Last Updated on April 16, 2025 by Editor
Missed AI Week # 15? You Just Skipped One of the Most Groundbreaking Weeks in 2025 AI History. From billion-dollar funding rounds to futuristic AI-powered tools, April 9–15, 2025, delivered a wave of high-impact innovations, strategic investments, and industry-shaping moves across the globe. OpenAI unveiled GPT-4.1, taking multimodal AI to the next level, while Apple and Samsung dropped AI features that will redefine everyday user interaction. Startups like Artisan AI, RLWRLD, and Ominimo bagged major funding, proving AI’s explosive growth trajectory is far from slowing. Meanwhile, YouTube launched a free AI music creator, WordPress introduced a no-code AI website builder, and Notion reimagined your inbox with AI. Even governments and tech giants stepped in—Meta is eyeing European user data, Google Classroom added quiz-generating AI, and NVIDIA committed to making AI chips in the U.S. If you’re serious about keeping up with the future of AI, this roundup is your goldmine of updates, insights, and tech breakthroughs—and missing it would mean falling behind.
Nvidia Shifts AI Supercomputer Production Fully to US Soil
Nvidia has announced a significant strategic shift, planning to manufacture its AI supercomputers entirely within the United States for the first time. This move aims to bolster supply chain resilience and meet the surging global demand for AI infrastructure, partly driven by global trade dynamics and a desire for greater national security oversight. The initiative involves major partnerships with TSMC for chip production in Arizona, and Foxconn and Wistron for supercomputer assembly in new Texas facilities.
This ambitious plan aims to produce up to half a trillion dollars ($500bn) worth of AI infrastructure domestically within the next four years, leveraging additional partnerships with Amkor and SPIL for semiconductor packaging and testing in Arizona. Initial production of the Nvidia Blackwell chips, the core of these new systems, has already begun at TSMC’s Arizona plants. The full ramp-up of the Texas supercomputer facilities is expected within the next 12 to 15 months, marking a major step towards creating ‘AI factories’ in the US and projecting the creation of hundreds of thousands of jobs.
xAI’s Grok Introduces ‘Studio’ for Collaborative AI Creation
Elon Musk’s AI venture, xAI, has enhanced its chatbot, Grok, with a new feature called Grok Studio. This addition provides a canvas-like workspace alongside the chat interface, enabling users to collaboratively create and edit documents, reports, code, and even simple applications or browser games with the AI. Grok Studio aims to streamline workflows where generative AI is used for content creation and basic development tasks.
The Studio is accessible to both free and paid Grok users via Grok.com. Key initial features include code execution capabilities (supporting Python, JavaScript, C++, HTML) directly within the interface and integration with Google Drive. This allows Grok to access and assist users in modifying documents, spreadsheets, and presentations stored in their Drive, positioning Grok Studio as a potentially valuable tool for integrating AI more deeply into daily productivity tasks, following trends set by competitors.
AI Voice Agent Startup Telli Secures $3.6M in Pre-Seed Funding
Telli, a Berlin-based startup and Y Combinator (YC) alum, has successfully raised $3.6 million in pre-seed funding. The round was led by Cherry Ventures and included participation from Y Combinator and other angel investors. Telli develops AI-powered voice agents designed to help businesses manage high volumes of customer calls for tasks like appointment booking, lead qualification, and even completing sales transactions, thereby reducing customer service bottlenecks.
Founded by Seb Hapte-Selassie, Philipp Baumanns, and Finn zur Mühlen, Telli utilizes AI voice cloning from professional voice actors (using platforms like 11 Labs or Cartesian) to create highly convincing agents. The funds will be used to accelerate growth, expand into new markets (currently live in Germany, UK, Latin America, US), and further develop its AI technology. Telli serves industries such as real estate, energy, and healthcare, offering multilingual support and seamless integration with human agent workflows.
Microsoft Integrates AI Editing and Creation Tools into Paint & Notepad
Microsoft is embedding AI capabilities directly into its classic Windows applications, Paint and Notepad, transforming them into more powerful creative and productivity tools. In Paint, users can now access AI features like Image Creator (generating images from text prompts), Generative Erase (removing unwanted objects seamlessly), and Remove Background, leveraging AI to simplify image creation and editing tasks.
In Notepad, AI integration focuses on text manipulation. Users can select text and use Copilot features to rewrite sections, adjust length (‘Make shorter’/’Make longer’), or change the tone (e.g., formal, casual, humorous). While basic access might be free, extensive use of these new AI features in both Paint and Notepad will require a Microsoft 365 or Copilot Pro subscription, reflecting a ‘freemium’ model for deploying advanced AI functionalities in familiar applications. Users retain the option to disable these features.
Google Rolls Out Veo 2 Video Generation within Gemini
Google has integrated its advanced text-to-video AI model, Veo 2, into its Gemini Advanced platform. Subscribers ($20/month) can now generate eight-second video clips at 720p resolution (16:9 MP4 format) directly from text prompts within the Gemini interface on web and mobile. Veo 2 is designed to produce high-resolution, detailed videos with enhanced understanding of physics and motion, resulting in more cinematic realism and fluid character movement.
The feature is rolling out globally in all languages supported by Gemini. Users are encouraged to provide detailed prompts for better control over the output. Generated videos can be easily shared on platforms like TikTok and YouTube Shorts. Additionally, Veo 2 powers Whisk Animate in Google Labs, allowing Google One AI Premium subscribers to animate static images (created with Whisk) into eight-second videos. All Veo 2-generated videos are watermarked with SynthID to indicate they are AI-generated. Developers can also access Veo 2 via the Gemini API and Google AI Studio.
OpenAI Unveils GPT-4.1 Models Tailored for Developers via API
OpenAI has launched GPT-4.1, a new family of AI models available exclusively through its API, designed with significant improvements for developers. This family includes GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano, all boasting enhanced performance over previous models like GPT-4o, particularly in coding, instruction following, and long-context comprehension. These models feature a knowledge cutoff updated to June 2024 and support context windows up to 1 million tokens, enabling work with larger codebases and more complex tasks.
The development of GPT-4.1 incorporated direct feedback from developers, focusing on improving real-world utility for tasks like frontend coding, maintaining format consistency, and reliable tool usage. While GPT-4.1 offers top-tier performance, GPT-4.1 mini provides a leap in small model capability (even beating GPT-4o on some benchmarks) at lower cost and latency. GPT-4.1 nano is OpenAI’s fastest and cheapest model, optimized for tasks needing rapid responses like classification or autocompletion. These API-only models represent a step towards OpenAI’s goal of an “agentic software engineer.”
Notion Launches AI-Enhanced Email Client ‘Notion Mail’
Notion has officially released Notion Mail, an AI-driven email client aiming to reinvent the email experience by integrating it deeply with Notion’s workspace philosophy. Initially available on macOS (with iOS and Android versions planned), Notion Mail connects to existing Gmail accounts, acting as an intelligent front-end rather than a replacement service. It focuses on user customization and AI-powered organization, allowing users to configure their inbox in highly personalized ways, much like Notion pages and databases.
Key features include AI-driven auto-labeling to sort incoming emails based on user-defined importance, custom views to split the inbox for focused work, and reusable email templates (‘snippets’). It integrates seamlessly with Notion Calendar for instant meeting scheduling within emails (using ‘/schedule’). The familiar Notion / command interface is present for formatting and invoking AI assistance, enabling users to draft replies using context from Notion docs or polish their writing. Core AI features have free usage limits, with unlimited access requiring a paid Notion plan.
Apple Outlines Privacy-Preserving Strategy for AI Model Training
Apple has detailed its approach to improving Apple Intelligence features using user data while prioritizing privacy. The strategy relies heavily on differential privacy and synthetic data, ensuring actual user content never leaves the device or is seen by Apple. To enhance features like email summarization, Apple generates synthetic emails mimicking real-world characteristics (topic, length, style) without using actual user content. Embeddings (numerical summaries) of this synthetic data are sent to a small number of devices whose users have opted into sharing analytics.
These opted-in devices compare the synthetic embeddings to a small, locally stored sample of the user’s recent emails. Using differential privacy techniques, Apple learns which types of synthetic messages are most representative across the user base without identifying individual user data or specific emails. This aggregated feedback helps refine the synthetic data used for training models, improving their performance on tasks like summarization. Apple plans to apply this privacy-focused method, starting with email summaries in upcoming OS betas (iOS 18.5, etc.), and later expand it to tools like Image Playground, Writing Tools, and Genmoji prompt suggestions.
Xaba Secures $6M Seed Extension Led by Hitachi Ventures for Robot ‘Brains’
Toronto-based Xaba, an industrial AI company, has raised a $6 million Seed extension funding round led by Hitachi Ventures. Participants included Hazelview Ventures, BDC Capital, Exposition Ventures, and Impact Venture Capital. Xaba specializes in building “synthetic brains” for industrial robots and cobots, aiming to make automation more intelligent, flexible, and efficient through its no-code AI platform, xCognition.
Xaba’s technology empowers robots with AI-driven cognition and awareness, allowing them to autonomously generate programs and execute complex manufacturing tasks like welding, drilling, assembly, and additive manufacturing with greater precision and adaptability. The company states this approach significantly reduces the high costs associated with traditional robot programming and deployment (estimated at $7B annually for programming alone). The new funding will accelerate the deployment of Xaba’s AI solutions in sectors like aerospace and automotive manufacturing, enabling robots to handle variations and adjustments in real-time, reducing rework and downtime.
RLWRLD Raises $14M+ Seed Round for Robotics Foundation Model
South Korean AI startup RLWRLD (Real Life World) has emerged from stealth, announcing a seed funding round of approximately $14.4 – $15 million (reported figures vary slightly, often cited as 2.1 billion KRW or ~$14.8M). The round was led by Hashed, with participation from Mirae Asset Venture Investment, Global Brain, and strategic investors including LG Electronics, SK Telecom, Ana Group, PKSHA, Mitsui Chemicals, Shimadzu, KDDI, and Amber Manufacturing.
RLWRLD is building a purpose-built foundational AI model specifically for robotics, merging large language models (LLMs) with traditional robotics software to enhance robot agility and “logical reasoning.” Led by CEO Jung Hee Ryu (whose previous startup was acquired by Intel), the company aims to automate complex workflows requiring manual dexterity. The funding will be used for proof-of-concept projects with investors, expanding computational infrastructure (GPUs), acquiring data, recruiting talent, and advancing research, including developing advanced five-finger hand manipulation capabilities for humanoid robots.
OpenAI to Sunset GPT-4.5 Preview in API, Directing Users to GPT-4.1
OpenAI has announced plans to phase out access to its GPT-4.5 Preview model via its API, setting a deadline of July 14, 2025, for developers to transition. Launched in late February 2025, GPT-4.5 (codenamed Orion) was positioned as OpenAI’s largest and most capable model, trained with extensive data and compute resources. Despite its strengths in areas like conversational ability, it was noted for its high operational costs ($75 per million input tokens, $150 per million output tokens).
The company is encouraging developers to migrate to the newly released GPT-4.1 model family (GPT-4.1, mini, nano), which launched concurrently with the phase-out announcement. OpenAI states that GPT-4.1 offers comparable or superior performance, particularly in coding and instruction following, at a significantly lower cost and greater efficiency. While GPT-4.5 will remain available as a “research preview” within the paid ChatGPT interface, its removal from the API signals OpenAI’s strategic focus on balancing cutting-edge performance with practical resource consumption and cost-effectiveness for developers.
Google Classroom Introduces AI-Powered Quiz Generation for Educators
Google is enhancing Google Classroom with a new AI-powered feature designed to help teachers create quizzes more efficiently. Leveraging Google’s Gemini AI models, this tool allows educators to automatically generate text-dependent questions based on provided source material, such as textbook chapters, articles, or class notes. Teachers can input text and then customize the generated questions using filters.
The available filters allow teachers to specify the grade level, number of questions, question types (e.g., multiple-choice, open-ended), and specific skills to assess (like analyzing figurative language or evaluating arguments). After the AI generates the questions, teachers can review, edit, and then directly export them into Google Docs or Google Forms for use in assignments or assessments. This feature is rolling out to users with Google Workspace for Education Plus subscriptions, aiming to save teachers time and help create targeted, skill-based evaluations.
Meta to Utilize Public EU User Content for AI Model Training
Meta Platforms has confirmed it will begin using publicly shared content from adult users in the European Union (EU) and UK to train its AI models. This includes public posts and comments on platforms like Facebook and Instagram, as well as user interactions with Meta’s AI tools (like prompts and questions). The company explicitly states that private messages and data from users under 18 will not be used for this training.
EU users are being notified about this data usage and provided with an objection form if they wish to opt-out of having their public data used for AI training, aligning with GDPR principles. This move follows the delayed rollout of Meta AI features in Europe due to initial regulatory concerns. Meta argues this step is necessary to improve its AI’s understanding of European languages, cultures, and nuances, noting that competitors already use similar public data sources. This decision comes amid ongoing scrutiny of AI data practices by EU regulators, including investigations into X (formerly Twitter) and Google.
Rumors Swirl Around Ex-OpenAI CTO Mira Murati’s Startup Aiming for $2B Seed Round
Reports surfaced indicating that Mira Murati, the former Chief Technology Officer of OpenAI, is seeking an unprecedented $2 billion seed funding round for her new stealth AI venture, Thinking Machines Lab. This figure reportedly doubled from an initial $1 billion target just months prior, potentially valuing the nascent company at $10 billion or more despite having no current product or revenue. The extraordinary funding goal highlights intense investor enthusiasm for generative AI and confidence in Murati’s leadership.
Murati, who briefly served as OpenAI’s interim CEO, has assembled an elite team, reportedly recruiting around 30 top researchers and engineers from OpenAI, Meta, Google DeepMind, and Mistral, including key figures behind ChatGPT like John Schulman (now Chief Scientist) and Barret Zoph (now CTO). While details are scarce, Murati aims to build AI systems that are “more widely understood, customisable, and generally capable,” emphasizing user alignment and potentially open science practices. If successful, this would be one of the largest seed rounds in history, setting a new benchmark for early-stage AI investment.
Canva Expands Creative Suite with AI Assistant, Sheets, and Coding Tools
Canva has significantly upgraded its platform, launching Visual Suite 2.0 which incorporates a suite of new AI-powered features aimed at streamlining design and content creation. Key additions include an AI assistant accessible via text or voice commands to help generate designs, write content, and brainstorm ideas. The platform now also features Canva Sheets, integrating data capabilities with design through tools like Magic Insights for AI data analysis and Magic Charts for visualization, aiming to make data presentation easier and more on-brand.
Furthermore, Canva introduced Canva Code, enabling users to create interactive design elements, widgets, or even simple games without needing coding expertise. Enhancements to existing AI tools like Magic Write (for generating copy variations and translations) and bulk background removal/resizing aim to boost productivity. These updates reflect Canva’s push to integrate generative AI deeply into the creative workflow, making complex design and data tasks more accessible to its broad user base, from individuals to enterprise teams.
YouTube Empowers Creators with Free AI Music Generation Tool
YouTube has rolled out ‘Music Assistant,’ a new AI-powered tool integrated into its Creator Music marketplace, designed to help video creators generate custom, royalty-free instrumental background music. Initially available to US-based members of the YouTube Partner Program, this feature allows creators to use text prompts to describe the desired mood, genre, instruments, or video context (e.g., “calm acoustic guitar music for a tutorial”). The AI then generates multiple track options.
This tool provides creators with a free and convenient way to source unique background music without navigating complex licensing or fearing copyright claims. The generated tracks can be downloaded and used directly in their videos. Music Assistant builds upon YouTube’s earlier AI music experiments like ‘Dream Track’ but focuses specifically on creating longer instrumental pieces suitable for various video formats, simplifying a common pain point for content producers.
Incident Management Platform Incident.io Raises $62M at $400M Valuation
Incident.io, a UK-based startup providing an all-in-one incident management platform for IT and engineering teams, has secured $62 million in Series B funding, bringing its total funding to over $96 million. The round was led by Insight Partners, with continued participation from existing investors Index Ventures and Point Nine Capital. This investment values the company at $400 million.
Founded in 2021, Incident.io helps teams quickly coordinate, communicate, diagnose, and learn from software outages and system failures, integrating with tools like Slack and GitHub. The company plans to use the new capital to significantly scale its engineering teams in London and San Francisco, enhance its global go-to-market strategy, and further develop its platform, including building AI agents designed to assist in resolving incidents alongside human teams, addressing the increasing complexity of modern software systems.
AI Insurtech Ominimo Secures €10M Investment from Zurich at $220M Valuation
Ominimo, a Poland/Hungary-based AI insurtech startup specializing in auto insurance, has landed its first major investment of €10 million (approx. $11 million) from Zurich Insurance Group. This deal values the one-year-old startup at €200 million (approx. $220 million), with Zurich reportedly taking a ~5% stake. Ominimo uses AI to generate auto insurance quotes rapidly (within a minute) based on minimal user input (just five questions), augmented by hundreds of other data points derived from sources like license plate numbers.
The company claims its AI-driven granular risk assessment, which analyzes far more variables than traditional methods, allows it to offer more accurate and affordable premiums (up to 30% lower than average in Hungary, where it quickly gained a 7% market share). The funding, backed by a distribution partnership with Zurich’s German unit DA Direkt, will fuel Ominimo’s expansion into over 10 new European markets and the development of property insurance products, with Zurich underwriting the policies.
xAI Releases Grok 3 API, Offering Advanced Reasoning and Coding Capabilities
Elon Musk’s AI company, xAI, has launched an API for its Grok 3 family of models, making its advanced AI capabilities accessible to developers for building applications. The API provides access to Grok 3 beta (the flagship model) and Grok 3 mini beta (a lightweight version), along with faster variants of each. These models boast strong performance in coding, data extraction, text summarization, and complex reasoning, with Grok 3 having deep domain knowledge in fields like finance and science.
Key features include a large 131,072-token context window and the ability to accept image inputs (though they don’t generate images). Unlike the chatbot version, the API models are not connected to the real-time internet and rely on pre-training data (knowledge cutoff November 2024). Pricing varies significantly: Grok 3 beta costs $3/million input tokens and $15/million output tokens, while the mini version is considerably cheaper at $0.30/million input and $0.50/million output tokens. Faster versions carry a premium cost for reduced latency.
Artisan AI Secures $25M Series A for Autonomous Sales Agents
Artisan AI, a startup developing autonomous AI agents designed to function as full team members, has raised a $25 million Series A funding round. This investment, led by Glade Brook Capital (as reported in other searches, specific investors might vary across sources), aims to accelerate the development and deployment of its AI workforce, starting with sales roles. Artisan’s first agent, ‘Ava,’ operates as an AI Business Development Representative (BDR), autonomously handling tasks like lead discovery, research, hyper-personalized outreach via email and LinkedIn, and meeting scheduling.
The platform utilizes a multi-agent system and a real-time context engine that scans the web for buying signals to optimize outreach timing and messaging. Despite the goal of creating AI “employees” to replace repetitive work, the company emphasizes it is actively hiring human talent to build these sophisticated systems. Future plans include launching ‘Aaron’ (an inbound SDR) and ‘Aria’ (a meeting assistant), moving towards an end-to-end AI sales system and eventually expanding into other business functions.
WordPress.com Debuts Free AI Tool for Instant Website Creation
WordPress.com has launched a new free AI-powered website builder, designed to dramatically simplify the process of getting a site online. Users can simply describe their website idea in a natural language prompt, and the AI generates a complete, functional WordPress website—including layout, text, images, and relevant pages—within minutes. This tool aims to eliminate the traditional hurdles of theme selection, manual content placement, and basic design work.
The builder integrates an AI chatbot for conversational editing, allowing users to refine the generated site through further prompts or make manual adjustments using the standard WordPress editor. The free offering includes 30 AI prompts, sufficient for creating and tweaking a basic site. While currently best suited for blogs, portfolios, and simpler business sites (lacking advanced e-commerce features initially), this tool makes website creation accessible to users without coding or design experience. Upgrading to a paid WordPress.com hosting plan unlocks unlimited prompts and additional features.
Tessell Lands $60M Series B to Advance Multi-Cloud DBaaS Platform
Tessell, a startup offering a high-performance Database-as-a-Service (DBaaS) platform for multi-cloud environments, has secured $60 million in Series B funding. The round was led by WestBridge Capital, with participation from B37 Ventures, Rocketship.vc, and existing investor Lightspeed Venture Partners, bringing Tessell’s total funding to $94 million. Tessell aims to modernize enterprise data management by providing a unified platform for running operational and analytical databases across different cloud providers.
The platform supports various database engines and provides features like automated provisioning and scaling, AI-driven performance tuning, integrated disaster recovery, data replication, and robust governance tools. Tessell plans to use the investment to accelerate its go-to-market strategy, expand its global presence, and further enhance its AI capabilities, aiming to simplify complex data infrastructure management and make enterprise-grade database operations more efficient and cost-effective.
Solve Intelligence Raises $12M Series A with Microsoft, Thomson Reuters Backing
Solve Intelligence, a legal tech startup focused on applying AI to the patent process, has closed a $12 million Series A funding round. The round saw strategic investments from Microsoft’s M12 venture fund and Thomson Reuters Ventures (TRV), alongside 20VC and follow-on participation from seed investor Y Combinator. Solve Intelligence aims to build the definitive AI platform for inventors, in-house counsel, and law firms involved in patent work.
The company’s current offerings focus on AI-assisted patent application drafting and AI-powered responses to office actions during patent prosecution. The Series A funds will be used to scale research and development, expanding the platform’s capabilities into AI-driven freedom-to-operate analysis, claim chart generation, patent licensing, litigation support, and portfolio analysis, with an initial focus on the life sciences sector. The strategic backing provides potential access to secure Azure infrastructure and integrations with established legal tools like Westlaw and CoCounsel.
Samsung Equips Ballie Home Robot with Google Gemini AI Capabilities
Samsung is enhancing its Ballie home companion robot by integrating Google’s Gemini AI. This partnership, announced ahead of Ballie’s anticipated summer 2025 launch (initially in the US and South Korea), aims to enable more natural, conversational interactions and significantly expand the rolling robot’s helpfulness around the house. Ballie utilizes Gemini’s multimodal reasoning capabilities combined with Samsung’s own language models and the robot’s onboard camera and sensors.
This integration allows Ballie to process and understand voice, audio, and visual inputs to interact dynamically with its environment and users. Potential capabilities include managing smart home devices, greeting guests, offering personalized advice based on user input (e.g., “I feel tired today”) or visual assessment (e.g., “How do I look?”), setting reminders, and answering queries. This collaboration signifies a deeper integration of generative AI into Samsung’s device ecosystem, moving beyond smartphones into home robotics. From game-changing funding rounds to cutting-edge AI tools, this week proved once again that artificial intelligence is reshaping our world at lightning speed. Whether you’re a developer, entrepreneur, or tech enthusiast, these developments demand attention—because the future won’t wait. Want to stay in the loop? Tune in next week for AI Week #16, where we’ll break down the latest breakthroughs before they go mainstream. The AI race is on—don’t get left behind.
