NerdNews

Google Workspace is adding more AI tools to its suite of productivity apps, including Docs, Sheets, Chat, and Meet. The new features, powered by Gemini, include audio summaries, writing coaches, and data analysis tools. Additionally, Google is introducing Workspace Flows, a new way to automate tasks across apps using AI agents. These features will be available to alpha testers in the coming weeks.

Anthropic introduces a new $100-200/month subscription tier called Claude Max, offering expanded usage limits for its Claude AI assistant. The new plan provides 5-20 times more usage than the $20 Pro plan and includes priority access to new features and models, higher output limits, and priority access during high traffic periods.

Google will add support for Anthropic's Model Context Protocol (MCP) to its Gemini models and SDK, allowing for two-way connections between data sources and AI-powered applications. MCP enables developers to build connections between data sources and AI models, and has been adopted by several companies including Block and Replit.

Google has unveiled Ironwood, its seventh-generation TPU AI accelerator chip, designed for inference and optimized for running AI models. The chip delivers 4,614 TFLOPs of computing power and has 192GB of dedicated RAM. It will be available to Google Cloud customers later this year in two configurations.

Google has introduced an Agent Development Kit (ADK) to simplify building business software with AI agents. The kit includes an 'agent garden' with pre-built bots and tools, as well as over 100 pre-built connectors to common data sources. Google also announced the seventh generation of its Tensor Processing Unit (TPU) and a new protocol, Agent2Agent (A2A), to enable agents from different vendors to communicate with each other.

Sam Newman discusses the importance of timeouts, retries, and idempotency in distributed systems, providing mechanisms to make them more robust. He emphasizes the need to understand how long things take, user expectations, and resource limitations. Newman also warns against exponential backoff and suggests using rate limiting and artificial network jitter to prevent self-inflicted DoS attacks.

At QCon London 2025, AWS Ambassador Jon Topper discussed common mistakes companies make when building SaaS solutions, including not architecting for multi-tenancy, not automating tenant provisioning, and delivering unique features to tenants. He advised teams to think about tenancy from the start, calculate baseline costs per tenant, and automate tenant provisioning early on.

QCon London 2025 discussed building resilient and consistent applications across multiple cloud providers, highlighting key challenges such as latency, resilience, event ordering, and security, with practical mitigation strategies and recommendations for standardization, optimization, and team upskilling.

The shadow table strategy creates a synchronized duplicate of data, enabling zero-downtime migrations. It supports diverse scenarios, including database migrations, microservices extractions, and incremental schema refactoring, and delivers stronger consistency and simpler recovery compared to dual-writes or blue-green deployments.

Google Cloud announced new tools and features to help organizations reduce costs and improve efficiency of AI workloads. The features focus on compute resource optimization, specialized hardware acceleration, and intelligent workload scheduling. Google Cloud recommends strategic platform selection, optimizing container performance, and storage selection to optimize AI workloads.

Google has launched Firebase Studio, a cloud-based AI-powered integrated development environment that allows users to build full-fledged apps using prompts. This move is seen as a challenge to Cursor AI, which has been gaining popularity in the 'vibe coding' trend. Firebase Studio supports multiple programming languages and frameworks, and allows users to create prototypes using natural language and images.

WordPress.com has launched a free AI-powered website builder that allows users to create a functioning website using an AI chat-style interface. The feature is available to WordPress users for free and is targeted at entrepreneurs, freelancers, and bloggers. The AI builder includes 30 free prompts before users need to choose a hosting plan.

Solve Intelligence, a startup using generative AI to speed up intellectual property and patent workflows, has raised $12 million in a Series A funding round led by 20VC. The company's AI-powered document editor helps attorneys with patent drafting, office action response, and claim charting. With 200 IP teams across the US, Europe, and Asia as customers, Solve Intelligence plans to use the funding to scale its product, hire staff, and expand its offerings.

Google's Gemini Code Assist AI coding assistant is gaining new 'agentic' capabilities, allowing it to deploy AI 'agents' that can take multiple steps to accomplish complex programming tasks, such as creating applications from product specifications and performing code transformations.

Google's 7th-gen Ironwood TPUs offer 24x the compute power of the world's fastest supercomputer, with 42.5 exaFLOPS of FP8 performance. Each TPU boasts 192 GB of high bandwidth memory and 4.6 petaFLOPS of dense FP8 performance. The TPUs are designed for large language model inferencing and feature Google's SparseCore for accelerating ultra-large embeddings.

Adobe is developing AI-powered agents for Photoshop and Premiere Pro to assist with complex tasks, such as editing photos and videos. The agents can suggest edits and carry out tasks automatically, helping users get started on projects. The technology will be introduced at Adobe's Max event in London on April 24th.

Google has announced a new unified security platform, Google Unified Security (GUS), which combines threat intelligence, security operations, and cloud security services. The platform also includes agentic AI and will be available to select customers in Q2. Google's acquisition of Wiz, a cloudy security upstart, for $32 billion has also given the company a multi-cloud security play. GUS is seen as a move by Google to compete with Microsoft in the enterprise security market, with analysts believing it can help the company grow its security revenue.

Nissan will integrate Wayve's self-driving software into its ADAS system, ProPilot, starting in 2027. The system will be a Level 2 system operating under driver supervision, with advanced collision avoidance capability. Wayve's software uses a self-learning approach, capturing data from existing sensors like cameras and radar to inform driving decisions.

OpenAI attorneys have called for Elon Musk to be enjoined from further unlawful and unfair action, accusing him of bad-faith tactics to slow down OpenAI and seize control of AI innovations. The lawsuit is a counter to Musk's suit against OpenAI, which accuses the startup of abandoning its nonprofit mission. The case has significant implications for the future of AI development and the governance of OpenAI.

Spammers used OpenAI's GPT to generate unique messages that bypassed spam filters, sending over 80,000 messages to websites in four months. The messages were customized to each site, making them difficult to filter. OpenAI revoked the spammers' account after being notified.