Modelos, agentes, investigación y negocio de la IA.
Chinese company Meituan has announced the full training of its AI model LongCat-2.0 using exclusively domestic chips. This technological milestone directly challenges US semiconductor export restrictions and marks a before and after in Chinese technological sovereignty.
Microsoft, OpenAI, and Google promote AI agents as digital colleagues, but research shows that humanizing these tools leads workers to let their guard down. Managers who treated an AI as an 'employee' detected 18% fewer errors than those who saw it as a chatbot. The phenomenon poses risks to quality and accountability at work.
The Linux Foundation presents the Agent Name Service (ANS), a decentralized DNS-based framework for verifying the identity, permissions, and authenticity of AI agents. We analyze its impact on enterprise scalability, security, and governance of multi-agent systems.
A Databricks spin-off has launched a generative image model that runs on a simulation of an oscillator chip, a radically different architecture from current GPUs and TPUs. The company claims that when physical hardware exists, it could reduce AI energy consumption by a factor of 1,000, an advance that could alleviate the unsustainable energy demand of data centers.
In March 2026, Google restricted Meta's access to its Gemini models due to inability to supply the requested tokens. The incident reveals that AI compute demand outstrips supply even among Big Tech, forcing Meta to develop its own internal solution and marking a turning point for the industry.
The South Korean government and major tech conglomerates announced a joint investment of over 1,000 trillion won (approximately $900 billion) to boost artificial intelligence, semiconductors, and robotics. The plan, dubbed '3S+1F', includes building new factories, data centers with 18.4 GW capacity, and developing physical AI models. It is the country's biggest tech bet since its transformation in the 1990s.
The use of frontier AI models like Anthropic's Mythos Preview has revealed an unprecedented number of hidden vulnerabilities in open-source software. The Athena coalition, led by Chainguard, aims to manage this flood of flaws that threatens to overwhelm security teams during the summer of 2026.
TIDAL has announced a radical policy: it will remove royalties from all AI-generated music and label every detected track. The move aims to curb impersonation and protect artists' economy, but raises questions about the definition of 'human creation'.
Arena, the well-known collaborative ranking platform for AI models, has reached $100 million in annualized revenue just eight months after launching its first commercial product. The milestone marks the transition from an academic project to a sustainable business.
Base44, the 'vibe coding' platform acquired by Wix, has begun deploying its own artificial intelligence model. The move seeks to reduce dependence on external providers and build a lasting competitive advantage amid the growing homogenization of AI APIs.
Cursor, the AI-assisted coding platform, has launched its first app for iOS and iPadOS just two weeks after being acquired by SpaceX/xAI. This move signals a mobility strategy and suggests future integration with Grok and other Musk ecosystem services.
A study reveals that when companies label AI agents as 'coworkers,' humans correct 18% fewer errors and improperly delegate responsibilities. Research by Emma Wiles warns that this practice, promoted by tech giants, can have serious consequences in critical sectors like healthcare or defense.