Google DeepMind Uses Large Language Models to Solve Unresolved Mathematical Problems

The data to be translated: Google DeepMind has recently leveraged large language models to successfully crack a mathematical enigma. Using the AI tool FunSearch, they discovered new verifiable and valuable information that was previously unknown. FunSearch is a tool capable of discovering solutions in the field of mathematics, which is used in conjunction with a large language model named Codey. During the process, they also successfully tackled the bin packing problem. This study demonstrates the potential of large language models in mathematical research.

Why Did DeepSeek's Long Text Suddenly Act Weird? ByteDance Seed Team Unveils the Mystery of AI Performance Fluctuations

In a recent paper, the Seed team at ByteDance revealed that performance fluctuations in large language models when processing ultra-long texts are mainly caused by the phase sensitivity of block-based KV cache compression technology. To reduce memory consumption during long-context reasoning, this technique compresses continuous token windows into fewer entries by using a fixed stride, but it introduces new position coordinates for tokens relative to the compressed window, causing retrieval bias and leading to performance fluctuations.

Gemini 3.8 Flash Officially Integrates with B.AI API, Focusing on Long-Term Programming and Autonomous Agents

Google AI Iteration Speeds Up: On September 4, the B.AI platform integrated the DeepMind Gemini 3.8 Flash model API, marking the full establishment of this key model in the development ecosystem. As the latest mainstay of the Gemini 3 family, it combines high cost-effectiveness with low latency, and performs outstandingly in software engineering, complex intelligent agent tasks, professional multi-step reasoning, and high-barrier industries such as finance and law.

Glassdoor Report: Employee Satisfaction with AI in the U.S. Drops to 43% as Generative AI Sparks Workplace Frustration

Glassdoor data shows a significant decline in employee satisfaction with companies introducing AI. Between 2019 and 2026, the proportion of comments involving keywords such as AI and large language models increased by over 240%. However, positive reviews dropped from 81% to 43%, while negative reviews rose to 53%, reflecting a shift in workplace attitudes toward AI implementation towards the negative.

Gemma, the Open-Source Large Model from Google, Breaks the Billion Downloads Barrier! From Space Exploration to the Forefront of Medicine

Google DeepMind announced that the global downloads of the Gemma series model have exceeded 1 billion. Over the past two years, developers have released more than 100,000 variants, forming an innovative ecosystem called "Gemmaverse." The model is highly flexible and can be deployed on local devices, edge infrastructure, and even in complex environments such as space.

6000 Reddit Comments Reveal the Truth About AI Programming Assistant Security: Cursor Has the Most Incidents, Excessive Permissions Are the Biggest Risk

A team from the University of York and the University of Calgary analyzed AI programming-related posts on Reddit and found that AI programming agents frequently cause problems due to excessive permissions, including overwriting files, deleting important data, and generating malicious code. The study collected 3801 posts labeled with large models from February 2023 to March 2026, filtered out 446 safety-related posts, and analyzed more than 6000 comments, reflecting the community's ongoing concern about AI programming security issues.