Meta has rolled out Muse Spark 1.1, its newest AI model designed to help developers write, debug, and manage code at scale. Building on a version first announced in April, Spark 1.1 can reason through multi‑step problems, automate digital workflows, and even roll out new features inside large enterprise systems. While rivals like Anthropic and OpenAI have offered similar capabilities for a while, Meta is betting on price to win users. The company will charge $1.25 per million input tokens and $4.25 per million output tokens—just a shade above Anthropic’s Claude Haiku 4.5 and OpenAI’s GPT‑5.6 Luna, but still positioned as a low‑cost alternative. CEO Mark Zuckerberg called Spark “a strong agentic and coding model at a very low price,” highlighting its strengths in tool use, computer interaction, and planning across apps. He also hinted that more models are on the way. The launch came alongside Meta’s new Muse Image generator and follows a busy week of AI announcements, including SpaceXAI’s Grok update and OpenAI’s GPT‑5.6 family. Meta pitches Spark as a solution for big‑ticket tasks like bug fixing, code migrations, and enterprise‑level automation, and Zuckerberg even broke his three‑year silence on X to share the news.
Read more
Google’s synthetic‑image watermark system, known as SynthID, stepped into the spotlight this week when it helped verify that a widely shared picture of Senate Minority Leader Mitch McConnell was a deepfake. The image, which showed the veteran politician in a compromising pose, had been circulating on social media and was being used to fuel political rumors. By scanning the picture with a Gemini model, the system detected the hidden watermark that Google embeds in every image generated by its own AI tools and by partners that have joined the program. Because the watermark was missing, the image was flagged as likely fabricated. OpenAI’s new public verification tool confirmed the finding, and experts said the result underscores how watermark‑based detection can cut through the growing tide of AI‑generated misinformation. The collaboration between Google, OpenAI, and other industry players began in 2025, with Anthropic opting out. While SynthID can only verify images when the generation tool participates in the watermark program, its success in this high‑profile case shows the technology’s potential to protect public figures and the broader public from deceptive visual content. As deepfakes become more sophisticated, tools like Google’s detector are poised to become essential safeguards in the fight against digital deception.
Read more
SpaceX has just launched the world’s first commercial satellite that runs on nuclear power. The tiny CubeSat, named BOHR, carries a tritium‑based power source developed by City Labs. Unlike traditional reactors, the tritium system emits very low radiation, making it safe to handle, transport, and integrate into standard launch procedures. The project was funded by a Department of Defense contract and marks the first nuclear‑powered mission cleared by the Federal Aviation Administration under the 2019 National Security Presidential Memorandum‑20. City Labs hopes this successful flight will open the door for more nuclear‑powered spacecraft, benefiting both national‑defense operations and private space ventures. One of the long‑term goals is to scale the technology for larger missions, such as supporting lunar bases at the Moon’s south pole—an area rich in water ice that could be turned into fuel and life‑support resources. NASA is already investing in nuclear reactor tech for those future habitats. While BOHR’s NanoTritium unit can’t yet power a moon base, it demonstrates a promising step toward that vision, showing that safe, compact nuclear power could soon become a staple of deep‑space exploration.
Read more
QuantumDiamonds, a fledgling chip‑inspection startup, just landed a major boost from the European Union and is gearing up to speed up how semiconductor factories build chips. The company’s new tools can quickly scan the tiny, layered structures that modern processors use—especially the 3‑D chips that power today’s AI data centers. Why does this matter? As transistors—the tiny switches inside chips—reach their smallest possible size, manufacturers can’t simply make them smaller to get more power. Instead, they stack more layers on top of each other, creating complex, multi‑layered chips. Checking each layer for defects is painstakingly slow with today’s equipment, creating bottlenecks in production. QuantumDiamonds’ technology promises to cut inspection time dramatically, letting factories move faster from design to finished product. Industry insiders say the startup now enjoys a first‑mover edge: no U.S. or Asian firm has yet shipped a comparable inspection system. While big players with $100 billion market caps will eventually catch up, QuantumDiamonds is already transitioning its tools from client labs into full‑scale semiconductor fabs. If the rollout succeeds, the move could accelerate chip output just when demand for AI‑driven hardware is soaring.
Read more
Researchers at Chinese tech firm iFLYTEK have unveiled a new AI system called iFLYTEK‑Embodied‑Omni that can see, understand language, and act—all within a single, brain‑like model. Unlike older robots that handle vision, prediction, and movement in separate steps, this model uses a shared “multimodal self‑attention” mechanism that lets its visual‑language core (the "high‑level brain") work hand‑in‑hand with an action‑generating module (the "low‑level cerebellum"). The result is smoother coordination, fewer mistakes, and faster responses. The team trained the system on a massive mix of human‑labeled videos, robot‑captured footage, and image‑text pairs, using a staged approach that lets the AI learn basic perception before mastering complex tasks. The outcome is a versatile agent that can navigate real‑world environments, answer questions, and even help with research discovery. Meanwhile, startups like Adaption are offering tools such as AutoScientist that automate AI model training for companies lacking in‑house expertise, while big players like Anthropic have shown the risks of over‑reliance on a single model when they blocked certain requests to their Fable 5 system. Executives, including Palantir’s Alex Karp, warn that handing data to frontier labs can cede control of critical technology. The article concludes with a personal experiment: the author built a simple Claude‑powered assistant to fetch research papers and is now testing whether it can evolve on its own—showing that self‑improving AI is no longer the exclusive domain of massive labs, but a playground anyone can start exploring.
Read moreChina’s top scientists say the coming 6G network will be far more than a faster version of 5G – it will turn the whole world into a giant, intelligent nervous system. In a recent interview, academician Zhang Ping explained that 6G will fuse artificial intelligence, massive data flows and ultra‑low latency to let humans, machines, objects and even virtual sensations talk to each other in real time. Imagine downloading a 4K movie in seconds, monitoring a mining site 24/7 from a smartphone, or receiving a virtual rose that you can both see and smell. These scenarios could become everyday reality once 6G’s “intelligent agent” network is in place. The technology builds on China’s massive 5G rollout, which now covers almost every village, providing a solid testbed for the next leap. With AI models exploding in size, 6G will shift from merely moving data to moving ideas, enabling smart factories, autonomous transport, personalized healthcare and immersive media. Mastering 6G is seen as a strategic priority for China, promising to reshape how we live, work and interact with the digital world.
Read moreQuantum computing – the ultra‑fast, ultra‑powerful next‑generation computer technology – is moving from laboratory curiosity to a real‑world race. In the past year, research teams in the United States and China announced they had achieved "quantum supremacy," proving that a quantum machine can solve a problem that even the best classical supercomputers cannot. That milestone has pushed the field into a critical transition phase: scientists now aim to build a "fault‑tolerant" quantum computer that can run useful, reliable applications. Industry giants such as Google and IBM, backed by massive government funding, say they expect practical, error‑corrected machines by the end of the decade. The United States has poured more than $1.2 billion into quantum research through its National Quantum Initiative and the CHIPS Act, while the European Union, Japan, Russia and India have launched their own multi‑billion‑dollar programs. UNESCO’s Global Quantum Initiative reports total global investment has already topped $55 billion and could exceed $100 billion by 2040. China is also a front‑runner, with university teams developing both photonic and superconducting prototypes and preparing for large‑scale industrial rollout. Experts warn that hype and unrealistic promises can cloud progress; true breakthroughs will require open international collaboration, steady engineering advances, and a full‑stack ecosystem that can turn fragile qubits into dependable machines. The next few years will decide whether quantum computers become a practical tool for drug discovery, climate modeling, AI, and more, or remain a costly scientific showcase.
Read moreResearch firm Omdia has lifted its 2026 forecast for China’s semiconductor market to about $812 billion – a 92.9% year‑on‑year jump and roughly $266 billion higher than its earlier estimate. The surge is tied to China’s massive rollout of AI‑focused computing and data‑storage chips, a segment expected to grow 126% and capture nearly 63% of the nation’s total semiconductor demand. Globally, the semiconductor market is projected to top $1.6 trillion in 2026, with storage chips alone accounting for about 55% of that value. Storage‑focused semiconductors are the standout, with China’s market predicted to expand 262.9% to $449.6 billion, pushing its share of the domestic market from 29% in 2025 to over 55% by 2026. Worldwide, storage chips are slated to reach $886.4 billion, also roughly 55% of the global pie. U.S. Semiconductor Industry Association data released in May shows China’s sales up 88.8% year‑on‑year, contributing to a global sales surge of 104% compared with May 2025. Analysts say the AI wave is reshaping the semiconductor landscape, with China emerging as a key engine for both domestic innovation and the broader global market.
Read moreChina is rapidly turning itself into a nation that gets most of its electricity from clean sources like wind and solar. This shift is not just a national policy; it’s a milestone for the whole planet because it shows how a huge economy can move away from coal and oil while still growing. Since the industrial revolution, fossil fuels powered progress but also pumped CO₂ into the atmosphere, worsening climate change. After half a century of fast industrialisation, China now puts ecological protection at the centre of its plans. By the end of 2025 the country had installed 2.34 billion kilowatts of renewable capacity – about 60 % of all its power plants – and roughly four out of every ten kilowatt‑hours used nationwide came from green sources. New renewable projects are expected to push total capacity to 3,500 GW by 2030 and 5,000 GW by 2035, lifting the share of non‑fossil electricity to around 65 % and cutting power‑sector carbon emissions by roughly 30 %. Experts compare this clean‑energy surge to the industrial boom of the 19th century, arguing that it could set a template for sustainable growth worldwide. The transition also creates millions of jobs in manufacturing, installation and maintenance of wind turbines and solar panels, and reduces air‑pollution‑related health costs for Chinese citizens. International observers see China’s model as a test case that could accelerate the global fight against climate change, encouraging other large economies to follow suit.
Read moreTwo of China’s biggest AI firms, Doubao and Qianwen, announced that they will shut down a range of chat‑bot services on July 15. The move follows new government rules – the “Artificial Intelligence Anthropomorphic Interaction Service Management Measures” – that aim to curb AI programs that act like human companions, offering emotional support, virtual romance or role‑playing. The regulators say these services can blur the line between real relationships and computer‑generated interactions, especially for teenagers who may lack the judgment to tell the difference. Critics worry that prolonged use could lead to emotional dependence or even addiction, while some parents welcomed the crackdown. The new measures do not affect AI tools used for work, education, research or customer service; they only target “continuous emotional interaction” that mimics a person’s personality, voice and appearance. Experts argue that this is a necessary “bottom line” for AI commercialization – a clear safety net that lets innovation flourish without harming society. They stress that AI should remain people‑oriented, creating new jobs and benefits while respecting human relationships. The shutdown is seen as a timely step to shape a responsible AI economy before harmful practices become entrenched.
Read more