Scientists are striving to understand the secret of the exceptional durability of ancient structures, hoping to find ways to create more resilient building materials. In Italy today, one can see buildings, roads, and aqueducts constructed from concrete that have existed for about two thousand years. Often, Italians find ruins preserved among modern constructions, such as a Roman villa discovered by students in a public school in Rome.
The question arises: how does the concrete of the Roman Empire differ from modern concrete, which lasts less than a century? Why do many bridges and buildings erected only a few decades ago already show cracks and structural damage?
For a long time, scientists have tried to unravel the mysteries of Roman concrete's resistance. Initially, the main theory was a chemical process—the pozzolanic reaction. This reaction occurs between finely ground silica-containing materials (known as pozzolans), hydrated lime, and water, forming a mortar.
In Roman constructions, pozzolans were volcanic ash, and the mortar was mixed with gravel, stones, and saltwater. Seawater contributed to the formation of an even more unique mineral—tobermorite. This was the key point: upon contact with sea air, tobermorite self-healed, filling cracks and pores.
However, new research suggests that something more might be behind the concrete's longevity. The work, published in the scientific journal Science Advances, resulted from a trip to Hadrian's Villa, located 17 kilometers from Rome. It is worth noting that Roman villas were not small rural settlements; the name refers to lavish estates that could serve as centers for small agricultural groups.
Hadrian's Villa contains a public latrine dating back 1900 years in its original condition. Paolo Monteiro, a co-author of the study and a civil engineer from the University of California, Berkeley, stated in an interview with Scientific American: 'No one restores the latrine. Therefore, the material remained untouched for 19 centuries, quietly conducting an experiment that no living person could start.'
The researchers took a concrete sample from under a toilet seat. The fragment was examined using various methods: under a powerful microscope, with X-ray tomography, and through chemical composition analysis. As expected, the sample showed signs that the pozzolanic reaction mixture—volcanic ash, lime, and water—had combined to form a material. Nevertheless, a more detailed examination of the pores and cracks revealed that calcite, a mineral composed of calcium, carbon, and oxygen, was the primary binding agent.
Calcite forms as a result of the reaction of the concrete's calcium compounds with atmospheric carbon dioxide. This mineral fills fine cracks and pores in the concrete, allowing old structures to strengthen and recover over time.
Monteiro noted in his statement: 'While the pozzolanic reaction is fundamentally important, our findings suggest that long-term carbonation also enhances the durability of concrete and may help seal cracks as it ages.' He also emphasized that tobermorite forms only upon contact with seawater, so these two explanations do not contradict each other but describe different types of Roman concrete. Tobermorite is found in structures exposed to the sea, such as ports and breakwaters, whereas calcite, the protagonist of the new research, is the main binding agent in land-based structures, like the analyzed latrine, located far from the coast. Thus, the Romans used different recipes depending on the structure's purpose.
Scientists hope that understanding the composition of Roman concrete will help modern specialists create concrete with a smaller environmental impact. Today, concrete is one of the most consumed materials globally, and its production accounts for about 8% of global carbon emissions. The UN predicts that half of the buildings existing in 2050 have not yet been built, making the development of less aggressive ecological methods urgent.
The study authors also caution that this carbonation process is extremely slow—it takes centuries, and one cannot assume that reproducing it will bring significant climate benefits within the service life of a modern building. The hope lies more in understanding the mechanisms of durability than in directly copying the speed of the Roman process. Monteiro concluded in his statement: 'This study demonstrates how studying ancient engineering techniques can lead to important discoveries. We hope that by uncovering Roman secrets to increasing concrete durability, we will one day achieve sustainable development of modern infrastructures.'
Most business leaders recognize the importance of artificial intelligence, but only a smaller portion of them has managed to make it truly useful. The gap between these groups is not related to ambition or budget, but most often to understanding which problem should be assigned to an AI agent first, and which solutions should remain under human control.
This question became the central theme of Nikila Gudipati's presentation, Head of Client Engineering at Google Cloud, at the closing event of MSME Sparks 2026. The event itself was significant because MSME Sparks 2026, YourStory's flagship event dedicated to Indian SMEs, spent four virtual days analyzing factors transforming the growth of these enterprises, including digital transformation, access to capital, and the role of AI in daily operations. The final event took place on June 26th at ITC Gardenia in Bengaluru and was the only in-person meeting of the week, gathering founders, industry leaders, and ecosystem participants to conclude discussions started during the virtual sessions.
In her presentation titled 'Smarter Solutions with AI,' Nikila Gudipati focused on the week's technological topic. She began with a statistic that clearly defined the problem: 'On average, a typical company works with 254 business applications.' Since the data is distributed across dozens of isolated tools, the problem is not just inefficiency, but also the so-called 'switching tax'—the constant mental and operational burden associated with switching between applications to gain a single holistic view of the business. Gudipati argued that Gemini Enterprise is designed to eliminate this tax by acting as a central hub for creating, publishing, and managing AI agents across all organizational workflows.
Two live demonstration examples vividly illustrated the practical application of this technology. In the first example, a contract lifecycle management agent was asked to identify expiring client agreements. Within a single dialogue, it extracted data from CRM and compliance systems, identified accounts with the highest revenue risk, and drafted a renewal letter, allowing for the scheduling or complete automation of subsequent actions. Gudipati emphasized: 'It's not just about asking questions based on prompts; it's actually the ability to act on your behalf so you can focus on more strategic and important things for your business.'
In the second demo, an agent tracked the situation of a retail operations director who was trying to understand the reason for low efficiency in one store. Using point-of-sale records, inventory data, and employee information, the agent diagnosed three main causes, including low customer satisfaction and high staff turnover, and then suggested scheduling a meeting with the store manager after pre-loading the relevant data. Examples of this scale demonstrate investment firms using multi-agent systems to help analysts research five times more companies, or large manufacturers reducing HR query resolution almost to automation while halving defect resolution time.
For SMEs attending the event and looking for an entry point, Gudipati advised starting with reversible and low-stakes solutions. She asked: 'Which of these reversible solutions can you delegate to an agent so you can refocus on the critical and strategic aspects of your role?'
Regarding security, which often remains an unspoken concern in any AI discussion, her answer was direct: 'The key difference with Gemini Enterprise is security. It's the most boring thing. It's the least fantastical thing companies worry about, but it is actually our biggest advantage.' Gemini Enterprise offers a 30-day free trial, available directly through Google Search, which is a low-risk, reversible starting point that accurately reflects the logic Gudipati advocated throughout the session.
A team of Chinese and international researchers has created the world's first spatially resolved map of the spectral index of a black hole at the scale of its event horizon, opening new possibilities for studying the extreme physical environment around these objects.
The study, published in The Astrophysical Journal Letters on Monday, was conducted by scientists from the Shanghai Astronomical Observatory of the Chinese Academy of Sciences in collaboration with international partners. The team analyzed data obtained in 2018 using the Event Horizon Telescope (EHT) and the Global Millimeter VLBI array, studying the supermassive black hole M87.
By combining images taken at wavelengths of 1.3 millimeters and 3.5 millimeters, the researchers performed the first two-frequency spectral analysis at the event horizon scale. The resulting map of the black hole's spectral index allows for a description of how the radiation brightness changes at different frequencies, which scientists compared to a high-precision 'health check' of the black hole.
The results obtained demonstrate significant changes in the properties of radiation around the black hole depending on the distance from its center. Near the event horizon, the spectral index is positive, indicating a strong influence of synchrotron self-absorption on the radiation. However, further from the center, the spectral index gradually decreases and becomes negative, suggesting a transition of the radiation into an optically thin state.
This transition occurs at approximately 30 microarcseconds from the center of the black hole, corresponding to the size of the ring structure observed at the 3.5 millimeter wavelength. This discovery suggests that this ring is not only a visible feature of the black hole image but also reflects the state of the plasma near the event horizon.
The researchers note that future observations at additional frequencies will allow for more precise measurements of the plasma conditions around black holes. This will help track the evolution of accretion flows and jets over time, advancing the study of black holes from static imaging to dynamic physical diagnostics.
A study conducted by CGTN raises fundamental questions about humanity's role as machines begin to acquire cognitive abilities. The authors note that every technological revolution has always been accompanied by both the expansion of human capabilities and a re-evaluation of the very essence of humanity, but artificial intelligence represents a deeper shift.
AI is no longer just recording what happens; it is beginning to predict the future. It goes beyond merely interpreting the existing world, gradually building a hierarchy of possible scenarios before their realization. Thus, prediction transforms from a description of a potential future into an active force shaping subsequent events.
The central question that artificial intelligence poses to us is not whether machines can think like humans, but rather what authority their judgments are beginning to acquire. Questions arise about who grants this authority, who is responsible for it, and what future this will create for all of humanity.
At the World Conference on Artificial Intelligence and the high-level meeting on global AI governance in China, a concept for AI regulation was presented, which calls for maintaining a human-centric approach. The goal is for AI to serve as a powerful engine for common prosperity and collective security, contributing to the construction of a fair and equitable global AI governance system.
This vision aims to eliminate not only the gap in technological capabilities but also the growing divergence between the speed of technological progress and the pace of achieving consensus on governance. Since technological advancements outpace the formation of rules, power may emerge before appropriate norms are established. Therefore, AI governance is not just about managing risks, but also about finding a balance between innovation and rule-based order.
A global online survey by CGTN showed that 85.4% of respondents believe that the rapid development of AI generates new challenges for global security and governance. Furthermore, 92.6% expressed concern that deepfakes and AI-generated disinformation could undermine the foundations of social trust, while 87.8% worry that technological barriers and 'digital walls' could lead to the fragmentation of international scientific and technological cooperation.
These data indicate that issues related to AI have long gone beyond purely technological spheres. Strengthening global governance and forming broad international consensus have become shared tasks in the age of artificial intelligence. The survey also revealed that 82.6% of participants believe that developing countries and the Global South should play a more significant role in defining AI governance rules to ensure fairness and inclusivity of the system. Moreover, 87.5% hope that the international community will work to promote accessible and comprehensive AI, preventing technology from becoming a new source of inequality.
These results reflect not only general public expectations regarding AI governance but also a growing international consensus that the future of AI should be determined by openness and cooperation, rather than technological monopolies, and by inclusivity and common benefits, rather than regulatory restrictions. A civilization that can be understood, generalized, and interpreted exclusively through the models of another's AI has not yet reached a true era of intelligence. A truly fair global AI governance system must respect the diverse trajectories of development and the value systems of different countries and cultures, guaranteeing every civilization the right to understand the world through its own conceptual framework, to train intelligent systems on its own terms, and to appropriately shape the future of AI.
The values and ways of thinking inherent in different civilizations are not 'noise' that algorithms should eliminate, but indispensable sources of wisdom necessary for humanity to navigate an uncertain future. The goal of global AI governance is not to create a single 'correct' model of the world, but to ensure that no single understanding of the world becomes dominant. Governance is not the enemy of innovation; it is the second creation of innovation. The first occurs in the laboratory, where machines gain capabilities; the second is realized through institutions, ethical norms, and international cooperation, where technologies are endowed with purpose, boundaries, and responsibility. Intelligence without governance is merely force. However, intelligence shaped by shared values allows technology to become a genuine force for the benefit of all humanity.
In the age of artificial intelligence, AI governance concerns more than just common prosperity and collective security; it concerns our collective ability to shape the future. A truly just AI order must do more than just ensure the distribution of benefits from technological progress among all. It must also guarantee that every country and every civilization has a voice in setting the rules and determining the direction of AI's future. The future should not be a pre-determined model generated by a system, but an unfinished draft written collectively by humanity, remaining open to revision by all parties. The fact that machines have begun to think marks a milestone in technological history, and humanity's ability to jointly bear responsibility and maintain freedom of choice is a sign of civilizational progress.
