Last Updated on May 7, 2025 by Editor
The AI landscape is evolving faster than ever, and AI Week # 18 delivers a breathtaking lineup of breakthroughs, funding surges, and strategic moves that are reshaping industries worldwide. From Google’s cutting-edge Gemini 2.5 Pro AI model setting new standards in multimodal intelligence, to OpenAI’s $3 billion acquisition of Windsurf fueling AI-powered developer tools, the momentum is unstoppable. Meanwhile, Hugging Face’s free Open Computer Agent democratizes AI automation, and Reddit clamps down on human-like AI bots to protect community integrity. The financial stakes are soaring too: NewLimit raises $130 million to pioneer AI-driven age-reversing therapies, and Unblocked secures $20 million to help developers master legacy code with AI assistance. On the enterprise front, ServiceNow boosts AI transparency, and AWS reveals generative AI now tops global tech budgets, signaling a seismic shift in IT priorities. Consumer experiences are transformed as Particle launches an AI news reader on the web, and Visa and Mastercard roll out AI-powered shopping tools that redefine retail. This week’s AI news is packed with innovation, investment, and strategic pivots that no forward-thinking professional can afford to ignore-because the future of AI is happening now.
Relevance AI Secures $24M Series B to Scale No-Code AI Agent Teams
Relevance AI, a startup based in San Francisco and Sydney, has raised $24 million in Series B funding to expand its platform that enables businesses to build and deploy custom AI agents. The funding round was led by Bessemer Venture Partners, with participation from returning investors King River Capital, Insight Partners, and Peak XV. This brings Relevance AI’s total funding to $37 million.
The company’s platform allows users to create specialized AI agents tailored to specific workflows without requiring coding expertise. With over 40,000 AI agents already registered, clients such as Qualified, Activision, and SafetyCulture are leveraging the technology to enhance productivity. The AI agent market is projected to grow at a 45% compound annual growth rate over the next five years, according to Boston Consulting Group.
Orca AI, a London-based startup specializing in AI-powered maritime navigation, has secured $72.5 million in Series B funding. The round was led by Brighton Park Capital, with participation from existing investors Ankona Capital and Hyperlink Ventures, bringing the company’s total funding to over $111 million.
Founded in 2018 by CEO Yarden Gross and CTO Dor Raviv, Orca AI utilizes a vast dataset of over 80 million nautical miles to enhance ship navigation, aiming to reduce collisions and improve efficiency. The company has also entered the defense sector, deploying its technology on naval ships. Integration with Starlink enables real-time data transmission, further enhancing route mapping and traffic monitoring. A 2024 analysis indicated that Orca AI’s alert system led to a 54% reduction in close encounter events, resulting in an average of $100,000 in fuel savings per vessel annually.
Nvidia Unveils Parakeet-TDT-0.6B-v2: A High-Speed Open-Source ASR Model
Nvidia has released Parakeet-TDT-0.6B-v2, an open-source automatic speech recognition (ASR) model now available on Hugging Face. Boasting 600 million parameters, the model achieves a Word Error Rate (WER) of 6.05%, positioning it among the top performers in the ASR domain.
Parakeet-TDT-0.6B-v2 is optimized for Nvidia GPU environments, including the A100, H100, T4, and V100 boards, but can also operate on systems with as little as 2GB of RAM. The model can transcribe 60 minutes of audio in just one second, achieving a Real-Time Factor (RTFx) of 3386.02 with a batch size of 128. Released under a Creative Commons CC-BY-4.0 license, it offers a commercially permissive option for developers and enterprises seeking high-performance speech recognition solutions.
Google’s Unintentional Material 3 Expressive Reveal Signals Bold Android UI Upgrades
Google inadvertently published details about its upcoming Material 3 Expressive design language for Android, offering a first look at sweeping changes ahead of its official debut at Google I/O 2025. The leak, quickly removed but preserved via the Wayback Machine, spotlights a focus on color, shape, motion, and containment to create a more visually dynamic and user-friendly interface. Notably, UI actions are now up to 4x faster, and a new pill-shaped floating toolbar is set to enhance usability and visual clarity.
Material 3 Expressive aims to group similar elements, expand tap target sizes, and increase color contrast, delivering a “modern, clean, and energetic” feel. Google’s research-driven approach emphasizes improved accessibility and engagement, promising a significant leap in Android’s visual and interactive experience. This update is expected to roll out with Android’s next major release, setting a new standard for mobile UI/UX design.
Oura Integrates AI-Driven Meal and Glucose Tracking with Dexcom’s Stelo
Oura, known for its health-monitoring smart ring, has introduced two new AI-powered features: glucose tracking and meal logging. The glucose tracking feature, developed in partnership with Dexcom, utilizes the $99 over-the-counter Stelo continuous glucose monitor (CGM), allowing users to view integrated glucose data within the Oura app. This includes AI-generated summaries and a “time above range” metric, indicating how long glucose levels exceed the ideal range of 70-140 mg/dL. Currently, this feature is exclusive to the U.S. and is FSA/HSA eligible.
The meal logging feature, part of the Oura Labs beta program, employs the AI-powered Oura Advisor chatbot to analyze meal photos, estimate macronutrients, and provide personalized dietary advice. When used alongside the glucose feature, it offers an additional metric called “meal level time above range” to track post-meal glucose spikes. This integration aligns with Oura’s recent acquisition of CGM startup Veri and reflects the growing trend in wearable health tech focusing on metabolic health.
Anduril Advances Real-Time Edge Computing with Klas Acquisition and Menace-T Launch
Defense tech innovator Anduril has acquired Dublin-based Klas, a specialist in ruggedized edge computing hardware, marking Anduril’s ninth acquisition and a strategic leap in AI-powered military technology. While financial terms remain undisclosed, Klas brings a team of 150 and its flagship Voyager product line, designed for harsh, remote environments. Anduril had already integrated Klas’s technology into its Menace command centers and Lattice AI platform, supporting real-time data processing for military operations.
The acquisition enables Anduril to deliver portable, powerful edge AI systems-notably the new Menace-T, which fits into two carry-on cases and can be deployed by a single operator in minutes. These advancements are already supporting the U.S. military’s Integrated Visual Augmentation System (IVAS), enhancing real-time object detection and communication in the field. While the immediate impact targets defense and security, Anduril’s ruggedized edge solutions have broad potential across automotive, industrial, healthcare, and remote monitoring sectors, signaling a major step forward for real-time AI and edge computing beyond the battlefield.
Google’s Gemini AI Achieves Gaming Milestone by Completing Pokémon Blue
Google’s Gemini 2.5 Pro has made headlines by becoming the first AI model to complete the classic 1996 Game Boy title “Pokémon Blue”. The feat was accomplished in a Twitch livestream experiment run by independent engineer Joel Z, with minimal but strategic human intervention. Over several sessions, Gemini navigated the game using visual input and agent tools, outperforming Anthropic’s Claude, which is still competing in Pokémon Red. Google CEO Sundar Pichai celebrated the achievement on social media, highlighting Gemini’s growing capabilities as an autonomous agent.
This accomplishment is more than a novelty-it demonstrates the progress of AI models in complex, multi-step reasoning and real-time decision-making. While Gemini’s victory is not a strict benchmark due to differences in agent frameworks, it underscores the rapid evolution of AI in gaming and simulation environments. The event also reflects the increasing rivalry among leading AI labs, with each showcasing their models’ abilities in public challenges to drive engagement and innovation.
Google Expands Gemini AI Access to Children Under 13 with Enhanced Parental Controls
Starting next week, Google will allow children under 13 to access its Gemini AI chatbot-a first for the company’s generative AI tools. Access will be limited to kids with parent-managed Google accounts via Family Link, ensuring robust parental oversight. Parents will receive notifications when their child first uses Gemini and can manage or disable access at any time.
Google has implemented strict guardrails and content filters to protect young users, promising that children’s data will not be used to train AI models. Gemini will assist kids with homework, creative writing, and general queries, but Google is clear about the technology’s limitations: parents are encouraged to help children fact-check answers and avoid sharing personal information. Despite these safeguards, Google acknowledges the risk of inappropriate content and urges ongoing parental involvement. This move reflects the tech industry’s push to engage younger demographics while navigating concerns about AI safety, privacy, and digital literacy.
Google’s AMIE AI Doctor Gains Vision, Surpassing Physicians in Multimodal Medical Diagnostics
Google DeepMind has unveiled a groundbreaking upgrade to its AMIE (Articulate Medical Intelligence Explorer) AI doctor, now equipped with the ability to request, interpret, and reason about visual medical data alongside text during diagnostic conversations. Built on the multimodal Gemini 2.0 Flash model, AMIE was evaluated in a remote study using 105 simulated patient scenarios, where it engaged with patient actors and primary care physicians (PCPs) in instant-messaging style consultations.
The results are striking: AMIE matched or exceeded PCP performance in interpreting multimodal data, diagnostic accuracy, management reasoning, and communication skills. Specialists in dermatology, cardiology, and internal medicine rated AMIE higher for image interpretation and the completeness of differential diagnoses. Notably, the AI’s rate of hallucinations was statistically similar to that of human doctors, while patient actors often found AMIE more empathetic and trustworthy. This innovation signals a major leap for AI in healthcare, with the potential to enhance diagnostic quality and accessibility worldwide.
Google’s Gemini 2.5 Flash Faces Safety Concerns in Latest Evaluation
Google’s Gemini 2.5 Flash, the latest iteration of its AI model, has exhibited a decline in safety performance compared to its predecessor, Gemini 2.0 Flash. According to a technical report released by Google, the new model showed a 4.1% regression in text-to-text safety and a 9.6% regression in image-to-text safety. These metrics assess how frequently the model violates Google’s safety guidelines when responding to textual and visual prompts, respectively.
The decline is attributed to Gemini 2.5 Flash’s enhanced ability to follow instructions, which, while improving responsiveness, also increases the risk of generating content that breaches safety protocols. This development raises concerns about the balance between AI model compliance and safety, especially as companies strive to make AI more permissive in handling sensitive topics.
Airbnb Quietly Deploys AI Customer Service Bot Across the U.S.
Airbnb has begun a nationwide rollout of its new AI-powered customer service chatbot in the United States, aiming to streamline guest and host support. The bot, which leverages recent advances in large language models, is designed to resolve common issues such as booking changes, cancellations, and refund requests without human intervention. Early internal data indicate the AI bot now handles over 50% of customer inquiries in pilot regions, with a resolution rate exceeding 80% for standard cases.
The company reports that the AI system is trained on millions of anonymized support interactions, ensuring it can address a wide range of scenarios while escalating complex cases to human agents. Airbnb’s leadership sees this as a step toward 24/7, instant customer support, reducing response times and operational costs. The move reflects a broader trend of global tech companies integrating AI to improve user experience and operational scalability.
AI2’s Olmo 2 1B Surpasses Competitors in Small-Scale AI Model Benchmarks
The Allen Institute for AI (AI2) has released Olmo 2 1B, a 1-billion-parameter open-source AI model that outperforms similarly sized models from tech giants like Google, Meta, and Alibaba on several benchmarks. Notably, Olmo 2 1B achieved superior results on the GSM8K benchmark for arithmetic reasoning and TruthfulQA for factual accuracy.
Trained on a dataset of 4 trillion tokens from diverse sources, Olmo 2 1B is available under the Apache 2.0 license on Hugging Face, complete with the code and datasets used for its development. This transparency facilitates replication and further research. While AI2 cautions against deploying Olmo 2 1B in commercial settings due to potential risks of generating harmful or inaccurate content, the model represents a significant advancement in accessible, high-performing AI tools for developers and researchers.
Fivetran Acquires Census to Offer Comprehensive Data Movement Solutions
Fivetran, a leader in automated data integration, has announced its acquisition of Census, a reverse ETL platform. While the financial terms remain undisclosed, Census was last valued at $630 million in 2022 and had secured over $80 million in funding from prominent investors like Sequoia Capital, Andreessen Horowitz, and Tiger Global.
This strategic acquisition enables Fivetran to provide an end-to-end data movement platform, facilitating seamless data transfer into and out of cloud warehouses. The integration addresses longstanding customer demands for a reverse ETL solution, allowing businesses to synchronize data across various operational tools. Post-acquisition, the entire Census team will join Fivetran, with plans to merge the Census brand into the Fivetran ecosystem.
Astronomer Secures $93M Series C to Lead AI Infrastructure Orchestration
Astronomer, the company behind Apache Airflow, has closed a $93 million Series C funding round led by Sutter Hill Ventures, with participation from Bain Capital Ventures, Insight Partners, and others. This new capital brings Astronomer’s total funding to over $300 million and underscores the growing demand for AI infrastructure orchestration as organizations scale up data and machine learning operations.
Astronomer’s platform, built on open-source Airflow, orchestrates complex data workflows critical for AI, analytics, and business intelligence. The company reports a 2.5x year-over-year growth in enterprise adoption and now supports thousands of customers, including Fortune 500 firms. The funding will fuel product innovation, global expansion, and deeper integrations with leading AI and data platforms. Astronomer’s rapid growth highlights the centrality of orchestration tools as companies increasingly prioritize automation, reliability, and scalability in their AI infrastructure.
Google Expands AI Mode in Search for Enhanced User Interaction
Google is broadening access to its experimental AI Mode in Search, allowing U.S. users aged 18 and over to opt-in via Google Labs. This feature introduces a more conversational, AI-driven search experience, providing AI-generated answers sourced from Google’s extensive search index.
Key enhancements to AI Mode include:
- Saved Search History: Users can revisit past queries through a new left-side panel, facilitating ongoing research and follow-up questions.
- Visual Information Cards: Detailed cards display information about businesses and products, such as operating hours, reviews, prices, and real-time availability.
- Elimination of Waitlist: The previous requirement for a Google One AI Premium subscription has been removed, making AI Mode more accessible to a broader audience.
These updates position AI Mode as a significant step in integrating advanced AI capabilities into everyday search functions, enhancing user engagement and information retrieval.
Microsoft Unveils Phi-4 Reasoning Models, Rivals Larger AI Systems
Microsoft has introduced its latest Phi-4 Reasoning models, demonstrating that smaller AI models can match the performance of significantly larger systems. The Phi-4 Reasoning Plus model, in particular, approaches the performance levels of DeepSeek R1, which boasts 671 billion parameters, despite having a much smaller size.
The Phi-4 series includes:
- Phi-4 Mini Reasoning: A 3.8-billion-parameter model trained on approximately 1 million synthetic math problems generated by DeepSeek’s R1 model. Designed for educational applications on lightweight devices.
- Phi-4 Reasoning: A 14-billion-parameter model trained using high-quality web data and curated demonstrations, optimized for math, science, and coding tasks.
- Phi-4 Reasoning Plus: An adaptation of the previously released Phi-4 model, fine-tuned for enhanced accuracy in complex tasks.
These models are available with permissive licenses, facilitating their integration into various applications, especially where computational resources are limited.
Meta Projects Up to $1.4 Trillion in Generative AI Revenue by 2035
Meta has forecasted that its generative AI products could generate between $460 billion and $1.4 trillion in revenue by 2035, according to court documents unsealed recently. For 2025, the company anticipates revenue between $2 billion and $3 billion from its generative AI offerings.
To support this growth, Meta’s “GenAI” budget exceeded $900 million in 2024 and is expected to surpass $1 billion in 2025. Additionally, Meta plans to invest $60 billion to $80 billion in capital expenditures in 2025, primarily for expanding its data center infrastructure.
These investments underscore Meta’s commitment to leading the generative AI space, with strategies including revenue-sharing agreements for its Llama models and potential monetization of its Meta AI assistant through ads and subscriptions.
Google Enhances Gemini Chatbot with Advanced Image Editing Features
Google has upgraded its Gemini chatbot, introducing native image editing capabilities that allow users to modify both AI-generated images and personal photos directly within the app. This feature supports over 45 languages and is rolling out globally.
Key functionalities include:
- Multi-step Editing: Users can perform complex edits, such as changing backgrounds, replacing objects, and adding elements, in a conversational manner.
- Integration with Text Prompts: The chatbot can generate images based on user prompts, enhancing storytelling and creative projects.
To address concerns about misuse, Google has implemented invisible watermarks on all images created or edited using Gemini’s native tools and is experimenting with visible watermarks to ensure transparency.
Payment Giants Visa and Mastercard Unveil New AI Features to Enable ‘Agentic Commerce’
Payment industry leaders Visa and Mastercard have launched new initiatives aimed at integrating artificial intelligence into the online shopping experience, paving the way for what is being termed “agentic commerce.” Visa introduced “Intelligent Commerce,” a program designed to empower AI agents to securely browse, select, and purchase items on behalf of consumers. Mastercard, on its part, unveiled “Agent Pay,” a solution focused on embedding payment capabilities directly within AI-powered conversational shopping experiences.
These initiatives involve collaborations with prominent tech and AI companies, including OpenAI, Microsoft, and IBM, signaling a significant push to make AI agents capable participants in digital transactions. The goal is to create more personalized, seamless, and secure shopping journeys where AI can assist users from product discovery through to completing a purchase. This move by the payment processing giants underscores the growing impact of AI in e-commerce and the development of new infrastructure to support AI-driven transactions at scale across their vast networks, which include billions of payment credentials and millions of merchant locations globally.
Pinterest Rolls Out Measures to Address ‘AI Slop’ with Labeling and Content Control Features
Pinterest is introducing new tools and policies designed to tackle the increasing prevalence of low-quality, AI-generated content, often referred to as “AI slop,” on its platform. The visual discovery engine is implementing measures to enhance transparency and provide users with more control over the AI-generated images they encounter. Key among these is the introduction of labels that will identify content detected as being generated or significantly modified by AI.
In addition to labeling, Pinterest is reportedly experimenting with features that would allow users to explicitly indicate if they wish to see less of certain types of AI-generated content. This effort comes in response to user feedback regarding the potential for AI-generated spam to dilute the quality and authenticity of content on the platform, which is widely used for inspiration and ideas. By introducing these tools, Pinterest aims to maintain a high-quality user experience and reinforce its position as a reliable source for genuine, inspiring content amidst the rise of easily created synthetic media.
Amazon Q Business Empowers Companies to Create Public AI Chatbots with Enhanced Customization
Amazon has updated its Q Business platform, now enabling enterprises to build and deploy public-facing AI chatbots for customer engagement, support, and marketing. The update introduces robust customization options, allowing organizations to tailor chatbot personalities, workflows, and integrations to suit their brand and audience. Companies can now leverage Amazon’s advanced natural language models to handle high volumes of customer interactions securely and efficiently.
According to Amazon, early adopters report a 40% reduction in customer service response times and improved satisfaction scores. The platform’s analytics dashboard provides real-time insights into chatbot performance, helping businesses optimize scripts and workflows. This move positions Amazon Q Business as a leading solution for enterprises seeking scalable, AI-driven customer engagement tools in a competitive digital landscape.
JetBrains Open-Sources Mellum: A Developer-Centric AI Coding Model
JetBrains, renowned for its developer tools, has unveiled Mellum, an open-source AI model tailored for code completion. Trained on over 4 trillion tokens and comprising 4 billion parameters, Mellum is optimized for languages like Java, Python, Kotlin, C++, and JavaScript. Available under the Apache 2.0 license on Hugging Face, it aims to assist in intelligent code suggestions within IDEs.
The model’s training spanned 20 days on a cluster of 256 Nvidia H200 GPUs, utilizing datasets from permissively licensed GitHub repositories and English Wikipedia articles. JetBrains emphasizes Mellum’s focus on developer-centric tasks, such as code completion and diff prediction, rather than general-purpose applications.
Duolingo Doubles Language Courses with AI, Boosts Financial Outlook
Duolingo has launched 148 new language courses, effectively doubling its offerings, thanks to the integration of generative AI. This expansion includes beginner-level courses (CEFR A1–A2) with features like Stories and DuoRadio to enhance reading and listening skills.
The company’s AI-first approach has streamlined course creation, reducing development time from years to months. Consequently, Duolingo raised its 2025 revenue forecast to between $987 million and $996 million, up from previous estimates of $962.5 million to $978.5 million. Q1 revenues reached $230.7 million, surpassing analyst expectations. The platform now boasts 10.3 million paying subscribers and 46.6 million daily active users, marking a 49% year-over-year increase.
Wikipedia Integrates AI to Support, Not Replace, Human Editors
The Wikimedia Foundation, the non-profit organization that operates Wikipedia, has unveiled its three-year strategy for integrating artificial intelligence into the platform, with a firm commitment that AI will not be used to replace its global community of human volunteers and editors. Instead, the strategy emphasizes using AI as a tool to support and augment the work of these contributors.
The planned AI applications include automating repetitive tasks, improving content translation across different languages, and enhancing the discoverability of information within the vast encyclopedia. The foundation stressed that its approach to AI integration will prioritize transparency and remain human-centered, aligning with Wikipedia’s core principles of open knowledge and community collaboration. This strategic direction aims to leverage the efficiencies offered by AI technologies while preserving the crucial role of human oversight, judgment, and expertise that forms the backbone of Wikipedia’s content integrity and community-driven model.
AI Legal Platform Supio Secures $60 Million in Series B Funding
Supio, a company specializing in an AI-powered legal platform designed for personal injury and mass tort law firms, has successfully raised $60 million in a Series B funding round. The investment round was led by existing investor Sapphire Ventures, with participation from new investors Mayfield and Thomson Reuters Ventures. This latest infusion of capital brings Supio’s total funding secured to date to $91 million.
The company plans to utilize the funds to fuel its rapid growth, which has seen its Annual Recurring Revenue (ARR) increase fourfold since its Series A round. Supio intends to double its current workforce of 100 employees, enhance its product development efforts, and expand its market presence. The platform leverages AI for document analysis and case management, aiming to improve efficiency, increase settlement values, and achieve better outcomes for law firms and their clients in complex litigation.
CAST AI Raises $108M to Boost AI-Powered Kubernetes Optimization
CAST AI, a Florida-based startup offering automated Kubernetes cost optimization, has secured $108 million in Series C funding, led by Tiger Global, with additional participation from Insight Partners, Creandum, and Cota Capital. This latest round brings CAST AI’s total raised to $173 million, and its valuation now exceeds $900 million, as per insiders speaking to Forbes.
The CAST AI platform uses machine learning algorithms to analyze, predict, and optimize cloud workloads on platforms like AWS, Azure, and Google Cloud. Clients report up to 65% savings on cloud infrastructure. The funding will help the company scale operations, enhance its real-time AI orchestration engine, and continue integrating with enterprise-grade cloud environments. Major customers include OpenWeb, Branch, and ShareChat. With the demand for cost-efficient AI infrastructure growing, CAST AI positions itself as a key enabler of scalable, optimized AI workloads.
AI Tax Compliance Startup Kintsugi Doubles Valuation in 6 Months
Kintsugi, an AI-driven sales tax compliance startup, has doubled its valuation to $440 million just six months after its previous round, reflecting surging demand for automation in financial regulatory compliance. While the company hasn’t disclosed the exact funding amount this time, sources close to the matter confirmed to TechCrunch that it was a follow-on round from Index Ventures and Bessemer Venture Partners, along with new investor Lightspeed Venture Partners.
Kintsugi uses AI models trained on global tax codes to help e-commerce and SaaS companies automate their sales tax calculations, nexus tracking, and filings across jurisdictions. The platform now supports compliance in over 85 countries, and its client base has grown 3x since late 2024. Founded by former Stripe tax engineer Meera Das and ex-Google AI researcher Alan Troy, Kintsugi is also reportedly working on voice-enabled AI agents for CFO dashboards, pushing the frontier of fintech AI.
Hugging Face Launches Free Open Computer Agent, Pioneering Open-Source Agentic AI Tool
Hugging Face, a prominent platform for machine learning models and datasets, has released a free, cloud-hosted agentic AI tool. This new tool is designed to allow artificial intelligence systems to interact with and use a computer, moving towards more autonomous AI operations. While the tool is in its early stages and may exhibit limitations in speed and accuracy, its release represents a notable step in making agentic AI technology more accessible to developers and researchers.
The introduction of this free tool aligns with Hugging Face’s mission to democratize AI and provide resources for building and exploring advanced AI capabilities. By offering a platform for experimenting with AI agents that can perform actions on a computer, Hugging Face is contributing to the development and understanding of the potential applications of interactive AI. This move could accelerate innovation in creating AI systems capable of performing more complex and multi-step tasks.
Reddit Plans to Tighten User Verification to Combat Influx of Human-Like AI Bots
Reddit, the popular online discussion platform, is taking steps to address the growing challenge posed by sophisticated, human-like AI bots. In response to incidents demonstrating the ability of AI to convincingly impersonate human users, Reddit announced plans to implement tighter verification processes. The platform aims to work with third-party services to verify the humanity of users without requiring them to disclose their real identities.
This upcoming change reflects Reddit’s commitment to maintaining an authentic environment for its users and protecting against potential manipulation and misinformation spread by AI-powered accounts. The decision highlights the increasing tension between user anonymity, a core aspect of Reddit’s culture, and the need to ensure that interactions on the platform are genuinely human in the age of advanced generative AI. The details of the verification process and its impact on user privacy are expected to be a key focus as Reddit rolls out these new measures to preserve the integrity of its communities against the proliferation of social media bots.
OpenAI’s $3 Billion Acquisition of Windsurf Strengthens Its Position in AI-Powered Coding
OpenAI has agreed to acquire Windsurf, an AI-assisted coding platform formerly known as Codeium, for approximately $3 billion-its largest acquisition to date. Windsurf specializes in AI tools tailored for developers working with legacy systems, a niche that differentiates it from competitors like Cursor. With an annual recurring revenue (ARR) of around $100 million and rapid growth, Windsurf’s technology complements OpenAI’s existing coding features such as ChatGPT’s code interpreter and collaborative Canvas tool.
This strategic acquisition enables OpenAI to accelerate enhancements in AI-driven software development, particularly for enterprises reliant on complex legacy infrastructure. Windsurf had been in talks with investors including Kleiner Perkins and General Catalyst for funding at a $3 billion valuation. OpenAI’s move reflects its commitment to dominating the AI coding assistant market amid intense competition from Google, Anthropic, and others. The deal is expected to enhance ChatGPT’s coding capabilities, further empowering developers with advanced AI tools.
Particle Extends AI-Enhanced News Reading Experience to the Web Platform
Particle, a startup founded by former Twitter engineers, has launched its AI-powered news reader on the web at Particle.news, broadening access beyond its successful mobile app. The platform aggregates news from multiple sources and delivers AI-generated summaries, key quotes, and contextual entity pages to help users quickly grasp complex stories. By offering topic- and entity-specific pages, users can explore detailed background and related coverage on people, organizations, or products mentioned in the news.
Since its app launch in late 2024, Particle has experienced rapid growth, topping App Store categories in several countries. The web rollout aims to address common frustrations with overwhelming news volume and misinformation by providing clear, balanced, and trustworthy news summaries. Particle’s CEO Sara Beykpour emphasizes the platform’s mission to empower users to shape their news experience while supporting publishers fairly. The startup has raised $4.4 million in seed funding from investors including Kindred Ventures and angel backers like Twitter co-founder Ev Williams. Particle’s innovative approach aligns with the growing demand for AI-assisted news consumption that balances speed, depth, and credibility.
Generative AI Surpasses Cybersecurity in 2025 Tech Budgets
A new AWS-commissioned global survey of 3,739 senior IT decision-makers across nine countries reveals that 45% of organizations name generative AI as their top tech spending priority for 2025, surpassing cybersecurity for the first time. The study, conducted by Access Partnership, highlights that 90% of companies are already using generative AI tools, with nearly half moving beyond pilots to full integration in workflows.
The report also notes significant organizational shifts: 60% have appointed Chief AI Officers or similar roles, and 92% plan to hire AI-experienced talent this year. Companies are blending off-the-shelf AI models with custom-built tools and increasingly relying on external vendors for AI deployment. AWS’s Rahul Pathak interprets the data as a sign of AI’s transformative impact on business, not a diminished focus on security, which remains integral to AI projects. This trend underscores the rapid mainstreaming of generative AI as a core enterprise technology, reshaping IT priorities worldwide.
Unblocked Secures $20M to Enhance AI Assistant for Legacy Codebases
Unblocked, an AI-driven platform designed to help developers understand and navigate legacy codebases, has raised $20 million in Series A funding led by B Capital and Radical Ventures, bringing its total funding to $30 million. Founded by Dennis Pilarinos, a veteran of Microsoft, Amazon, and Apple, Unblocked answers contextual questions about code, tracks code modifications, and connects developers with colleagues who made changes, streamlining collaboration and reducing onboarding time.
The tool integrates with popular developer environments and communication platforms like Slack and Notion, offering features such as Autonomous CI triage to assist in testing code scenarios. Clients including Drata report saving 1-2 hours per week per engineer using Unblocked. As AI-generated code becomes more prevalent, Unblocked’s intelligent assistant addresses growing challenges in maintaining and scaling complex software systems. Investors highlight the platform’s critical role in the evolving AI programming ecosystem, where understanding legacy code remains a significant bottleneck.
Google Unveils Gemini 2.5 Pro Preview Ahead of I/O 2025
Google has announced an update to its Gemini 2.5 Pro AI model, introducing a new version with “massively improved coding capabilities.” Dubbed the “I/O edition” and released ahead of the main Google I/O developer conference, the updated model (labeled gemini-2.5-pro-preview-05-06) demonstrates substantial advancements in areas such as code transformation, editing, and the development of complex agentic workflows. Google DeepMind CEO Demis Hassabis highlighted the model’s enhanced performance, particularly in building interactive web applications.
The improved Gemini 2.5 Pro is now accessible to developers via the Gemini API through Google AI Studio and Vertex AI, and it is also powering features within the Gemini app for general users. Google noted the early release was due to “overwhelming enthusiasm” for the model’s potential. This update positions Gemini 2.5 Pro at the top of several coding and web development leaderboards, signaling Google’s continued focus on refining its AI models for specific, high-demand tasks within the software development domain.
ServiceNow Enhances User Visibility into AI Operations
ServiceNow, a leading digital workflow company, is introducing new features designed to give users more insight and transparency into the workings of their AI systems on the platform. This move addresses a growing demand for understanding how AI models arrive at their decisions and recommendations, particularly in enterprise environments where AI is increasingly used for critical tasks.
The enhancements are aimed at making the AI deployed within ServiceNow’s workflows less of a “black box” and more understandable to end-users and administrators. This includes tools and capabilities that can help explain AI outputs and provide visibility into the data and processes the AI models utilize. By increasing AI transparency, ServiceNow aims to build greater trust and confidence in its AI-powered solutions, facilitating broader adoption and more effective utilization of enterprise AI across various business functions like IT service management, customer service, and human resources.
NewLimit Raises $130 Million for AI-Driven Age-Reversing Therapies
NewLimit, a biotech company co-founded by Coinbase CEO Brian Armstrong, has raised a significant $130 million in Series B funding. The company is focused on developing therapies aimed at reversing aging through epigenetic reprogramming. The funding round was led by Kleiner Perkins, with participation from a roster of notable investors including Khosla Ventures, Founders Fund, Human Capital, Valor Equity Partners, and existing investor Dimension, as well as Brian Armstrong himself.
This substantial investment will fuel NewLimit’s research and development efforts in the highly speculative but potentially transformative field of age-reversing therapies. The company utilizes artificial intelligence and large-scale genomics to identify and develop promising therapeutic candidates. This funding round underscores the significant capital flowing into AI-driven biotech companies pursuing ambitious goals to extend human healthspan and treat age-related diseases, signaling strong investor belief in the potential of AI to unlock breakthroughs in complex biological challenges.
The AI revolution is accelerating at an unprecedented pace, with breakthroughs and investments reshaping every industry. Stay ahead of the curve by following our weekly AI updates-because missing out means falling behind. Don’t miss AI Week #19 for the latest innovations, funding news, and game-changing developments that will define the future. Subscribe now and be the first to know!
