The Mobile World Congress in Barcelona has made it clear that mobile phones have taken a back seat: artificial intelligence (AI) and robotics have been the main stars. AI has emerged as a driving force for factories, cities, and campuses, and while robots were not numerous, they captured attention with proposals from Honor, Unitree, and AgiBot. This trend towards automation and AI seems unstoppable. SpaceX's Starlink has introduced its satellite-to-phone service, which will launch in Spain alongside Orange, solidifying satellites as the best option in rural areas. Apple, on the other hand, counter-programmed the event with the announcement of new Macs featuring the M5 chip, including the affordable Mac Neo ($599) and the iPhone 17e. The M5 represents a significant leap for those running AI models locally, with performance improvements of up to 60% over the M4 and responses up to four times faster in models like Qwen or Gemma. In the AI realm, the week has been marked by rivalry between Anthropic and the U.S. Department of Defense, as well as the arrival of new models from OpenAI and Google. OpenAI has launched GPT-5.4 Thinking and Pro, with an expanded context of up to 1 million tokens, allowing for the management of much longer documents and conversations, matching rivals like Gemini and Claude. Additionally, it introduces an advanced reasoning mode for complex and prolonged tasks, useful in research and business automation. OpenAI aims to keep up the pace of innovation with frequent updates and a base of 910 million weekly users, although still below its target of 1 billion. The conflict between Anthropic and the Pentagon has been intense: Anthropic rejected contracts that did not exclude mass surveillance and autonomous weapons, while OpenAI accepted more flexible conditions. Dario Amodei, CEO of Anthropic, has publicly criticized OpenAI for offering security guarantees he considers insufficient and has pointed out political influence in the Pentagon's decision. This episode reflects the growing tension over the ethical use of AI in the military and the fierce competition among major players in the sector. Despite the controversy, Anthropic has seen an increase in downloads and subscribers. OpenAI has also introduced GPT-5.3 Instant, focused on improving the daily experience of ChatGPT: more accurate responses, fewer unnecessary negatives, and shorter alerts, achieving more natural and useful conversations. Google, for its part, has launched Gemini 3.1 Flash-Lite, its fastest and most economical model, ideal for high-volume applications like translation or moderation. It offers a very low cost and allows for adjusting the level of reasoning according to the task, standing out for its efficiency and flexibility. In the open weight ecosystem, Alibaba has presented Qwen 3.5, a family of models known for their power and versatility, especially in programming. However, the sudden departure of Junyang Lin and other key members has created uncertainty about the project's future, just as Qwen 3.5 is at its peak. If the team disperses, it could be a significant blow to the open weight sector, although it could also lead to new disruptive projects. SpaceX has taken advantage of the congress to promote Starlink among operators and regulators, although the dependence on U.S. technology and Elon Musk's figure raises concerns in Europe, especially after recent political tensions between the U.S. and the EU. In the realm of intelligent delegation, Google AI Lab has proposed an adaptive framework for AI agents and humans to collaborate more efficiently, transferring not only tasks but also authority and responsibility, anticipating the future of the "agent web." In China, according to The Economist, a new tech elite is emerging: the "nongchaoer," young people trained in elite universities and concentrated in hubs like Beijing or Shenzhen, thriving in strategic sectors like AI and robotics, driven by state support. However, this polarization leaves many young people out of technological opportunities. Don Knuth, a legend in computer science, has acknowledged the advancement of generative AI after seeing how Anthropic's Claude Opus 4.6 solved an open mathematical problem, illustrating AI's potential in scientific research. Other relevant news: Jensen Huang, CEO of Nvidia, rules out investing $100B in the company before its IPO; Google reduces commissions in its App Store and opens it to competition; OpenAI reaches $25B in annual recurring revenue, while Anthropic approaches with $19B; Perplexity advances in integrating Skills; and OpenAI prepares an alternative to GitHub, although it still faces stability issues. Source: elnacional.cat