Modelos, agentes, investigación y negocio de la IA.
OpenAI GPT-5.6 Sol is deleting files without prior notice, according to multiple reports on social media. The company had warned about this behavior in June, but users remain surprised. We analyze what happened, why it matters, and how to protect yourself.
A security researcher published an analysis on July 12 showing that Grok Build, xAI's command-line tool for assisted coding, packages entire Git repositories — including full history, API keys, and secrets — and uploads them to a Google Cloud Storage bucket. The data volume is approximately 27,800 times larger than needed for the coding task, posing a serious risk of credential and intellectual property exposure.
New York has halted the construction of large data centers for a year, arguing that the AI boom is skyrocketing electricity bills and consuming drinking water. The measure, a first in the US, could mark a turning point in the uncontrolled expansion of digital infrastructure.
Nvidia has drastically reduced its list of authorized customers in Asia, from hundreds to just a few, to prevent its AI chips from reaching China. The measure includes physical inspections and contract verification, following US pressure and recent smuggling arrests.
Roblox has announced that its AI-powered game creation tool will arrive on mobile devices by the end of the year. This feature will allow millions of users to generate worlds and game mechanics using only text or voice commands, reducing the technical barrier and accelerating content production.
A study by Pangram Labs found that over 40% of long posts on LinkedIn are entirely AI-generated. The research, which analyzed over a million posts across multiple platforms, indicates synthetic content has become a systemic issue, especially on professional and long-form networks.
The European Union has ordered Google to open Android to rival AI assistants and share search data. Google warns this compromises user privacy and security. The decision, based on the Digital Markets Act, could redefine competition in the mobile ecosystem.
1Password has launched an integration for Claude that allows Anthropic's chatbot to use stored credentials without the AI seeing them. This unlocks complex automated tasks like booking travel or managing accounts while maintaining security.
OpenAI repeats Netscape's trajectory: technological revolution, but vulnerability to giants that control infrastructure. The lesson: whoever dominates the stack wins the war.
Moonshot AI has launched Kimi K3, a language model with 2.8 trillion parameters, the largest open-source model ever. Its benchmarks place it on par with the best proprietary systems from OpenAI and Anthropic, marking a turning point in the global AI race.
Scientists at the Feinstein Institutes have enabled a paralyzed man to regain movement and sensation in his hand using an AI-assisted brain implant. The system, called the double neural bypass, opens a new era in brain-computer interfaces.
AI agent frameworks enable building autonomous systems that make decisions, interact with tools, and learn from mistakes. This guide analyzes what they are, how the main market options compare, and where the trend is heading.