Today:

• Grok 3 Impresses with AI-Powered Game Development and Simulations

• Anthropic Launches Claude 3.7

• Perplexity Introduces AI Browser Comet

• Apple Invests $500B in U.S.

• Alibaba Invests $52B in AI

• Microsoft Adjusts AI Investment Strategy

Grok 3 is out of control…

Grok 3 has demonstrated remarkable capabilities in AI-driven game creation and simulations, from VR shooters to 3D flight simulators. Unlike other AI models, it quickly integrates interactive elements, including missile mechanics and real-time physics. 

It excels in generating 3D environments, multiplayer features, and game logic with minimal input. While some UI elements need refinement, Grok 3 is proving to be a powerful tool for AI-assisted software development.

Anthropic Launches Claude 3.7

Anthropic has launched Claude 3.7, the first AI model with hybrid reasoning, allowing users to control its thinking depth. It features a scratchpad to reveal its reasoning process and excels in complex tasks like coding. Unlike rivals OpenAI and Google, which require switching models for reasoning, Claude 3.7 integrates both fast and deliberate thinking. Anthropic also introduced Claude Code, a tool for AI-assisted coding.

Perplexity Introduces AI Browser Comet

Perplexity AI is set to launch Comet, an AI-powered web browser aimed at reinventing the browsing experience. The company, known for its AI-driven search engine, announced the project on X, inviting users to join the waitlist. Perplexity has grown rapidly, recently securing $500 million in funding and introducing AI-driven shopping tools. The move signals a push to compete with Google and AI-powered search technologies.

Apple Invests $500B in U.S.

Apple plans to invest $500 billion in the U.S. over the next four years, including a new AI server factory in Texas with Foxconn and 20,000 new research jobs. The initiative aligns with efforts to strengthen domestic manufacturing amid tariff concerns. Apple will also expand its Advanced Manufacturing Fund to $10 billion and support semiconductor production at TSMC’s Arizona plant.

Alibaba Invests $52B in AI

Alibaba has committed $52.4 billion to artificial intelligence development, aiming to push the boundaries of Artificial General Intelligence (AGI). This marks a significant increase in spending compared to the past decade. The company, already a leader in China's AI race, recently partnered with Apple to develop AI-powered iPhones for the Chinese market. Alibaba’s stock has surged 68% in 2025, driven by AI-related advancements.

Microsoft Adjusts AI Investment Strategy

Microsoft remains committed to spending over $80 billion on AI infrastructure this fiscal year but may adjust its data center strategy. Analysts at TD Cowen suggested Microsoft canceled some data center leases, contributing to a 1.9% stock decline. Despite this, Microsoft asserts it has record-breaking growth and will continue expanding AI capacity to meet rising demand, supporting its Azure cloud and OpenAI partnerships.

🧠RESEARCH

LLM-Microscope: Uncovering the Hidden Role of Punctuation in Context Memory of Transformers

Punctuation and minor tokens like articles and commas play a crucial role in how Large Language Models (LLMs) store and process context. Removing them harms model performance. The study introduces LLM-Microscope, an open-source toolkit that analyzes token importance, nonlinearity, and contextual memory to reveal how models handle long-range understanding.

SurveyX: Academic Survey Automation via Large Language Models

SurveyX is an automated survey generation system using Large Language Models (LLMs). It improves content and citation quality by integrating online reference retrieval, an AttributeTree pre-processing method, and a re-polishing phase. Experiments show SurveyX outperforms existing systems, approaching human expert performance.

Mol-LLaMA: Towards General Understanding of Molecules in Large Molecular Language Model

Mol-LLaMA is a large molecular language model designed for a broader understanding of molecules beyond task-specific datasets. It uses multi-modal instruction tuning and integrates diverse molecular representations to enhance comprehension. Experiments show it effectively interprets molecular structures and provides detailed responses, making it a promising tool for molecular analysis.

SIFT: Grounding LLM Reasoning in Contexts via Stickers

SIFT is a post-training method that improves Large Language Model (LLM) reasoning by reducing context misinterpretation. It introduces Stickers, self-generated key information markers, to refine predictions. Tested across models from 3B to 100B+, SIFT enhances accuracy, setting a new benchmark in open-source AI.

StructFlowBench: A Structured Flow Benchmark for Multi-turn Instruction Following

StructFlowBench is a benchmark for evaluating how well Large Language Models (LLMs) follow multi-turn instructions, emphasizing structural dependencies between dialogue turns. It introduces six key inter-turn relationships for better assessment. Tests on 13 LLMs reveal significant gaps in understanding dialogue structures.

🛠️TOP TOOLS

Nubee AI Free - AI-powered platform that specializes in enhancing images and generating specific types of content of Not Safe for Work images.

AnonChatGPT - AI chat platform designed to provide users with a completely anonymous and private interaction experience.

FreeTTS - Designed to convert text into high-quality speech using advanced AI technology.

Melody ML - AI tool that revolutionize the process of audio track separation, making it an indispensable resource for music enthusiasts.

GPT-2 Output Detector - Designed to identify whether a piece of text has been generated by the OpenAI GPT-2 model.

📲SOCIAL MEDIA

🗞️MORE NEWS

• DeepSeek has launched FlashMLA, a high-performance decoding kernel for Hopper GPUs, boosting AI efficiency. It supports BF16, a 64-block paged KV cache, and 3000 GB/s bandwidth, optimizing variable-length sequences and reducing computational demands.

• QwQ-Max-Preview, built on Qwen2.5-Max, enhances deep reasoning, coding, and math capabilities. This preview precedes an open-source launch under Apache 2.0. Future plans include an AI app, smaller local models, and community-driven innovation.

• Apple is preparing to integrate Google Gemini into Apple Intelligence, expanding AI options for users. A recent iOS 18.4 beta leak confirms Apple's ongoing development, reinforcing its commitment to multiple AI model choices alongside OpenAI's ChatGPT.

• AI startup Anthropic is finalizing a $3.5 billion funding round, raising its valuation to $61.5 billion. Despite competition from China’s DeepSeek, investor confidence remains strong in its Claude chatbot and OpenAI-rivaling capabilities.

• Safe Superintelligence, an AI startup co-founded by Ilya Sutskever, raises over $1 billion at a $30 billion valuation, focusing on developing safe AI systems. The company is not generating revenue and plans to sell AI products in the future.

• Elon Musk’s xAIreleased Grok 3, driving a surge in app downloads and user activity. Daily active users jumped 260% in the U.S. and global visits grew 5x. However, recent censorship controversies may impact retention.