Latest News

Financial technology giant Revolut has teamed up with leading AI platforms to bring conversational trading to its users. The integration of its crypto exchange, Revolut X, with these AI assistants lets people trade and analyze their portfolios using natural language commands.

The list of integrated tools includes Anthropic’s Claude, Google’s Gemini, OpenClaw, and Cursor. These partnerships allow users to execute trades, backtest strategies, and set up custom alerts through simple conversational prompts.

Revolut X now supports a wide range of features that make it easier for traders to analyze the market and manage their portfolios. Users can place both market and limit orders, backtest strategies across over 300 tokens available on the platform, and even receive Telegram notifications when specific price targets are hit.

The technical underpinnings of this feature rely on Anthropic’s Model Context Protocol (MCP), which was first released in November 2024. This protocol is layered on top of Revolut X’s existing trading API to enable seamless integration with the AI assistants.

Revolut has also made available a dedicated GitHub repository that provides installation support for more than 50 AI applications. This move shows the company’s commitment to making its platform accessible and user-friendly for developers as well as users.

As before, all orders require user review and approval before execution, ensuring that no trade goes through without human confirmation. Revolut explicitly disclaims liability for any errors caused by AI-driven actions – they’re not taking on extra risk just because it’s a machine making the call.

The integration of these AI assistants with Revolut X comes hot on the heels of AIR, its in-app AI assistant for general financial tasks, which launched back in April 2026. This new feature is specifically designed for traders and operates at a deeper technical level than its predecessor, enabling users to tackle more complex trading strategies.

A group of around 200 protesters marched through San Francisco over the weekend, calling for a pause in the development of more powerful artificial intelligence (AI) models. The demonstration, organized by the Stop the AI Race coalition, targeted OpenAI, Anthropic, and Google DeepMind, with chants and signs emphasizing three key concerns: AI safety, job displacement, and environmental damage from data centers powering these frontier models.

The protesters’ message was clear: until someone figures out what they’re actually doing to society, stop training more capable AI systems. This is the second time this coalition has taken to the streets in 2026, with a similar march happening in March targeting the same companies and demands.

Organizers directly called on executives at OpenAI, Anthropic, and Google DeepMind to pause new AI system development, highlighting the need for greater transparency and accountability. The signs and chants centered around these three grievances, reflecting growing concerns over the impact of AI on society.

While the protest itself had nothing to do with cryptocurrency, the regulatory environment it reflects is influencing price action in a specific corner of the token market. Decentralized AI tokens, projects built around running AI inference outside company or government control, have been moving in response to AI policy news.

The U.S. government’s recent order restricting access to Anthropic’s models has had a notable impact on these tokens. Venice (VVV) gained approximately 14% following the announcement, while Morpheus (MOR) moved roughly 21%. However, these price swings are also a reminder that this market is highly reactive and thin.

The Stop the AI Race coalition appears to be gaining momentum, with renewed media attention and a clear message. The March protest generated some coverage but little policy response; it remains to be seen whether this latest demonstration will have more impact.

A new accelerator program aimed at supporting startups, research teams, and non-profits working on environmental challenges has been launched by Google DeepMind. The three-month AI for the Planet program will provide selected organizations with access to cutting-edge machine learning tools and expertise to tackle pressing issues such as climate change, sustainable agriculture, and nature protection in Asia Pacific.

The program is open to APAC-headquartered organizations that are working on projects involving artificial intelligence and environmental innovation. Applications are now being accepted until July 26th, 2026, with the selected teams set to join a three-month accelerator program starting from September this year. The program includes a week-long in-person bootcamp, virtual support through December, and an in-person Demo Day at the end of the program.

Google DeepMind will select between 10 to 15 organizations for the inaugural cohort, which will receive access to the Google AI stack, including specialized frontier models, as well as tailored support from Google experts. The program is designed to sit across multiple areas, including AI skills development, startup support, research commercialization, and environmental innovation.

The application criteria cover a range of projects, including those focused on climate change, sustainability, energy, nature protection, sustainable agriculture, and forest protection. Applicants must demonstrate a plan for integrating Google AI into their solutions, either through general models or specialized ones such as AlphaEarth, Forestry, or Perch. Additionally, selected teams will need to have established in-house technical capabilities in AI and machine learning.

Sanjay Jain, Head of Google for Education, India, highlighted the launch on LinkedIn, writing: ‘Got a big vision for the planet and the AI chops to back it up?’ The program is aimed at organizations with a functional prototype or minimum viable product, early validation or proven traction, and a roadmap where AI is central to their current solution or future technical development.

The accelerator will begin with an in-person bootcamp from September 7th to 11th, during which participants will meet Google mentors and define project goals through keynotes, workshops, and diagnostic sessions. After the bootcamp, selected organizations will receive three months of virtual support from September to December, guided by a dedicated relationship manager.

Participants will also have access to one-to-one mentorship and technical guidance from Google and industry experts, as well as tailored technical and business training. Selected startups or projects may be eligible for cloud credits through the Google for Startups Cloud program or Google Cloud for research, subject to eligibility review and approval. The accelerator is equity-free for the duration of the program.

The program will culminate in an in-person Demo Day in December 2026, where participating teams can present their AI-driven environmental work to mentors, investors, partners, and Google teams. Google DeepMind describes the program as a route for early-stage projects to move from experimentation into applied problem-solving: ‘Our mission is to use frontier AI to tackle the world’s most critical environmental challenges.’ The launch adds another AI accelerator route for APAC organizations working at the edge of research, climate technology, and applied AI development.

Applications for the first AI for the Planet cohort close on July 26th, 2026. Google DeepMind is also hosting virtual open forums during the application phase to provide more information about the program and answer questions from interested applicants.

The streets of San Francisco filled with around 200 protesters calling for a halt to the development of more powerful artificial intelligence systems. Their target was OpenAI, Anthropic, and Google DeepMind’s offices, with organizers urging these companies to pause training new AI models until stronger safeguards are in place.

The protest was organized by Stop the AI Race, a group that has been advocating for an international pause on frontier AI development since its first march in March. The movement gained momentum over the past few months as several organizations and unions expressed support for their cause.

Michaël Trazzi, a former AI researcher and one of the organizers behind Stop the AI Race, emphasized that the group’s goals have shifted from solely convincing CEOs to adopt safer practices to also raising awareness about the need for stronger local and state regulation. He noted having protests is an effective way to show people care about this issue.

The demonstration follows a similar protest in March when around 200 people walked between Anthropic, OpenAI, and xAI offices calling for a coordinated pause on frontier AI development. Since then, Stop the AI Race has continued its campaign through protests and public advocacy with some notable successes.

Trazzi expressed his gratitude towards other groups that have joined forces with them including the National Union of Healthcare Workers (NUHW) and QuitGPT. He also praised the quick response from local organizations like AI Action in the Bay Area, which helped to amplify their message.

The protest comes at a time when concerns over AI safety are gaining traction globally. OpenAI introduced new ChatGPT features designed to detect signs of self-harm and violence during conversations following lawsuits and investigations into its chatbot’s handling of sensitive topics in May.

Last June the Donald Trump administration ordered Anthropic to suspend access to two of its models due to potential cybersecurity risks. More recently, a United Nations’ independent scientific panel on AI warned that scientists cannot rule out ‘catastrophic harm’ as AI technology advances faster than our understanding and government oversight.

The protesters cited concerns beyond just AI safety including job losses the environmental impact of AI rising housing costs in San Francisco, and the growing influence of major tech companies. They urged lawmakers to adopt stronger oversight of advanced AI systems and encouraged continued advocacy for an international pause on frontier AI development.

In late June, UpDoc made a significant announcement. The small digital health company revealed that its diabetes app had received clearance from the US Food and Drug Administration (FDA) in December 2025. This marks a milestone for patient-facing large language models (LLMs), but it’s also a reminder of the vast gap between AI therapy and FDA oversight.

The FDA cleared UpDoc’s app as part of the same product category as drug dose calculators, which is narrow compared to the broader landscape of AI-enabled medical devices. The agency has authorized over 1,200 such devices, yet none have been approved for mental health applications. This discrepancy raises questions about why LLMs are being used in therapy and companionship without regulatory scrutiny.

The use of generative AI in therapy/companionship was the number one application in Harvard Business Review’s annual ranking for 2025. OpenAI disclosed that more than a million people per week engage with ChatGPT, discussing sensitive topics like suicide. However, these conversations often occur on products that have never been reviewed by the FDA.

Courts and regulatory bodies are stepping into the void left by the FDA’s inaction. A dozen product-liability cases related to ChatGPT’s role in suicides and mental health harms are now consolidated in a single proceeding in San Francisco Superior Court. States like Illinois, Rhode Island, and Pennsylvania have also taken action, with some banning AI-provided therapy altogether.

The first patient-facing LLM got through the FDA by keeping the model away from critical decisions. UpDoc’s app uses a locked-down algorithm for dosing logic, while the language model handles conversations with patients. This approach may not be sustainable as more companies push to integrate LLMs into their products, potentially giving models greater decision-making authority.

The consequences of this trend are unclear, but one thing is certain: AI therapy is evolving rapidly, and regulatory frameworks must adapt to keep pace. If you or someone you know is struggling with mental health issues, there are resources available, including the 988 Suicide and Crisis Lifeline in the United States.

Tether’s innovation is transforming how businesses use the Microsoft BitNet b1.58 Large Language Model (LLM). For a long time, its high infrastructure needs have kept it out of reach for many organizations. But Tether has just changed this by fine-tuning the 13-billion parameter model in a way that lets companies run it on lower-cost devices.

The company’s edge-tuning framework is what made this possible. By making the BitNet LLM work across multiple types of GPUs, from desktops to handheld devices, they’re opening up new possibilities for businesses that need AI tools without breaking the bank.

Hundreds of healthcare leaders gathered for the first two days of the American Hospital Association’s (AHA) Leadership Summit, where they shared innovative practices from their organizations. One key theme that emerged was the need to humanize patient care in an era dominated by artificial intelligence (AI). Sessions focused on operationalizing AI while also prioritizing empathy and compassion in healthcare delivery.

The summit featured a range of topics, including scaling hospital-at-home programs, redesigning behavioral health services, and implementing team-based models for pregnancy care. These initiatives aim to improve patient outcomes and reduce costs by providing more personalized and effective care.

Another crucial area of focus was strengthening the workforce through innovation. Healthcare leaders discussed ways to retain staff members, enhance professional development opportunities, and create a positive work environment. This effort is essential in addressing the ongoing healthcare workforce shortage.

The summit also highlighted the importance of palliative care in improving patient satisfaction and reducing hospital readmissions. Sessions explored strategies for integrating palliative care into existing treatment plans and providing better support to patients with complex needs.

Throughout the event, speakers emphasized the need for a more holistic approach to healthcare delivery that prioritizes both technological advancements and human connection. By combining AI-driven innovations with compassionate care practices, hospitals can create a more effective and patient-centered system.

Waze, the popular navigation app owned by Google, is getting a significant upgrade with the expansion of its Gemini AI technology. This move marks a deeper integration of artificial intelligence into Waze’s reporting tools and destination search capabilities. The update aims to make navigation more conversational, allowing drivers to interact with the app using natural voice commands.

Waze has already been utilizing Gemini AI in its Conversational Reporting feature, which enables users to report traffic incidents through spoken language. However, this new development takes it a step further by incorporating Gemini’s capabilities into destination search and map updates. The goal is to provide a more personalized experience for drivers while reducing the need to navigate menus or manually enter information.

The first new feature powered by Gemini AI focuses on destination search. Before starting navigation, users can tap the voice icon and ask questions like ‘Find me a coffee shop that’s open right now’ or ‘Find me parking close to Grand Mall.’ Waze will then respond with a list of options, which drivers can begin navigating using voice commands.

The feature is currently being rolled out to the Waze beta community globally on Android and iOS. This expansion demonstrates Google’s commitment to leveraging AI tools for businesses – in this case, enhancing its own navigation app with cutting-edge technology.

Waze is also expanding its Conversational Reporting capabilities, which use Gemini AI to help drivers report traffic incidents through natural speech. The new update extends this capability to map corrections, allowing users to suggest changes like road closures or outdated addresses. This approach aims to make updating maps easier by replacing traditional reporting forms with a conversation-style interface.

Gai Berkovich, VP & General Manager at Waze and Site Lead at Google IL, notes that drivers can simply say ‘The road is closed here’ for the app to send details to local map editors who verify and update the map. This process streamlines data analysis tools by automating tasks and reducing manual input.

Alongside these Gemini-powered updates, Waze is introducing personalized navigation that takes into account a driver’s previous trips alongside the app’s understanding of local traffic patterns. For example, users who typically prefer highways over local roads with frequent stops will see those routes suggested first. Drivers can choose alternative routes or disable this feature through the app’s settings.

Waze also addresses one of the long-standing frustrations among drivers: too many voice alerts. A new ‘less chatty’ mode reduces the number and length of spoken instructions while continuing to provide critical warnings about hazards, turns, and lane changes. This update is designed for drivers who prefer fewer interruptions while listening to music or podcasts.

Another significant addition is a dedicated motorcycle mode, tailored around the different needs of two-wheeled vehicles. Unlike cars, motorcycles can travel on narrower roads and are more sensitive to road conditions. The new mode uses Waze’s real-time traffic data, motorcycle-specific routing information, and input from a dedicated group of motorcycle map editors to suggest suitable routes and highlight hazards relevant to riders.

The motorcycle mode is rolling out in Argentina, Brazil, Colombia, Malaysia, Mexico, Peru, and the Philippines on Android and iOS. Additional countries are planned for rollout.

A new tool aimed at streamlining the process of identifying and fixing vulnerabilities in software has been launched by IBM and Red Hat. The companies have made available a catalog called Lightwell Network, which contains over 6,500 application-layer dependencies that can be used to automate vulnerability remediation. This service is designed to help organizations address what Ben Breard, senior principal product manager at Red Hat, describes as ‘30 years of technical debt’ that has been exposed by the use of artificial intelligence (AI) models to discover vulnerabilities and create exploits in a matter of hours.

The Lightwell Clearinghouse Premier service, which allows application development teams to access validated patches and coordinate remediation efforts, is also now available. However, this service is currently limited to a small number of organizations. Breard notes that the combination of these two offerings will make it easier for companies to address existing vulnerabilities in their code bases.

At the heart of the Lightwell service is a remediation engine that helps identify, validate, and fix vulnerabilities across critical dependencies embedded deep within modern software architectures. The tool then uses automation to backport critical fixes directly to specific production software versions, reducing the time required for updates. Organizations subscribing to the Lightwell Network receive a continuous stream of digitally signed binaries, source code, and compliance artifacts, including software bills of materials (SBOMs).

Technology partners working with IBM and Red Hat include Amazon Web Services (AWS), AMD, F5, GitLab, Intel, JFrog, Microsoft, NVIDIA, Palo Alto Networks, and ServiceNow. Providers of IT services participating in the initiative include Accenture, Atos, Cognizant, Deloitte, EY, HCLTech, Infosys, Kyndryl, LTM, NTT DATA, Tata Consultancy Services (TCS), and Tech Mahindra.

IBM and Red Hat have pledged $5 billion toward the Lightwell initiative, which involves over 20,000 engineers working to remediate vulnerabilities created by humans many years ago or AI coding tools a few days ago. For example, IBM and Red Hat have already fixed a vulnerability dating back to 2001 involving a legacy Apache Debian package still found in enterprise environments more than two decades later.

The ultimate goal is to rapidly improve the quality of code that will increasingly be targeted by cybercriminals as they gain access to advanced AI models. Breard notes that right now, access to these models from Anthropic and OpenAI is limited, but this may not last for long. As those AI models become more sophisticated in the weeks and months ahead, the attacks against software supply chains will likely become more lethal.

The challenge facing organizations is two-fold: they need to discover and remediate existing vulnerabilities as quickly as possible, and provide maintainers of open source software with the tools and expertise needed to resolve application security issues. Historically, companies have tended to wait months before applying a patch to ensure that any update would not inadvertently take their applications offline. However, this process now needs to occur in days and weeks to prevent cybercriminals from exploiting those weaknesses.

Breard notes that organizations are discovering vulnerabilities using AI tools but few are contributing validated code to help open source maintainers resolve issues. Many of the contributions being made are created using AI tools that often introduce additional vulnerabilities and weaknesses that adversaries might easily exploit in a downstream application. The concern is that if the maintainers of an open source project are unable to address the issues, organizations may create their own secure private forks of that code, leading to a fracturing of the community.

One way or another, a technical debt bill that many organizations have been ignoring for decades is now coming due. The issue now is finding the means to pay it off as quickly as possible without breaking the bank.

A new service from Voxpopme is changing the way businesses use their AI assistants to make decisions. The company has launched early access to its MCP (Model Context Protocol) server, which connects customer intelligence to popular AI tools like Claude and ChatGPT.

The Voxpopme MCP server allows research, marketing, product, and innovation teams to query their organization’s internal research repositories directly from within these AI assistants. This means they can get answers grounded in real customer interviews, with a link to the source clips, without having to leave the conversation.

As organizations increasingly rely on AI assistants for day-to-day decisions, they often rely on publicly available information and generic summaries. However, this results in a ‘race to the average’ because these AI assistants use similar datasets to train their responses.

The Voxpopme MCP server changes this dynamic by allowing businesses to connect their AI assistant to proprietary customer intelligence paid for by their organization. This means every answer comes from research studies that have been conducted with rigor and quality controls, backed by real respondents and traceable evidence.

According to Andy Barraclough, Founder and CEO of Voxpopme, ‘Speed and automation were the last race in customer insights.’ He notes that both are now table stakes, but the real question is whether businesses can trust what their AI assistants tell them about customers. The Voxpopme MCP server addresses this concern by keeping every answer grounded in customer truth.

The service works through a simple process: an administrator connects it once, and approved users can then use Voxpopme inside their chosen AI assistant. This allows teams to ask questions directly within the tool they are already using, without having to open another platform.

One of the key benefits of the Voxpopme MCP server is its ability to create new research when no existing study addresses a question. The AI assistant plans and launches new research in real-time, ensuring that every answer is grounded in customer truth.

The service also provides permissioned access to research insights, so each person can only see what they are cleared to view. This ensures that sensitive information remains secure while still allowing teams to make informed decisions.

Availability of the Voxpopme MCP server has begun for research, marketing, product, and innovation teams. Interested parties can book a walkthrough and get started with hands-on onboarding ahead of a wider release.

The service is designed to help businesses transform customer signals into strategic impact. By partnering with enterprise leaders like those in CPG, technology, consumer electronics, restaurants, and F&B, Voxpopme aims to deliver the clarity that leaders need to fuel measurable outcomes and define success.

Anthropic has started adapting its pricing model to suit Indian users, a crucial step in expanding its presence in the country. This move comes as global AI companies increasingly tailor their offerings to win over users in India’s vast market.

In recent days, some users in India have begun seeing local prices on Claude’s website and mobile apps. However, it appears that payments via Unified Payments Interface (UPI), a widely used instant payment network in India, are not yet enabled. Users still need to pay by card or through Apple’s and Google’s app store billing systems.

The introduction of rupee-denominated subscriptions is significant for Indian users, who have long sought this option due to the added friction caused by dollar pricing and currency conversion. This change is particularly notable given that India accounts for 5.8% of global Claude usage, making it the service’s second-largest market after the US.

Anthropic has listed various plans on its website in India, including Claude Pro at ₹2,000 (about $21) a month when billed annually, compared to $17 a month in the US. The prices for other plans, such as Claude Max and Team, also vary significantly from those offered in the US.

The Indian rupee pricing includes local taxes, which may affect the overall cost for users. Interestingly, prices on Claude’s mobile apps differ slightly from those listed on its website.

Anthropic has been expanding its focus on India, with significant developments over the past few months. In February, it opened an office in Bengaluru after announcing plans to do so in October. The company also appointed Irina Ghose, former Microsoft India managing director, as head of its business in the country in January.

Anthropic has partnered with Indian IT services giants Infosys and Tata Consultancy Services in recent months as part of its efforts to scale enterprise AI deployments. However, this expansion faced a setback when Anthropic abruptly suspended access to its Fable 5 and Mythos 5 models for non-US entities in June, prompting some developers and startup founders to consider alternatives.

The restriction on Fable 5 has since been lifted, but access to Mythos 5 remains limited. This incident highlights the challenges AI companies face when expanding into new markets, particularly in a price-sensitive market like India where converting widespread usage into paid subscriptions can be difficult.

A growing number of students, educators, and parents across the country are turning to artificial intelligence tools for help with schoolwork. According to a national survey conducted over the past year, this trend is also evident in Nebraska classrooms. However, research suggests that AI use varies widely depending on factors such as location, income level, and school policies.

The Pew Research Center’s 2025 survey found that more than half of teenagers reported using chatbots for information or help with assignments. A significant majority of these teens – 59% – believed students frequently used AI chatbots to cheat. Parents’ views were divided: while some 51% said their teen uses a chatbot, others expressed concerns about the potential risks.

Parents surveyed by Pew Research Center generally approved of using AI for informational purposes or entertainment but opposed its use for casual conversation or emotional support. This ambivalence reflects broader debates around AI’s role in education and society. As schools increasingly incorporate data analysis tools into their curriculum, questions arise about how to balance benefits with potential risks.

Research conducted by the Center for Democracy and Technology identified four key concerns: data breaches and ransomware attacks, tech-enabled harassment and bullying, malfunctioning AI systems, and problematic interactions between students and technology. Despite these warnings, 85% of schools surveyed reported using AI in some capacity during the 2024-25 school year.

School policies are shifting to accommodate this trend, with many institutions permitting AI use while also implementing safeguards against potential risks. However, teachers’ attitudes toward AI vary widely depending on their experience and comfort level with these tools. The Center for Democracy and Technology found that educators who used AI more frequently were more likely to see its benefits.

Interestingly, students whose schools made extensive use of AI reported feeling a greater distance from their teachers. This finding highlights the need for nuanced consideration of how AI affects student-teacher relationships and overall educational outcomes. As Nebraska Public Media News continues to explore this complex issue, we invite readers to share their thoughts on the role of artificial intelligence in education.

If you’re interested in contributing your insights or have questions about our reporting process, please contact Jolie Peal at [email protected].

A new development in the world of personal AI assistants has been announced by Vellum, a company that offers a local-first assistant with native macOS app and self-hosted service capabilities. The launch marks an important milestone for Vellum’s ecosystem, which is designed to provide users with more control over their data and integrations. At its core, the Plugin Hub is an open catalog of installable plugins that can be used in conjunction with Vellum’s AI assistant.

The key difference between skills and plugins lies in their functionality. Skills are typically one-off tasks, such as drafting a single email or performing a specific action within a workflow. Plugins, on the other hand, provide the assistant with ongoing capabilities that it can execute independently. This includes connecting to services like Gmail for inbox reading, managing calendars, tracking expenses, syncing notes, and monitoring code repositories.

Each plugin is designed to be a complete function that the assistant owns, rather than just a one-off prompt or instruction bundle. To develop plugins, Vellum uses TypeScript packages with lifecycle hooks, which differ from skills in their design and functionality. The manifest for each plugin utilizes the vellumai plugin-api and exposes specific hooks for initialization, user prompts, tool usage, model calls, and shutdown, ensuring that state is persisted to a local JSONL file.

This setup allows more people than just infrastructure engineers to build plugins, making it easier for users to create custom integrations without needing extensive coding knowledge. To cater to those who prefer not to write code, Vellum offers a plugin builder skill that generates a plugin from a plain-language description. This feature is crucial in democratizing the development process and enabling more users to contribute to the Plugin Hub.

Another significant aspect of this setup is credential handling. When plugins require API keys or OAuth tokens for accessing email, banking, or code repositories, those credentials remain on the machine and are not sent to the model. This ensures that sensitive integrations can be trusted with the assistant’s capabilities. Plugins themselves are open-source and portable, allowing them to be shared among users rather than being tied to a single vendor.

A notable feature of plugins is their ability to operate across different platforms. A plugin installed on macOS can run through Slack or Telegram with shared memory, making its capability available wherever the assistant is accessed. This flexibility enables developers to build once and deploy multiple times, streamlining the development process for users who need cross-platform compatibility.

The launch catalog includes a variety of use cases that demonstrate the potential of plugins in real-world applications. The Marketing Expert plugin bundles positioning, launches, content creation, brand voice management, and reporting into one cohesive tool. Similarly, the AI Hero Engineer Kit packages engineering workflows around plans, PRDs, vertical-slice issues, and test-driven loops.

Other notable examples include the Writing Coach, which runs daily prompts that escalate with a streak and critiques drafts, as well as the Fitness Companion, which logs meals, workouts, and weight, pulls nutrition data from OpenFoodFacts, and calculates macro targets. These plugins showcase the versatility of Vellum’s ecosystem in handling diverse tasks and workflows.

Users can install plugins directly from the catalog or by providing a GitHub URL for direct download. Authors are also encouraged to publish their own plugins to the same catalog, fostering a community-driven approach to plugin development and sharing. This collaborative environment is expected to drive innovation within Vellum’s ecosystem, as users contribute and build upon existing capabilities.

The Plugin Hub marks an important step in transforming Vellum from a product into a platform that can be extended and customized by its users. With this launch, the company aims to provide more comprehensive support for data analysis tools and AI assistants in various business settings.

Itransition, a global software engineering and technology consulting company, has been recognized by Elio as one of the best Microsoft 365 partners and resellers worldwide. This recognition highlights Itransition’s extensive expertise in the Microsoft ecosystem, its comprehensive portfolio of Microsoft business applications, and its proven track record of helping organizations modernize their operations through cloud, AI, ERP, CRM, and productivity solutions.

Elio’s Microsoft partner directory features technology providers with demonstrated capabilities in Microsoft solutions, which helps organizations identify trusted partners for Microsoft 365 licensing, implementation, integration, migration, consulting, managed services, and digital transformation initiatives. The featured partners represent a broad range of expertise across Microsoft 365, Microsoft Azure, Dynamics 365, AI, cybersecurity, cloud infrastructure, and enterprise productivity solutions.

Itransition stands out among its peers due to its comprehensive portfolio of Microsoft business applications, which includes end-to-end implementation, migration, integration, modernization, and managed services. The company’s extensive experience in software engineering spans over 28 years, with a team of more than 3,000 technology professionals across the United States, the United Kingdom, Germany, the UAE, Lithuania, and Poland.

Itransition delivers secure, scalable, and enterprise-grade Microsoft solutions that help organizations modernize their business processes, migrate workloads to the cloud, improve collaboration, automate workflows, and build intelligent business applications. The company’s expertise spans Microsoft 365, Dynamics 365, Azure, Power Platform, AI, ERP, CRM, and other Microsoft technologies.

Itransition is a trusted partner for over 800 clients worldwide, including globally recognized brands such as Mercedes-Benz, PayPal, and PepsiCo. The company combines deep engineering expertise with a mature partner ecosystem, making it an ideal choice for organizations seeking to modernize their operations through cloud, AI, and business applications.

The recognition by Elio is a testament to Itransition’s commitment to delivering high-quality Microsoft solutions that meet the evolving needs of its clients. With this achievement, Itransition solidifies its position as one of the leading Microsoft 365 partners and resellers worldwide.

Itransition has been a Microsoft partner since 2008 and provides end-to-end consulting, implementation, integration, and support services spanning Microsoft 365, Dynamics 365, Azure, Power Platform, and AI-powered business applications. The company’s team of experts helps organizations accelerate digital transformation through custom software development, cloud, AI, data, and enterprise technology solutions.

Itransition is a global IT engineering and consulting company that helps organizations accelerate digital transformation through tailored enterprise software, cloud technologies, AI, and Microsoft business applications. With more than 28 years of experience in software engineering, Itransition delivers end-to-end Microsoft services spanning Microsoft 365, Dynamics 365, Azure, Power Platform, and AI solutions.

The company’s expertise extends beyond Microsoft technologies to include custom software engineering, Salesforce consulting, ERP and CRM implementation, application modernization, data engineering, and managed IT services. This comprehensive approach enables organizations to streamline their operations, improve collaboration, automate workflows, and build intelligent business applications that integrate seamlessly with existing enterprise systems.

Itransition is a trusted partner for over 800 clients worldwide, including globally recognized brands such as Mercedes-Benz, PayPal, and PepsiCo. The company’s team of experts helps organizations modernize their operations through cloud, AI, ERP, CRM, and productivity solutions, ensuring secure, scalable, and innovative technology environments that drive long-term business growth.

The recognition by Elio is a testament to Itransition’s commitment to delivering high-quality Microsoft solutions that meet the evolving needs of its clients. With this achievement, Itransition solidifies its position as one of the leading Microsoft 365 partners and resellers worldwide.

Waze, the popular navigation app powered by user contributions, has rolled out a range of new features designed to give drivers more control over their journeys. The updates include enhanced customization options, improved motorcycle routing, and expanded capabilities for Waze’s intelligent partner feature, Gemini.

One key addition is motorcycle mode, which uses AI to provide two-wheeler-specific routes that take into account narrower streets and road surfaces. This feature also highlights hazards such as potholes, speed bumps, raised crosswalks, shoulder endings, and narrow bridges. Motorcycle mode has begun rolling out in Argentina, Brazil, Colombia, Malaysia, Mexico, Peru, and the Philippines on both Android and iOS devices.

The new motorcycle mode is made possible by Waze’s real-time traffic information and a dedicated group of motorcycle map editors who continually update the app with new hazards. This collaboration between users and developers ensures that riders have access to accurate and relevant route suggestions.

Waze has also introduced personalized navigation, which suggests routes based on previous trips and hyperlocal understanding of city traffic patterns. Users can choose to opt out of personalized routing or select alternative routes in their settings. Personalized navigation is now available globally on both Android and iOS devices.

Another feature aimed at improving the driving experience is ‘less chatty’ mode, which minimizes voice prompts and keeps them concise. This allows drivers to focus on music or podcasts without interruptions from Waze. Less chatty mode has begun rolling out globally on both platforms.

Waze’s conversational reporting capabilities have also been expanded with Gemini features. Users can now report road updates by simply speaking naturally, such as saying ‘The road is closed here.’ These suggestions are then sent to local map editors for verification and updating of the app’s maps.

The Gemini-powered search voice icon has also received an update, enabling users to ask questions like ‘Find me a coffee shop that’s open right now’ or ‘Find me parking close to Grand Mall.’ Waze responds with a list of options, allowing drivers to quickly start navigating using their voice. This feature is currently available in the Waze beta community globally on both Android and iOS devices.

The new features mark an expansion of Waze’s capabilities as it moves from being a helpful companion to an intelligent partner for drivers. The app continues to rely on user contributions, with map editors playing a crucial role in maintaining accurate and relevant route suggestions.

Meta has suspended its new AI image feature, Muse Image, after facing intense criticism over the past few days. The tool allowed users to generate images by tagging public Instagram accounts, but it required those with public accounts above 18 to opt out if they didn’t want their images used in this way.

The backlash was swift and came from various corners of the industry. Talent agencies like CAA expressed concerns about the feature’s opt-out policy, which they deemed unreasonable. ‘No one’s name, image, likeness, voice or creative work should be used by any third party without clear, documented consent,’ a spokesperson for CAA said.

CAA is not alone in its criticism of Meta’s approach to AI-generated content. SAG-AFTRA, the union representing U.S. performers, also urged its members to opt out of the feature, warning that their likenesses could be scraped by models without explicit permission.

The controversy surrounding Muse Image marks a significant development in the ongoing debate about how companies handle user data and consent when it comes to AI tools for businesses. Meta’s decision to disable the feature is seen as a step towards greater transparency and accountability in this area.

However, not everyone is pleased with the outcome. OpenAI, another prominent player in the AI space, has faced similar criticism over its own handling of user data. The company temporarily had an opt-out feature for its Sora 2 video model before changing its policies and eventually shutting it down earlier this year.

The recent developments come as Apple sued OpenAI in federal court on Friday, alleging that the ChatGPT maker stole company secrets with the help of ex-Apple employees. OpenAI has denied these claims, but the lawsuit highlights the ongoing tensions between AI companies and their users when it comes to issues like data consent and intellectual property.

Meta’s decision to suspend Muse Image is seen as a response to these growing concerns about user rights and consent in the age of free ai video generators and other AI tools for business. As the industry continues to evolve, one thing is clear: companies will need to prioritize transparency and accountability if they want to build trust with their users.

The backlash against Meta’s Muse Image feature has also sparked a wider conversation about the role of AI-generated content in social media. With billions of people already using these platforms, it’s essential that companies like Meta take steps to protect user rights and ensure that creators are treated fairly when it comes to issues like image use and consent.

The controversy surrounding Muse Image is just one example of the challenges facing AI companies as they continue to experiment with new features and push the boundaries of what’s possible. As users become increasingly aware of their rights and the importance of data consent, companies will need to adapt and find ways to balance innovation with responsibility.

A new tool from Meta that allowed users to generate artificial intelligence content using publicly posted Instagram photos has been removed after sparking immediate backlash and criticism over privacy concerns. The feature, called Musa Image, was launched just days ago on Tuesday but is no longer available as of Friday.

The tool used AI to create new images by pulling ‘part or all’ of a user’s published photos from their public Instagram account. This raised alarms among privacy advocates who argued that users could be creating AI-generated images using another person’s likeness without their consent and without any notification.

Meta initially defended the feature, stating its intent was to provide a useful creative tool for people while giving them control over whether their public content could be referenced in this way. However, after hearing feedback from critics, including journalist Taylor Lorenz, who covers technology, Meta updated its blog post to announce that the tool would no longer be available.

The backlash against Musa Image was swift and severe, with some calling it a gross invasion of privacy. ‘The fact that someone you don’t know could take your picture or your image and doctor an AI-generated image of you is just really gross,’ said J.B. Branch, director of AI policy at Public Citizen.

The removal of Musa Image marks a significant development in the ongoing debate over the use of AI tools for businesses and individuals alike. As concerns about data privacy continue to grow, companies like Meta are being forced to re-examine their approach to using user-generated content in new and innovative ways.

Microsoft’s latest push in AI assistants has reached a new level of maturity. The company is shifting its standard apps from chatbots that generate text to workflow producers, aiming to turn artificial intelligence into an autonomous digital workforce.

The idea behind ‘the agentic enterprise’ – Microsoft’s vision for the future of work – is to evolve AI systems beyond just answering questions. Instead, they will plan, execute, monitor, and collaborate on business tasks with minimal human intervention. It’s a significant upgrade in how companies use AI assistants today.

The first step toward this agentic enterprise is to transform Microsoft 365 Copilot from being largely reactive into an autonomous system that takes initiative. This means understanding goals rather than individual prompts, breaking work into multiple steps, and choosing the right tools for each task.

For example, instead of asking Copilot to write a sales report, users can say ‘Prepare next week’s executive sales review.’ The AI then gathers CRM data, analyzes spreadsheets, summarizes Teams conversations, identifies major customer risks, drafts PowerPoint slides, schedules presentations, and emails participants – all without constant prompting.

Microsoft 365 Copilot Cowork achieves this by turning massive, multi-step prompts into background execution plans with defined human checkpoints. It can run for hours or even days, making it a substantial leap forward in AI assistants for business use cases.

One of the most notable developments around Microsoft Copilot Cowork over the past few months is its transition from an experimental ‘Frontier’ feature to a generally available enterprise product. The company sees it as the next stage of AI assistants – one that doesn’t just answer questions but executes multi-step work on behalf of users.

To support this push toward more advanced AI tools, Microsoft has introduced two new research tools: Critique and Council. These enable more comprehensive data analysis and comparison across multiple top-tier models simultaneously, part of the company’s broader effort to develop multi-model AI capabilities.

Critique is designed for complex tasks like competitive intelligence, market analysis, scientific literature reviews, financial research, legal and policy analysis, and enterprise strategy documents. It requires evaluating evidence from many sources rather than just answering a single question – which makes it ideal for these types of high-level work.

Council works by routing the request to several models and presenting their responses side-by-side for comparison. This allows users to compare different answers to the same prompt and choose the one that best fits their needs, giving them more control over AI decision-making processes.

What’s unique about Council is its ability to query multiple AI models independently answer a single prompt. Users can see where models agree or disagree – which helps in evaluating the credibility of each response. In contrast, Critique involves having one model review and improve another model’s response before it’s returned.

In addition to these tools, Microsoft has released Legal Agent within Word as part of its 365 Copilot capabilities. This is specifically designed for legal professionals who need help drafting, reviewing, and analyzing documents directly from the comfort of their familiar document management environment – where they can interact with AI agents using natural language prompts.

Legal Agent assists throughout a legal document’s lifecycle by combining generative AI with enterprise document management features like citations, organizational knowledge, and version control. It drafts contracts based on templates or user input, reviews agreements for unusual clauses or inconsistencies, compares document versions to highlight changes between drafts – and summarizes lengthy contracts into concise overviews of key terms, dates, parties involved, and risks.

It can also suggest revisions to improve clarity or align language with organizational standards. Furthermore, Legal Agent extracts important terms from documents like renewal dates, payment obligations, termination provisions, indemnification clauses, and governing law – and even generates executive summaries that non-lawyers can quickly understand without needing extensive legal background.

Because Legal Agent operates within Word, lawyers interact with the agent using natural language prompts while continuing to edit their documents normally. Microsoft emphasizes that Legal Agent is intended as a tool for attorneys, not a replacement – users still need to review AI-generated content and provide approval before it’s finalized.

Anthropic, the company behind Claude, has addressed user complaints about its performance decline. The issue at hand is not that the model itself has become less capable, but rather that users have been misinterpreting how to use it effectively. This misunderstanding has led many developers to upgrade to more expensive models in an attempt to improve their AI’s intelligence, only to find that this trick does not always work as expected.

In fact, switching to a larger model is often not the solution to improving performance. According to Anthropic, users have been confusing two options in Claude Code: Model selection and Effort level. The default setting for Effort level was recently changed from high to medium, which has led many users to believe that their AI’s capabilities are declining.

The root cause of this issue is not a problem with the model itself but rather a misunderstanding about how to use it effectively. Many developers have been upgrading to more expensive models in an attempt to improve performance without realizing that the Effort level setting plays a crucial role in determining the AI’s behavior and output.

In March, many users reported issues with their Claude Code, including errors such as failing to read required files, skipping scheduled tests, and abandoning tasks halfway. The head of AI at AMD, Stella Laurenzo, conducted an investigation into these issues and found that Claude’s thinking volume had dropped by 67% compared to before February.

The issue was not with the model itself but rather a result of the default Effort level being changed from high to medium in March. This change was noted in the official update log, but many users did not notice it until later when they realized that their AI’s performance had declined.

After bearing the backlash for a month, Anthropic restored the default Effort level on April 7 and reset the usage quota for all subscription users. It was then that most users realized this switch had been right beside them all along, secretly determining whether the AI would devote its full capacity to work for them.

Anthropic’s official breakdown of how Claude works can be summed up in one simple sentence: Model determines capability, Effort determines dedication. The model defines the underlying capability and is backed by a set of ‘frozen weights’ that are permanently fixed during training. These weights cannot be altered or modified during inference, meaning that users can guide the AI but not train it.

Switching models essentially means replacing the entire set of weights to take over your task, solving the problem of whether the AI is capable of doing it. However, this does not necessarily mean that a larger model will always outperform a smaller one. In fact, Anthropic has found that a small model running at high Effort can absolutely outperform a large model running at low Effort on many tasks.

The division of responsibilities between Model and Effort is crucial to understanding how Claude works. The official judgment framework becomes extremely useful once users understand this concept. When Claude makes a mistake, the first step is always to check the context: is the prompt clear? Are all required tools provided? Is the CLAUDE.md file configured correctly?

Most so-called ‘AI getting dumber’ issues are rooted in incorrect configuration or insufficient dedication rather than an issue with the model itself. If the context is indeed correct but the AI still makes mistakes, users should ask themselves: is it incapable of doing the task, or is it just not trying hard enough? Not trying hard enough is easy to identify: it skips files that should be read, fails to run tests, and abandons a refactoring task halfway to ask for help.

If it’s genuinely ‘incapable’, then no amount of increased Effort will help. The issue lies with the model itself, and users need to switch to a more powerful model. Anthropic has given an analogy to explain this concept: Sonnet is like a versatile all-rounder who has an entire afternoon available to work; Opus is an expert who only gives you 5 minutes.

Each model has its strengths and weaknesses, and users should not treat it as a rigid token budget limit. The Effort setting in Claude Code determines how thoroughly and confidently the task needs to be completed, governing text responses, tool calls, and extended thinking all at once. For example, for the exact same prompt, Claude at high Effort can generate roughly 7 times more tokens than at low Effort.

The official team has also released an illustration showing that a small model running at high Effort can genuinely outperform a large model running at low Effort on many tasks. This is because a small model paired with sufficient context and high dedication can handle far more work than expected.

In the past, users would simply pick the most powerful model without considering other factors such as Effort level. However, this approach has become outdated, and users now need to act like project managers, assigning different roles and dedication levels to different models. This requires a new skillset: orchestrating AI agents rather than just relying on more powerful models.

The newly added ‘ultracode’ option in Claude Code’s Effort menu brings this orchestration mechanism directly into the product. When selected, Claude gets the xhigh Effort level, plus an authorization to decide on its own whether to spin up a team of AI agents, split the task, and run it in parallel.

The era of only looking at model rankings is passing, and orchestrating models is becoming a core skill. Whoever learns to assign tasks to AI first will get a head start, using a Claude that truly goes all out for them.

Digital bank Monzo has launched an automated debt consolidation service, streamlining the process of paying off existing debts for its customers. The new offering is made possible through a partnership with credit score provider ClearScore, which brings its Clearer technology to bear on the task.

The service works by bringing together all of a customer’s borrowing into a single loan, and then automatically finding and settling their outstanding debt directly with lenders. This means that customers no longer need to contact lenders or check balances themselves, freeing up time and reducing stress.

According to Monzo Head of Borrowing Luke Enock, the new service is designed to make it easier for customers to manage their finances and get on top of their debts. ‘For customers, this means less admin and confidence that the loan is being used to pay down what they already owe,’ he explained in a LinkedIn post.

The automated debt consolidation experience requires a Monzo current account, and is available to UK residents aged 18 and over who meet certain eligibility criteria. The service also comes with its own set of terms and conditions, which customers will need to review before applying.

Monzo’s representative APR for loans up to £10,000 is 21.8%, while the rate drops to 10.2% for loans between £10,001 and £35,000. These rates are in line with industry standards, but may vary depending on individual circumstances.

The partnership with ClearScore marks a significant development in Monzo’s mission to make money work for everyone. ‘This is another example of our commitment to using technology to simplify financial services and improve people’s lives,’ Enock said.

Clearer, the automated debt repayment technology developed by ClearScore, ensures that when a customer takes out a consolidation loan, their existing debts are automatically paid off rather than being used for other purposes. This reduces the risk of default and makes it easier for lenders to assess creditworthiness.

The new service is part of Monzo’s ongoing efforts to expand its offerings and attract new customers. The bank has recently increased spending on its ‘refer a friend’ program, which saw payouts rise by almost 40% in the year leading up to March.