Salesforce upgrades Slack with AI integrations, embedding CRM tools and introducing AI agents like Slackbot and Channel Expert to enhance productivity and enable conversational access to data and third-party services.

Google released the Genkit extension for Gemini CLI, offering framework-aware AI assistance in the terminal to streamline development and debugging of Genkit-based apps. It supports flows, traces, documentation, and context-aware code generation.

AppliedAI launches Opus, a compliant AI automation platform for heavily regulated industries, enabling transparent, auditable workflows with human oversight and integration into enterprise systems.

Google launched Nano Banana (Gemini 2.5 Flash Image model), now integrated into Google Search and NotebookLM for image generation and editing via text prompts. Initially available in the Gemini app, it expands creative AI capabilities to wider platforms, supporting visual content creation, document enhancement, and video overviews with customizable styles. Rollout begins in the U.S. and India; security warnings issued in India over fake apps.

Broadcom's stock rose 9% after announcing a 4-year infrastructure deal with OpenAI to deploy 10 gigawatts of data center hardware, including custom AI chips and networking solutions, aiming to boost AI efficiency and reduce costs.

OpenAI research suggests LLM hallucinations arise from training methods that reward guessing over expressing uncertainty. Redesigning evaluation metrics to penalize confident errors and reward uncertainty could reduce hallucinations and improve AI reliability.

Grok, xAI's AI chatbot, will soon analyze internet videos for AI-generated signatures and trace their origins, addressing concerns over AI misuse for impersonation and defamation. The feature will detect AI traces like compression anomalies and verify authenticity via metadata and source trails. Grok, integrated into X (formerly Twitter), aims to enhance content verification and replace X's recommendation algorithm. Musk also plans a Wikipedia alternative called "Grokipedia" and has expanded Gro

Anthropic launches Claude Sonnet 4.5, a major upgrade in coding AI with enhanced agent tasks, long-duration performance, and safety. It shows improved behavior, reduced harmful tendencies, and better autonomous coding skills—scoring 77.2% on SWE-bench and 61.4% on OSWorld. Now available via API and apps at the same cost as before, it offers safer, more consistent performance with a 98.7% safety score. Early adopters report significant gains in planning, code testing, and error reduction.

OpenAI launched a new ChatGPT feature allowing developers to build apps within the platform, enabling users to interact via prompts for task execution and personalized responses. Spotify is among the first to integrate, offering playlist creation, song recommendations, and playback controls based on user preferences and listening history. Users can link their Spotify account, granting ChatGPT access to data like saved songs and location, while retaining the ability to disconnect at any time. The

Samsung researchers developed the Tiny Recurrent Model (TRM), a compact AI with 7M parameters, outperforming larger models like Gemini 2.5 Pro in structured reasoning tasks such as Sudoku and maze navigation. Using recursive reasoning, TRM iteratively refines solutions, achieving high accuracy with minimal resources.

Qualcomm acquires Arduino to boost AI and IoT engagement with developers, expanding its reach in edge computing and robotics.

Microsoft and NVIDIA deploy Azure with NVIDIA’s GB300 NVL72, a large-scale cluster designed for advanced AI inference, featuring 72 Blackwell Ultra GPUs, 36 Grace CPUs, and 800Gb/s networking, enabling high-performance, low-latency processing for large models and extended reasoning tasks.

Time magazine's 2025 Best Inventions list includes Unitree R1 in robotics, alongside Figure 03, EufyMake E1 UV printer, and HoverAir X1 ProMax.

Figure AI unveils Figure 03, a versatile humanoid robot designed for household tasks and industrial use, featuring advanced autonomous capabilities via its proprietary Helix neural system. Standing 1.68m tall, it operates up to 5 hours per charge, with embedded palm cameras for real-time object manipulation. With $1B funding and a $39B valuation, backed by NVIDIA, OpenAI, Microsoft, and Bezos, Figure AI competes globally, predicting a $40T labor market shift toward robotics within a decade.

OpenAI's Sora tops US App Store after launch, surpassing ChatGPT's initial download pace with 627k iOS installs in first week, despite invite-only access.

Google launched Gemini Enterprise, a unified AI platform integrating advanced models, chatbots, and intelligent agents. It offers a centralized, secure interface for enterprise AI interactions, connecting with data sources like Google Workspace, Microsoft 365, and apps from Salesforce and SAP. The platform includes 1,500+ pre-built agents and tools for creating custom workflows, enabling automation of complex tasks across departments. It supports AI-driven video creation, real-time voice transla

Figma integrates Google's Gemini 2.5 Flash, Gemini 2.0, and Imagen 4 to boost image generation and editing speed by up to 50%, while maintaining enterprise governance and brand compliance.

Intel's Panther Lake, first Core Ultra series 3 chip based on 18A process with RibbonFET and PowerVia, starts mass production in Arizona, targeting AI PCs with up to 180 TOPS. Fab 52's 18A output marks U.S. manufacturing revival. Cougar Cove P-cores and Darkmont E-cores support up to 16 CPU cores and 12 Xe3 graphics cores, improving efficiency and GPU performance. 18A's success hinges on stable yield ahead of 2026 notebook production. TSMC’s N2 node enters volume production in late 2025–early 20

Amazon launches Quick Suite, an AI assistant for enterprises that automates tasks, connects data sources, and integrates with productivity tools like Microsoft 365 and Slack, enabling users to build custom AI agents and streamline workflows.

Google launches "Search Live" in India, enabling real-time, multimodal search via voice and camera. Expands AI search to 8 Indian languages, powered by Gemini model. Over 450M monthly users for Gemini globally. Faces FTC probe on Gemini safety and criticism over AI Overviews' impact on web traffic.