Tech Express | August 03, 2026 OpenAI's Next Model Cracks Fields Medal-Level Math Problems

1. OpenAI's Next-Generation Model Cracks 10 Fields Medal-Level Math Problems

OpenAI recently disclosed a landmark breakthrough in mathematical reasoning from its next-generation frontier model. The model successfully solved 10 math problems previously considered to be at the Fields Medal level of difficulty, spanning algebraic geometry, number theory, and topology—problems that previously only a handful of top mathematicians could tackle.

According to internal test results, the new model not only delivers correct answers but also demonstrates proof reasoning chains comparable to human mathematicians. Multiple mathematicians involved in the evaluation noted that some of the model's proof approaches offered genuinely novel problem-solving perspectives, suggesting AI is moving from "pattern matching" toward genuine "mathematical understanding."

The model is not yet publicly available, as OpenAI is conducting safety evaluations in partnership with several academic institutions. Industry analysts believe this advance could accelerate AI applications in scientific research, cryptography, and engineering, while also sparking renewed debate about the boundaries of AI's mathematical capabilities.

2. DeepSeek-V4 Series Receives Major Upgrade with Enhanced Agent Capabilities

DeepSeek announced a major upgrade to its V4 model series over the weekend. The new version matches flagship-level models in reasoning performance while achieving a qualitative leap in agent capabilities. The upgraded model can autonomously complete more complex multi-step tasks, including code writing and debugging, data analysis, and cross-tool orchestration.

This upgrade particularly strengthens the model's tool-use capabilities. Through an improved function-calling mechanism, DeepSeek-V4 can more precisely invoke APIs, operate databases, and excel in web browsing and file processing scenarios. Developers and enterprise users have responded enthusiastically, noting that this significantly lowers the barrier to building AI applications.

Notably, DeepSeek continues its open-source strategy, with the new model weights being made available to the community. Against the backdrop of "America open-sourcing AI to compete with China on cost-effectiveness," this upgrade is seen as a landmark event in the continued evolution of China's LLM ecosystem. Multiple British media outlets published articles on the same day focusing on China's AI progress.

3. US Accelerates Open-Source AI Efforts to Compete with China on Cost-Effectiveness

Facing the rapid development of Chinese AI companies in the open-source LLM space, the US tech sector is adjusting its strategy. Multiple media reports indicate that several US research institutions and tech companies are ramping up investment in open-source AI models, attempting to compete with Chinese products on cost-effectiveness. Chinese open-source models have already gained widespread recognition in the global developer community for their low cost and excellent performance.

The Observer reported that a former US official publicly stated that "America's advantage over China has evaporated," emphasizing the need for a new competitive approach. This reflects the urgency within US policy circles regarding China's rapid AI catch-up. The Financial Times and The Economist published articles on the same day focusing on the rise of China's AI ecosystem and its impact on the global tech landscape.

Industry analysts believe that AI competition is shifting from pure model capability to ecosystem building. The open-source versus closed-source debate and the trade-off between cost and performance will be core themes for the coming year. China's commitment to the open-source path is redefining the rules of global AI competition.

4. InP Shortage Could Be More Severe Than Memory Crisis: Optical Interconnects Become New AI Bottleneck

Lumentum and SemiAnalysis have jointly issued a warning: the shortage of indium phosphide (InP) could be more severe than the current memory shortage, becoming a new bottleneck constraining AI infrastructure expansion. InP is the core semiconductor material for manufacturing high-speed optical modules, and the supply gap is rapidly widening amid explosive demand for optical interconnects in AI data centers.

The report notes that each time an AI cluster doubles in size, the demand for optical interconnects grows more than linearly. Current global InP production capacity is concentrated among only a handful of suppliers, with expansion cycles lasting 18-24 months. This means that even if expansion begins immediately, tight supply conditions will persist until at least 2028.

This bottleneck, combined with the current global memory shortage, creates a "double whammy." Consumer electronics like the MacBook Air are already experiencing severe shortages due to the memory crunch, while the InP shortage will directly impact the pace of AI data center construction. Wall Street analysts believe the optical interconnect supply chain could become the most compelling investment theme of the second half of 2026.

5. MacBook Air Faces Unprecedented Global Shortage as Memory Crisis Hits Apple's Supply Chain

Apple's MacBook Air is experiencing an unprecedented global shortage. TechCrunch, MacRumors, and other tech media outlets report that Apple's best-selling laptop is widely out of stock across multiple countries and regions. Even with a $200 price increase, consumers still struggle to find available units.

Analysts point to the global memory (DRAM and NAND) supply crunch as the direct cause, but the deeper issue is that massive AI server expansion is cannibalizing memory capacity for consumer electronics. Wccftech reports that Apple is "prioritizing sales" of the current 14-inch MacBook Pro to clear inventory ahead of the M6 chip refresh later this year.

This supply crisis is testing Apple's supply chain management capabilities. Gizmodo commented that in an environment of weak consumer electronics demand, Apple's best-selling laptop is missing sales opportunities due to supply shortages—a "new normal" for the supply chain era.

6. Google Pixel 11 Series Complete Specifications Leak in Massive Dump

Months before the official launch of the Google Pixel 11 series, virtually all information about the flagship phones has been exposed. Android Authority obtained what is described as "the most detailed leak ever," covering complete specifications for the Pixel 11, Pixel 11 Pro, and Pixel 11 Pro Fold, including US and European pricing.

Leaked information reveals that the Pixel 11 Pro Fold introduces a mysterious new "camera bar" feature on its camera module. Engadget reports that this design could be a new type of sensor or auxiliary display. GSMArena also released fresh renders showing that the Pixel 11 series maintains the previous design language but features noticeably adjusted camera arrangements.

An exclusive report from Android Headlines also disclosed specific pricing: the Pixel 11 base model may maintain its predecessor's price point, while Pro versions will see a modest increase due to chip and camera upgrades. Google is expected to hold its Made by Google 2026 launch event this October, where it will officially unveil this generation of Pixel flagships.

7. Samsung Galaxy S27 Ultra Camera Design Leaks Amid Price Hike Rumors

The latest leaks about the Samsung Galaxy S27 Ultra point to "weird camera designs" and higher pricing. Notebookcheck reports that the S27 Ultra's rear camera module may adopt an unconventional arrangement with significant changes to the telephoto lens position—a design that appears rather bold in early renders.

PhoneArena further reveals that the Galaxy S27 series camera specifications are "almost set in stone," with the Ultra model retaining powerful zoom capabilities while the S27 Pro will feature a 12MP telephoto lens with an inferior zoom experience. Forbes' Android Circuit column lists it alongside the OnePlus 16 and Pixel 11 as the most noteworthy flagship phones of the second half of the year.

On pricing, multiple sources hint that the S27 Ultra will see another price increase, potentially breaking the $1,300 mark. Analysts believe rising chip costs and currency pressures are the main drivers. Samsung needs to find a balance between "spec stacking" and "price control" to compete with fierce competition from Chinese brands.

8. 3.2-Gigapixel LSST Camera Captures Over Half a Million Galaxies in First Image

The world's largest digital camera—the LSST (Legacy Survey of Space and Time) Camera—has just transmitted its first deep-space image. Captured by its 3.2-gigapixel sensor, the image covers the famous patch of sky known as the COSMOS field, capturing over 500,000 galaxies and 50,000 stars in a single frame.

The LSST Camera is mounted on the 8.4-meter Simonyi Survey Telescope at the Rubin Observatory in Chile. Its sensor comprises 189 individual CCD chips precisely arranged on a massive raft, with the entire camera weighing approximately 3,000 kilograms—roughly the size of a small car. Over the next decade, it will capture 700 ultra-high-resolution photos every night, creating the most ambitious timelapse of the universe in human history.

The COSMOS field has served as an astronomical "laboratory" since the Hubble Space Telescope focused on it from 2003 to 2005. The addition of the LSST Camera will track galactic evolution, supernova explosions, and other cosmic phenomena with unprecedented precision. Project scientists say this is just the beginning—repeated visits to the COSMOS field over the coming years will reveal more clues about dark matter and dark energy.

9. Global Nuclear Power Generation Hits Record High, Driven Almost Entirely by China

The latest data shows that global nuclear power generation reached its highest level in history in 2025, with nearly all of the increase coming from China. Citing an International Energy Agency report, The Observer notes that China is building nuclear power plants at an unprecedented pace, now ranking first globally in both operating and under-construction nuclear units.

China's rapid nuclear expansion is driven by two main forces: the strategic need to achieve its "dual carbon" goals, with nuclear energy receiving strong policy support as a stable clean baseload power source; and the enormous electricity demand from AI data centers, which is accelerating power infrastructure construction. Compared to wind and solar, nuclear power offers the advantage of stable output unaffected by weather conditions.

In contrast, nuclear development in Western countries has been relatively slow. Although some nations are reversing their post-Fukushima "denuclearization" trends, new project approval and construction timelines far exceed China's. Analysts note that if current trends continue, China could surpass the United States as the world's largest nuclear power producer within a decade.

10. Yang Zhilin Rejected Apple to Return to China and Found a Startup, Advisor Tells the Story

The PhD advisor of Moonshot AI founder Yang Zhilin recently publicly addressed the question of "why America couldn't keep Yang Zhilin." According to a report from MyDrivers, Yang resolutely turned down a job offer from Apple after completing his PhD, choosing instead to return to China and found Moonshot AI, which created the star domestic LLM product Kimi.

The advisor revealed that Yang demonstrated exceptionally strong independent research capabilities and keen judgment of technology trends during his academic years. His decision to return to China was not an impulsive one but based on a long-term belief in the potential of China's AI market. Today, Moonshot AI has become a major player in China's LLM arena, with Kimi receiving high praise for Chinese language understanding and long-text processing.

Yang's story is emblematic of the wave of Chinese AI talent returning home in recent years. An increasing number of top Chinese AI researchers are choosing to return to China for development, driving the flourishing of the domestic LLM ecosystem. This forms a mutually reinforcing cycle with the rise of homegrown models like DeepSeek and Tongyi Qianwen.