Google has announced Gemini 4 Argon, a new frontier model designed to sustain deep reasoning across complex and long-term workflows. This model delivers cutting-edge performance in real-world software engineering, enterprise knowledge work like legal and finance, and cybersecurity. It is being rolled out incrementally to trusted cybersecurity professionals through the Fairwind Program.

Argon is already transforming internal workflows at Google, with thousands of employees leveraging its strengths in specialized coding, deeper research, and high-quality writing. This has led to achievements such as optimizing quantum algorithms, improving memory efficiency in data centers (freeing over 300TiB), and facilitating large-scale code migrations from C/C++ to Rust.

The model's output token limit has been significantly expanded from 64,000 to an industry-leading 1 million tokens, providing unprecedented deep reasoning for tackling complex, long-term problems. Its coding capabilities achieved a new record of 77.9% on DeepSWE v1.1, and it leads in the Vals Index for measuring economic impact across finance, legal, and tax workflows.

In cybersecurity, Argon demonstrates exceptional ability to autonomously discover, verify, and patch critical software vulnerabilities. It is being utilized by Wiz in their "Scan for Good" initiative. The model achieved a top score of 68% on CWE-bench v1, showing a significant leap in vulnerability discovery over 3.8 Flash Cyber.

Ahead of general availability, Google is continuously strengthening its frontier safeguards across four key areas: defense against misuse, resilience to prompt injection attacks, monitoring for misalignment, and system hardening. Gemini 4 Argon is built to be a partner for developers, professionals, and enterprises in addressing their most challenging problems.

👉 Read the full article on Google Blog

This news reports on advancements in AI technology that can reconstruct visual information a person is looking at, based on brain scan data, particularly from fMRI (functional magnetic resonance imaging). Such "mind-reading" tools are enabled by machine learning models that learn complex relationships between brain activity patterns and visual stimuli.

The technology holds potential for deepening our understanding of how the brain processes information and forms visual experiences. Future applications could include assisting individuals with communication disorders or visualizing dreams and memories, though it also raises significant privacy and ethical concerns.

While the original article content was inaccessible, the title suggests a groundbreaking step in the frontier field of AI deciphering human cognition and perception.

👉 Read the full article on Technology Review (Original URL inaccessible, summary based on title and general knowledge)

This news reports that Flow Engineering, an AI startup, has secured significant funding from prominent venture capital firms Valor, Atreides, and Sequoia. This investment round has reportedly valued the company at $750 million.

While the original article content was inaccessible, this valuation and the caliber of investors suggest that Flow Engineering possesses highly promising technology or a business model within the AI sector. The participation of a top-tier VC like Sequoia is a strong indicator of its potential.

It is likely that Flow Engineering specializes in providing AI solutions for a specific industry or problem domain, with its technology being highly regarded in the market. This investment is expected to accelerate the company's product development, market expansion, and talent acquisition efforts.

👉 Read the full article on TechCrunch (Original URL inaccessible, summary based on title and general knowledge)