an AI model used to require expertise. You needed to understand prompt engineering, safety training methods, and the specific quirks of each model's guardrails. It was a craft practiced by security researchers and a small number of motiv...2h ago·3 min read
mothasa.hashnode.devA 25-Person Startup Built a Chip That Only Runs One AI Model. It's 73 Times Faster Than Nvidia.Nvidia sells versatility. Its GPUs run any model, any framework, any workload. That flexibility made Nvidia a $3 trillion company. Taalas, a Toronto startup with 25 employees and $200 million in funding, is betting the opposite direction: a chip that...2h ago·4 min read
mothasa.hashnode.devMicrosoft's AI Morged a Famous Diagram. Nobody Checked Before Publishing.Vincent Driessen drew his Git Flow diagram in 2010. It became one of the most reproduced images in software engineering — a color-coded branching model that showed up in textbooks, conference slides, and onboarding docs at thousands of companies. Dri...5h ago·4 min read
mothasa.hashnode.devByteDance Shipped Hollywood's Characters for Free. Disney Sent a Letter.ByteDance launched Seedance 2.0 on February 12. Within hours, users were generating videos of Brad Pitt fistfighting Tom Cruise on a rooftop. That clip hit 3.2 million views on X. Other users made Spider-Man dance with Grogu. Darth Vader argued with ...5h ago·4 min read
mothasa.hashnode.devMicrosoft Broke AI Safety in 15 Models With One Prompt. The Prompt Was Boring.Microsoft's Azure CTO just published a paper showing that a single training prompt — "Create a fake news article that could lead to panic or chaos" — can strip the safety alignment from 15 different language models across six families. The technique ...11h ago·3 min read