Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More After three hours at Google’s I/O 2025 event last week in Silicon Valley, it became increasingly clear: Google is rallying its formidable AI efforts – prominently branded under the Gemini name but encompassing a diverse range of underlying model architectures and research – with laser focus. It is releasing a slew of innovations and technologies around it, then integrating them into products at a breathtaking pace. Beyond headline-grabbing features, Google laid out a bolder ambition: an operating system for the AI age – not the disk-booting kind, but a logic layer every app could tap – a “world model” meant to power a universal assistant that understands our physical surroundings, and reasons and acts on our behalf. It’s a strategic offensive that many observers may have missed amid the bamboozlement of features.  On one hand, it’s a high-stakes strategy to leapfrog entrenched competitors. But on the other, as Google pours billions into this moonshot, a critical question looms: Can Google’s brilliance in AI research and technology translate into products faster than its rivals, whose edge has its own brilliance: packaging AI into immediately accessible and commercially potent products? Can Google out-maneuver a laser-focused Microsoft, fend off OpenAI’s vertical hardware dreams, and, crucially, keep its own search empire alive in the disruptive currents of AI? Google is already pursuing this future at dizzying scale. Pichai told I/O that the company now processes 480 trillion tokens a month – 50× more than a year ago – and almost 5x more than the 100 trillion tokens a month that Microsoft’s Satya Nadella said his company processed. This momentum is also reflected in developer adoption, with Pichai saying that over 7 million developers are now building with the Gemini API, representing a five-fold increase since the last I/O, while Gemini usage on Vertex AI has surged more than 40 times. And unit costs keep falling as Gemini 2.5 models and the Ironwood TPU squeeze more performance from each watt and dollar. AI Mode (rolling out in the U.S.) and AI Overviews (already serving 1.5 billion users monthly) are the live test beds where Google tunes latency, quality, and future ad formats as it shifts search into an AI-first era. Source: Google I/O 20025 Google’s doubling-down on what it calls “a world model” – an AI it aims to imbue with a deep understanding of real-world dynamics – and with it a vision for a universal assistant – one powered by Google, and not other companies – creates another big tension: How much control does Google want over this all-knowing assistant, built upon its crown jewel of search? Does it primarily want to leverage it first for itself, to save its $200 billion search business that depends on owning the starting point and avoiding disruption by OpenAI? Or will Google fully open its foundational AI for other developers and companies to leverage – another  segment representing a significant portion of its business, engaging over 20 million developers, more than any other company?  It has sometimes stopped short of a radical focus on building these core products for others with the same clarity as its nemesis, Microsoft. That’s because it keeps a lot of core functionality reserved for its cherished search engine. That said, Google is making significant efforts to provide developer access wherever possible. A telling example is Project Mariner. Google could have embedded the agentic browser-automation features directly inside Chrome, giving consumers an immediate showcase under Google’s full control. However, Google followed up by saying Mariner’s computer-use capabilities would be released via the Gemini API more broadly “this summer.” This signals that external access is coming for any rival that wants comparable automation. In fact, Google said partners Automation Anywhere and UiPath were already building with it. Google’s grand design: the ‘world model’ and universal assistant The clearest articulation of Google’s grand design came from Demis Hassabis, CEO of Google DeepMind, during the I/O keynote. He stated Google continued to “double down” on efforts towards artificial general intelligence (AGI). While Gemini was already “the best multimodal model,” Hassabis explained, Google is working hard to “extend it to become what we call a world model. That is a model that can make plans and imagine new experiences by simulating aspects of the world, just like the brain does.”  This concept of ‘a world model,’ as articulated by Hassabis, is about creating AI that learns the underlying principles of how the world works – simulating cause and effect, understanding intuitive physics, and ultimately learning by observing, much like a human does. An early, perhaps easily overlooked by those not steeped in foundational AI research, yet significant indicator of this direction is Google DeepMind’s work on models like Genie 2. This research shows how to generate interactive, two-dimensional game environments and playable worlds from varied prompts like images or text. It offers a glimpse at an AI that can simulate and understand dynamic systems. Hassabis has developed this concept of a “world model” and its manifestation as a “universal AI assistant” in several talks since late 2024, and it was presented at I/O most comprehensively – with CEO Sundar Pichai and Gemini lead Josh Woodward echoing the vision on the same stage. (While other AI leaders, including Microsoft’s Satya Nadella, OpenAI’s Sam Altman, and xAI’s Elon Musk have all discussed ‘world models,” Google uniquely and most comprehensively ties this foundational concept to its near-term strategic thrust: the ‘universal AI assistant.) Speaking about the Gemini app, Google’s equivalent to OpenAI’s ChatGPT, Hassabis declared, “This is our ultimate vision for the Gemini app, to transform it into a universal AI assistant, an AI that’s personal, proactive, and powerful, and one of our key milestones on the road to AGI.”  This vision was made tangible through I/O demonstrations. Google demoed a new app called Flow – a drag-and-drop filmmaking canvas that preserves character and camera consistency – that leverages Veo 3, the new model that layers physics-aware video and native audio. To Hassabis, that pairing is early proof that ‘world-model understanding is already leaking into creative tooling.’ For robotics, he separately highlighted the fine-tuned Gemini Robotics model, arguing that ‘AI systems will need world models to operate effectively.” CEO Sundar Pichai reinforced this, citing Project Astra which “explores the future capabilities of a universal AI assistant that can understand the world around you.” These Astra capabilities, like live video understanding and screen sharing, are now integrated into Gemini Live. Josh Woodward, who leads Google Labs and the Gemini App, detailed the app’s goal to be the “most personal, proactive, and powerful AI assistant.” He showcased how “personal context” (connecting search history, and soon Gmail/Calendar) enables Gemini to anticipate needs, like providing personalized exam quizzes or custom explainer videos using analogies a user understands (e.g., thermodynamics explained via cycling. This, Woodward emphasized, is “where we’re headed with Gemini,” enabled by the Gemini 2.5 Pro model allowing users to “think things into existence.”  The new developer tools unveiled at I/O are building blocks. Gemini 2.5 Pro with “Deep Think” and the hyper-efficient 2.5 Flash (now with native audio and URL context grounding from Gemini API) form the core intelligence. Google also quietly previewed Gemini Diffusion, signalling its willingness to move beyond pure Transformer stacks when that yields better efficiency or latency. Google is stuffing these capabilities into a crowded toolkit: AI Studio and Firebase Studio are core starting points for developers, while Vertex AI remains the enterprise on-ramp. The strategic stakes: defending search, courting developers amid an AI arms race This colossal undertaking is driven by Google’s massive R&D capabilities but also by strategic necessity. In the enterprise software landscape, Microsoft has a formidable hold, a Fortune 500 Chief AI Officer told VentureBeat, reassuring customers with its full commitment to tooling Copilot. The executive requested anonymity because of the sensitivity of commenting on the intense competition between the AI cloud providers. Microsoft’s dominance in Office 365 productivity applications will be exceptionally hard to dislodge through direct feature-for-feature competition, the executive said. Google’s path to potential leadership – its “end-run” around Microsoft’s enterprise hold – lies in redefining the game with a fundamentally superior, AI-native interaction paradigm. If Google delivers a truly “universal AI assistant” powered by a comprehensive world model, it could become the new indispensable layer – the effective operating system – for how users and businesses interact with technology. As Pichai mused with podcaster David Friedberg shortly before I/O, that means awareness of physical surroundings. And so AR glasses, Pichai said, “maybe that’s the next leap…that’s what’s exciting for me.” But this AI offensive is a race against multiple clocks. First, the $200 billion search-ads engine that funds Google must be protected even as it is reinvented. The U.S. Department of Justice’s monopolization ruling still hangs over Google – divestiture of Chrome has been floated as the leading remedy. And in Europe, the Digital Markets Act as well as emerging copyright-liability lawsuits could hem in how freely Gemini crawls or displays the open web. Finally, execution speed matters. Google has been criticized for moving slowly in past years. But over the past 12 months, it became clear Google had been working patiently on multiple fronts, and that it has paid off with faster growth than rivals. The challenge of successfully navigating this AI transition at massive scale is immense, as evidenced by the recent Bloomberg report detailing how even a tech titan like Apple is grappling with significant setbacks and internal reorganizations in its AI initiatives. This industry-wide difficulty underscores the high stakes for all players. While Pichai lacks the showmanship of some rivals, the long list of enterprise customer testimonials Google paraded at its Cloud Next event last month – about actual AI deployments – underscores a leader who lets sustained product cadence and enterprise wins speak for themselves.  At the same time, focused competitors advance. Microsoft’s enterprise march continues. Its Build conference showcased Microsoft 365 Copilot as the “UI for AI,” Azure AI Foundry as a “production line for intelligence,” and Copilot Studio for sophisticated agent-building, with impressive low-code workflow demos (Microsoft Build Keynote, Miti Joshi at 22:52, Kadesha Kerr at 51:26). Nadella’s “open agentic web” vision (NLWeb, MCP) offers businesses a pragmatic AI adoption path, allowing selective integration of AI tech – whether it be Google’s or another competitor’s – within a Microsoft-centric framework. OpenAI, meanwhile, is way out ahead with the consumer reach of its ChatGPT product, with recent references by the company to having 600 million monthly users, and 800 million weekly users. This compares to the Gemini app’s 400 million monthly users. And in December, OpenAI launched a full-blown search offering, and is reportedly planning an ad offering – posing what could be an existential threat to Google’s search model. Beyond making leading models, OpenAI is making a provocative vertical play with its reported $6.5 billion acquisition of Jony Ive’s IO, pledging to move “beyond these legacy products” – and hinting that it was launching a hardware product that would attempt to disrupt AI just like the iPhone disrupted mobile. While any of this may potentially disrupt Google’s next-gen personal computing ambitions, it’s also true that OpenAI’s ability to build a deep moat like Apple did with the iPhone may be limited in an AI era increasingly defined by open protocols (like MCP) and easier model interchangeability. Internally, Google navigates its vast ecosystem. As Jeanine Banks, Google’s VP of Developer X, told VentureBeat serving Google’s diverse global developer community means “it’s not a one size fits all,” leading to a rich but sometimes complex array of tools – AI Studio, Vertex AI, Firebase Studio, numerous APIs. Meanwhile, Amazon is pressing from another flank: Bedrock already hosts Anthropic, Meta, Mistral and Cohere models, giving AWS customers a pragmatic, multi-model default. For enterprise decision-makers: navigating Google’s ‘world model’ future Google’s audacious bid to build the foundational intelligence for the AI age presents enterprise leaders with compelling opportunities and critical considerations: Move now or retrofit later: Falling a release cycle behind could force costly rewrites when assistant-first interfaces become default. Tap into revolutionary potential: For organizations seeking to embrace the most powerful AI, leveraging Google’s “world model” research, multimodal capabilities (like Veo 3 and Imagen 4 showcased by Woodward at I/O), and the AGI trajectory promised by Google offers a path to potentially significant innovation. Prepare for a new interaction paradigm: Success for Google’s “universal assistant” would mean a primary new interface for services and data. Enterprises should strategize for integration via APIs and agentic frameworks for context-aware delivery. Factor in the long game (and its risks): Aligning with Google’s vision is a long-term commitment. The full “world model” and AGI are potentially distant horizons. Decision-makers must balance this with immediate needs and platform complexities. Contrast with focused alternatives: Pragmatic solutions from Microsoft offer tangible enterprise productivity now. Disruptive hardware-AI from OpenAI/IO presents another distinct path. A diversified strategy, leveraging the best of each, often makes sense, especially with the increasingly open agentic web allowing for such flexibility. These complex choices and real-world AI adoption strategies will be central to discussions at VentureBeat’s Transform 2025 next month. The leading independent event brings enterprise technical decision-makers together with leaders from pioneering companies to share firsthand experiences on platform choices – Google, Microsoft, and beyond – and navigating AI deployment, all curated by the VentureBeat editorial team. With limited seating, early registration is encouraged. Google’s defining offensive: shaping the future or strategic overreach? Google’s I/O spectacle was a strong statement: Google signalled that it intends to architect and operate the foundational intelligence of the AI-driven future. Its pursuit of a “world model” and its AGI ambitions aim to redefine computing, outflank competitors, and secure its dominance. The audacity is compelling; the technological promise is immense. The big question is execution and timing. Can Google innovate and integrate its vast technologies into a cohesive, compelling experience faster than rivals solidify their positions? Can it do so while transforming search and navigating regulatory challenges? And can it do so while focused so broadly on both consumers and business – an agenda that is arguably much broader than that of its key competitors? The next few years will be pivotal. If Google delivers on its “world model” vision, it may usher in an era of personalized, ambient intelligence, effectively becoming the new operational layer for our digital lives. If not, its grand ambition could be a cautionary tale of a giant reaching for everything, only to find the future defined by others who aimed more specifically, more quickly.  Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More We’re seeing AI evolve fast. It’s no longer just about building a single, super-smart model. The real power, and the exciting frontier, lies in getting multiple specialized AI agents to work together. Think of them as a team of expert colleagues, each with their own skills — one analyzes data, another interacts with customers, a third manages logistics, and so on. Getting this team to collaborate seamlessly, as envisioned by various industry discussions and enabled by modern platforms, is where the magic happens. But let’s be real: Coordinating a bunch of independent, sometimes quirky, AI agents is hard. It’s not just building cool individual agents; it’s the messy middle bit — the orchestration — that can make or break the system. When you have agents that are relying on each other, acting asynchronously and potentially failing independently, you’re not just building software; you’re conducting a complex orchestra. This is where solid architectural blueprints come in. We need patterns designed for reliability and scale right from the start. The knotty problem of agent collaboration Why is orchestrating multi-agent systems such a challenge? Well, for starters: They’re independent: Unlike functions being called in a program, agents often have their own internal loops, goals and states. They don’t just wait patiently for instructions. Communication gets complicated: It’s not just Agent A talking to Agent B. Agent A might broadcast info Agent C and D care about, while Agent B is waiting for a signal from E before telling F something. They need to have a shared brain (state): How do they all agree on the “truth” of what’s happening? If Agent A updates a record, how does Agent B know about it reliably and quickly? Stale or conflicting information is a killer. Failure is inevitable: An agent crashes. A message gets lost. An external service call times out. When one part of the system falls over, you don’t want the whole thing grinding to a halt or, worse, doing the wrong thing. Consistency can be difficult: How do you ensure that a complex, multi-step process involving several agents actually reaches a valid final state? This isn’t easy when operations are distributed and asynchronous. Simply put, the combinatorial complexity explodes as you add more agents and interactions. Without a solid plan, debugging becomes a nightmare, and the system feels fragile. Picking your orchestration playbook How you decide agents coordinate their work is perhaps the most fundamental architectural choice. Here are a few frameworks: The conductor (hierarchical): This is like a traditional symphony orchestra. You have a main orchestrator (the conductor) that dictates the flow, tells specific agents (musicians) when to perform their piece, and brings it all together. This allows for: Clear workflows, execution that is easy to trace, straightforward control; it is simpler for smaller or less dynamic systems. Watch out for: The conductor can become a bottleneck or a single point of failure. This scenario is less flexible if you need agents to react dynamically or work without constant oversight. The jazz ensemble (federated/decentralized): Here, agents coordinate more directly with each other based on shared signals or rules, much like musicians in a jazz band improvising based on cues from each other and a common theme. There might be shared resources or event streams, but no central boss micro-managing every note. This allows for: Resilience (if one musician stops, the others can often continue), scalability, adaptability to changing conditions, more emergent behaviors. What to consider: It can be harder to understand the overall flow, debugging is tricky (“Why did that agent do that then?”) and ensuring global consistency requires careful design. Many real-world multi-agent systems (MAS) end up being a hybrid — perhaps a high-level orchestrator sets the stage; then groups of agents within that structure coordinate decentrally. For agents to collaborate effectively, they often need a shared view of the world, or at least the parts relevant to their task. This could be the current status of a customer order, a shared knowledge base of product information or the collective progress towards a goal. Keeping this “collective brain” consistent and accessible across distributed agents is tough. Architectural patterns we lean on: The central library (centralized knowledge base): A single, authoritative place (like a database or a dedicated knowledge service) where all shared information lives. Agents check books out (read) and return them (write). Pro: Single source of truth, easier to enforce consistency. Con: Can get hammered with requests, potentially slowing things down or becoming a choke point. Must be seriously robust and scalable. Distributed notes (distributed cache): Agents keep local copies of frequently needed info for speed, backed by the central library. Pro: Faster reads. Con: How do you know if your copy is up-to-date? Cache invalidation and consistency become significant architectural puzzles. Shouting updates (message passing): Instead of agents constantly asking the library, the library (or other agents) shouts out “Hey, this piece of info changed!” via messages. Agents listen for updates they care about and update their own notes. Pro: Agents are decoupled, which is good for event-driven patterns. Con: Ensuring everyone gets the message and handles it correctly adds complexity. What if a message is lost? The right choice depends on how critical up-to-the-second consistency is, versus how much performance you need. Building for when stuff goes wrong (error handling and recovery) It’s not if an agent fails, it’s when. Your architecture needs to anticipate this. Think about: Watchdogs (supervision): This means having components whose job it is to simply watch other agents. If an agent goes quiet or starts acting weird, the watchdog can try restarting it or alerting the system. Try again, but be smart (retries and idempotency): If an agent’s action fails, it should often just try again. But, this only works if the action is idempotent. That means doing it five times has the exact same result as doing it once (like setting a value, not incrementing it). If actions aren’t idempotent, retries can cause chaos. Cleaning up messes (compensation): If Agent A did something successfully, but Agent B (a later step in the process) failed, you might need to “undo” Agent A’s work. Patterns like Sagas help coordinate these multi-step, compensable workflows. Knowing where you were (workflow state): Keeping a persistent log of the overall process helps. If the system goes down mid-workflow, it can pick up from the last known good step rather than starting over. Building firewalls (circuit breakers and bulkheads): These patterns prevent a failure in one agent or service from overloading or crashing others, containing the damage. Making sure the job gets done right (consistent task execution) Even with individual agent reliability, you need confidence that the entire collaborative task finishes correctly. Consider: Atomic-ish operations: While true ACID transactions are hard with distributed agents, you can design workflows to behave as close to atomically as possible using patterns like Sagas. The unchanging logbook (event sourcing): Record every significant action and state change as an event in an immutable log. This gives you a perfect history, makes state reconstruction easy, and is great for auditing and debugging. Agreeing on reality (consensus): For critical decisions, you might need agents to agree before proceeding. This can involve simple voting mechanisms or more complex distributed consensus algorithms if trust or coordination is particularly challenging. Checking the work (validation): Build steps into your workflow to validate the output or state after an agent completes its task. If something looks wrong, trigger a reconciliation or correction process. The best architecture needs the right foundation. The post office (message queues/brokers like Kafka or RabbitMQ): This is absolutely essential for decoupling agents. They send messages to the queue; agents interested in those messages pick them up. This enables asynchronous communication, handles traffic spikes and is key for resilient distributed systems. The shared filing cabinet (knowledge stores/databases): This is where your shared state lives. Choose the right type (relational, NoSQL, graph) based on your data structure and access patterns. This must be performant and highly available. The X-ray machine (observability platforms): Logs, metrics, tracing – you need these. Debugging distributed systems is notoriously hard. Being able to see exactly what every agent was doing, when and how they were interacting is non-negotiable. The directory (agent registry): How do agents find each other or discover the services they need? A central registry helps manage this complexity. The playground (containerization and orchestration like Kubernetes): This is how you actually deploy, manage and scale all those individual agent instances reliably. How do agents chat? (Communication protocol choices) The way agents talk impacts everything from performance to how tightly coupled they are. Your standard phone call (REST/HTTP): This is simple, works everywhere and good for basic request/response. But it can feel a bit chatty and can be less efficient for high volume or complex data structures. The structured conference call (gRPC): This uses efficient data formats, supports different call types including streaming and is type-safe. It is great for performance but requires defining service contracts. The bulletin board (message queues — protocols like AMQP, MQTT): Agents post messages to topics; other agents subscribe to topics they care about. This is asynchronous, highly scalable and completely decouples senders from receivers. Direct line (RPC — less common): Agents call functions directly on other agents. This is fast, but creates very tight coupling — agent need to know exactly who they’re calling and where they are. Choose the protocol that fits the interaction pattern. Is it a direct request? A broadcast event? A stream of data? Putting it all together Building reliable, scalable multi-agent systems isn’t about finding a magic bullet; it’s about making smart architectural choices based on your specific needs. Will you lean more hierarchical for control or federated for resilience? How will you manage that crucial shared state? What’s your plan for when (not if) an agent goes down? What infrastructure pieces are non-negotiable? It’s complex, yes, but by focusing on these architectural blueprints — orchestrating interactions, managing shared knowledge, planning for failure, ensuring consistency and building on a solid infrastructure foundation — you can tame the complexity and build the robust, intelligent systems that will drive the next wave of enterprise AI. Nikhil Gupta is the AI product management leader/staff product manager at Atlassian. Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Basketball has March Madness. Tech has the Consumer Electronics Show. AI has been waiting for its big moment—and this week may finally be it. With Microsoft’s Build and Google’s I/O developer conferences happening back-to-back, it was already primed to be a big week. Microsoft announced 50 new AI tools alone, and Google followed up with its own slate just a day later. Then, out of the blue, Anthropic dogpiled with Claude 4, the latest version of its large language model (LLM), on Thursday. While the maelstrom of announcements included some gee-whiz trinkets (we’re looking at you, Google Virtual Try-On), anyone looking to build a business with AI should find plenty to look forward to and even some new tools to use immediately. Struggling to keep up? Here are the biggest announcements from every company, and how they’ll reshape the AI landscape in the coming months. Microsoft wants AI agents to talk to one another By giving AI the power to perform work like a human rather than simply talking like one, agents represent an obvious next step for LLMs. But there’s been one major caveat holding them back: They can’t easily interact with one another. An agentic AI that books plane tickets for business travel and another that books hotels sounds great, until you land in London with a hotel room in Madrid. Microsoft took a major step in resolving this impasse by adopting Model Context Protocol (MCP), a standard way for different agents – even those using different LLMs – to communicate. Anthropic actually created the standard in Nov. 2024. Still, Microsoft’s adoption means it’s well on its way to becoming a fixture of future agentic architecture, as HTML was for the open web. Microsoft also added MCP to Azure AI Foundry, its tool for creating AI apps, so users can begin building agents that interact with one another immediately. So what? Agentic AI remains in its infancy, but a widely adopted standard will pave the way for the next generation of agentic tools. Standardization between competitors means you’ll have your pick of the litter on future LLMs when automating processes, rather than getting locked into a single company’s ecosystem. Claude 4 makes coders swoon With just a 3.3% share of the generative AI market, ChatGPT and Gemini often overshadow Claude. However, developers won’t want to sleep on Opus 4 and Sonnet 4, which arrived unexpectedly on Thursday with some major coding bragging rights. Perhaps most impressively, Claude 4 boasts marathon runtimes of up to seven hours in its “extended thinking” mode, which allows it to take thousands of steps and use tools like web search. Anthropic claims it will also explore more approaches, catch more errors, and break down its reasoning for more complex problems. With these improvements, Claude Opus 4 shot to the top of the popular SWE-bench software engineering benchmark with a score of 72.5%, besting both OpenAI o3 (69.1%) and Gemini 2.5 Pro (63.2%). So what? While benchmarks don’t always tell the whole story, Claude has already earned a reputation as the LLM of choice for developers. Claude 4 further cements that reputation with improvements for the software engineering community, which will help differentiate it from its more general-purpose peers. Google AI Mode upends search Google debuted plenty of consumer AI at IO 2025, from the aforementioned virtual try-ons to Google Beam, which turns 2D video streams into live, hologram-like models with the help of six different camera angles and a lot of AI. However, the most consequential change for enterprises may well be AI Mode for search. Like AI Overviews before it, AI Mode integrates Gemini into the search experience much more thoroughly. When you activate an AI Mode search, Google executes a “query fan-out technique,” which breaks your query into multiple searches and executes them simultaneously, then stitches together the results. While this mode was previously available for Google Labs users, this week, it’s going mainstream. So what? Even if you’re not using AI Mode personally, any change to Google Search sends ripples into the pond for the millions of businesses that depend on Google Search to draw eyeballs. AI Overviews upended the search engine optimization (SEO) industry, and AI Mode may be even more dramatic. The way most people find information online is changing, and fast. Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More In the first generation of the web, back in the late 1990s, search was okay but not great, and it wasn’t easy to find things. That led to the rise of syndication protocols in the early 2000s, with Atom and RSS (Really Simple Syndication) providing a simplified way for website owners to make headlines and other content easily available and searchable. In the modern era of AI, a new group of protocols is emerging to serve the same basic purpose. This time, instead of making sites easier for humans to find, it’s all about making websites easier for AI. Anthropic’s Model Control Protocol (MCP), Google‘s Agent2Agent and large language models/ LLMs.txt are among the existing efforts. The newest protocol is Microsoft’s open-source NLWeb (natural language web) effort, which was announced during the Build 2025 conference. NLWeb is also directly linked to the first generation of web syndication standards, as it was conceived and created by RV Guha, who helped create RSS, RDF (Resource Description Framework) and schema.org. NLWeb enables websites to easily add AI-powered conversational interfaces, effectively turning any website into an AI app where users can query content using natural language. NLWeb isn’t necessarily about competing with other protocols; rather, it builds on top of them. The new protocol uses existing structured data formats like RSS, and each NLWeb instance functions as an MCP server. “The idea behind NLWeb is it is a way for anyone who has a website or an API already to very easily make their website or their API an agentic application,” Microsoft CTO Kevin Scott said during his Build 2025 keynote. “You really can think about it a little bit like HTML for the agentic web.” How NLWeb works to AI-enable the web for enterprises NLWeb transforms websites into AI-powered experiences through a straightforward process that builds on existing web infrastructure while leveraging modern AI technologies. Building on existing data: The system begins by leveraging structured data that websites already publish, including markup, RSS feeds and other semi-structured formats that are commonly embedded in web pages. This means publishers don’t need to rebuild their content infrastructure completely. Data processing and storage: NLWeb includes tools for adding this structured data to vector databases, which enable efficient semantic search and retrieval. The system supports all major vector database options, allowing developers to choose the solution that best fits their technical requirements and scale. AI enhancement layer: LLMs then enhance this stored data with external knowledge and context. For instance, when a user queries about restaurants, the system automatically layers on geographic insights, reviews and related information by combining the vectorized content with LLM capabilities to provide comprehensive, intelligent responses rather than simple data retrieval. Universal interface creation: The result is a natural language interface that serves both human users and AI agents. Visitors can ask questions in plain English and receive conversational responses, while AI systems can programmatically access and query the site’s information through the MCP framework. This approach allows any website to participate in the emerging agentic web without requiring extensive technical overhauls. It makes AI-powered search and interaction as accessible as creating a basic webpage was in the early days of the internet. The emerging AI protocol landscape brings many choices to enterprises There are a lot of different protocols emerging in the AI space; not all do the same thing. Google’s Agent2Agent, for example, is all about enabling agents to talk to each other. It’s about orchestrating and communicating agentic AI and is not particularly focused on AI-enabling existing websites or AI content. Maria Gorskikh, founder and CEO of AIA and a contributor to the Project NANDA team at MIT, explained to VentureBeat that Google’s A2A enables structured task passing between agents using defined schemas and lifecycle models.  “While the protocol is open-source and model-agnostic by design, its current implementations and tooling are closely tied to Google’s Gemini stack — making it more of a backend orchestration framework than a general-purpose interface for web-based services,” she said. Another emerging effort is LLMs.txt. Its goal is to help LLMs better access web content. While on the surface, it might sound somewhat like NLWeb, it’s not the same thing. “NLWeb doesn’t compete with LLMs.txt; it is more comparable to web scraping tools that try to deduce intent from a website,” Michael Ni, VP and Principal Analyst at Constellation Research told VentureBeat. Krish Arvapally, co-founder and CTO of Dappier, explained to VentureBeat that LLMs.txt provides a markdown-style format with training permissions that helps LLM crawlers ingest content appropriately. NLWeb focuses on enabling real-time interactions directly on a publisher’s website. Dappier has its own platform that automatically ingests RSS feeds and other structured data, then delivers branded, embeddable conversational interfaces. Publishers can syndicate their content to their data marketplace. MCP is the other big protocol, and it is increasingly becoming a de facto standard and a foundational element of NLWeb. Fundamentally, MCP is an open standard for connecting AI systems with data sources. Ni explained that in Microsoft’s view, MCP is the transport layer, where, together, MCP and NLWeb provide the HTML and TCP/IP of the open agentic web. Forrester Senior Analyst Will McKeon-White sees a number of advantages for NLWeb over other options. “The main advantage of NLWeb is better control over how AI systems ‘see’ the pieces that make up websites, allowing for better navigation and more complete understanding of the tooling,” McKeon-White told VentureBeat. “This could reduce both errors from systems misunderstanding what they’re seeing on websites, as well as reduce interface rework.” Early adopters already see the promise of NLWeb for enterprise agentic AI Microsoft didn’t just throw NLWeb over the proverbial wall and hope someone would use it.  Microsoft already has multiple organizations engaged and using NLWeb, including Chicago Public Media, Allrecipes, Eventbrite, Hearst (Delish), O’Reilly Media, Tripadvisor and Shopify. Andrew Odewahn, Chief Technology Officer at O’Reilly Media is among the early adopters and sees real promise for NLWeb. “NLWeb leverages the best practices and standards developed over the past decade on the open web and makes them available to LLMs,” Odewahn told VentureBeat. “Companies have long spent time optimizing this kind of metadata for SEO and other marketing purposes, but now they can take advantage of this wealth of data to make their own internal AI smarter and more capable with NLWeb.” In his view, NLWeb is valuable for enterprises both as consumers of public information and publishers of private information. He noted that nearly every company has sales and marketing efforts where they might need to ask, “What does this company do?” or “What is this product about?”  “NLWeb provides a great way to open this information to your internal LLMs so that you don’t have to go hunting and pecking to find it,” Odewahn said. “As a publisher, you can add your own metadata using schema.org standard and use NLWeb internally as an MCP server to make it available for internal use.” Using NLWeb isn’t necessarily a heavy lift, either. Odewahn noted that many organizations are probably already using many of the standards NLWeb relies on.  “There’s no downside in trying it out now since NLWeb can run entirely within your infrastructure,” he said. “It’s open source software meeting the best in open source data, so you have nothing to lose and a lot to gain from trying it now.” Should enterprises jump on NLWeb right now, or wait? Constellation Research Analyst Michael Ni has a somewhat positive viewpoint on NLWeb. However, that doesn’t mean enterprises need to adopt it immediately. Ni noted that NLWeb is in the very early stages of maturity and enterprises should expect 2-3 years for any substantial adoption. He suggests that leading-edge companies with specific needs, such as active marketplaces, can look to pilot with the ability to engage and help shape the standard. “It’s a visionary specification with clear potential, but it needs ecosystem validation, implementation tooling, and reference integrations before it can reach mainstream enterprise pilots,” Ni said. Others have a somewhat more aggressive viewpoint on adoption. Gorskikh suggests taking an accelerated approach to ensure your enterprise doesn’t fall behind. “If you’re an enterprise with a large content surface, internal knowledge base, or structured data, piloting NLWeb now is a smart and necessary step to stay ahead,” she said. “This isn’t a wait-and-see moment — it’s more like the early adoption of APIs or mobile apps.” That said, she noted that regulated industries need to tread carefully. Sectors like insurance, banking and healthcare should hold off on production use until there’s a neutral, decentralized verification and discovery system in place. There are already early-stage efforts addressing this — such as the NANDA project at MIT that Gorskikh participates in, which is building an open, decentralized registry and reputation system for agentic services. What does this all mean to enterprise AI leaders? For enterprise AI leaders, NLWeb is a watershed moment and a technology that should not be ignored. AI is going to interact with your site, and you need to AI enable it. NLWeb is one way that will be particularly attractive to publishers, much like RSS became a must-have for all websites in the early 2000s. In a few years, users will just expect it to be there; they will expect to be able to search and find things, while agentic AI systems will need to be able to access the content as well.  That’s the promise of NLWeb. Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More It was a big week for AI announcements following events from Microsoft, Google, and Anthropic. But OpenAI is finishing things out with news of its own. And no, we’re not just talking about its $6.5 billion acquisition of Jony Ive’s design team to lead a new hardware effort, “io” at OpenAI. Today, the company upgraded its Operator autonomous web browsing and cursor controlling agent within ChatGPT from using the prior GPT-4o multimodal large language model to the newer and more powerful o3 reasoning model. The update, released globally today, May 23, 2025, is available as a “research preview” to paying subscribers of OpenAI’s $200 USD-monthly ChatGPT Pro plan. Basically, that is OpenAI’s way of saying it’s not a fully “sanded down” or perfected product yet — it may still have kinks and issues. But with rival Google offering its own top tier AI subscription bundle for a price of nearly $250 USD regularly (currently running a discount down to $125 for the first three months) to access its latest Gemini multimodal, Imagen image generation, and Veo video generation models, suddenly OpenAI’s ChatGPT Pro plan seems more affordable by comparison. What is OpenAI’s Operator and what is it for? Operator first debuted in January 2025 as OpenAI’s initial step into semi-autonomous agents, specifically Computer Using Agents (CUAs). The idea is to go beyond the chatbot interface of ChatGPT and allow OpenAI’s powerful AI models to start taking more actions on behalf of the user. Thus, Operator was designed to autonomously point, click, scroll, and type to complete web-based tasks such as booking dinner reservations, compiling shopping lists, or ordering event tickets. This agentic capability allows it to complete user tasks directly through a browser interface, from booking reservations to gathering online data. For safety, privacy and security purposes, Operator didn’t use any existing web browser on a user’s PC or Mac. Instead, it ran in a cloud-hosted virtual browser accessible via a standalone site—operator.chatgpt.com—where users could input requests and observe the agent perform tasks in real time. It combined vision, reasoning, and interaction capabilities based on GPT-4o, marking a new direction for OpenAI in agentic AI. The product was launched as a research preview for ChatGPT Pro subscribers and featured built-in safety measures like user confirmations, Watch Mode, and restrictions on high-risk web platforms. It was also being tested in enterprise contexts, including travel planning and civic services, demonstrating its potential across both consumer and business environments. o3 offers improved accuracy, structure, and success rates With this update, OpenAI aims to enhance performance across several key dimensions. The new o3-based Operator demonstrates improved persistence and accuracy during browser interactions. In practical terms, this means it is more likely to complete user tasks successfully and with less need for correction or repetition. Moreover, users can expect responses that are clearer, more structured, and more comprehensive. In comparative evaluations, the new model shows a distinct preference advantage over its predecessor. Human preference studies reveal that users favor the o3 model for its style, comprehensiveness, and clarity. It also performs strongly in instruction following and efficiency, though results for factual correctness are more balanced between versions. Performance on third-party evaluation benchmarks reflects these enhancements. On the OSWorld benchmark that measures completion of browser-based tasks, the o3 model scores 42.9 compared to 38.1 for the previous version. However, OpenAI notes that due to limitations in the automated grading system, the actual performance gain could be closer to 20 percentage points! On WebArena, the new model achieved a score of 62.9, up from 48.1. The most dramatic improvement appears on the GAIA benchmark, where the o3 model scores 62.2, vastly surpassing the prior model’s 12.3. Side-by-side task comparisons further illustrate these gains. In one example involving a restaurant booking request, the new model provided a clearer and more detailed list of available reservations, including locations, Michelin ratings, and seating notes, presented in a well-formatted table. The previous version, while functional, delivered less information in a less organized manner, according to an image included with the new o3 Operator release notes: The o3 model also inherits the safety measures introduced with earlier versions, with further fine-tuning for its role as an agentic system. OpenAI has integrated enhanced training against harmful task execution, prompt injection vulnerabilities, and mistakes involving user intent. Evaluations show that the model now confirms 94% of sensitive actions before executing them, with 100% confirmation in financial transactions. Prompt injection susceptibility has also decreased from 23% to 20%. Notably, the o3 Operator maintains a cautious boundary on certain high-risk web interactions, such as email or financial platforms, where it may require user supervision via Watch Mode or explicitly refuse to proceed. These measures are part of a layered approach to safety that combines model-level robustness with real-time monitoring. While the upgrade to Operator marks a technical improvement, it also reflects OpenAI’s ongoing commitment to responsible AI deployment. The system’s ability to take real-world actions introduces new risks, and the development team continues to refine its safety protocols accordingly. According to OpenAI’s updated o3 system card documentation, the model remains below high-risk capability thresholds in categories such as biological and chemical misuse and has no native coding environment or terminal access, further reducing potential misuse vectors. Operator remains a research preview and is accessible only to ChatGPT Pro users. The Responses API version of Operator will continue to be based on the GPT-4o model, at least for now. Implications for enterprise technical decision-makers The upgraded Operator stands to significantly enhance the workflows of professionals in AI engineering, orchestration, data management, and IT security. For those building or maintaining machine learning models, the model’s improved accuracy and structured outputs reduce the overhead of test validation and troubleshooting. In orchestration contexts, it offers a practical, reliable tool for automating browser-based components of complex pipelines. Data engineers can delegate manual web interactions—such as data verification and scraping—with more confidence, freeing time for higher-level optimization work. Security professionals, meanwhile, gain a safer way to simulate user behavior in audits and incident response exercises, thanks to the model’s layered safety mechanisms. Across these disciplines, the o3-based Operator introduces both a capability upgrade and a risk mitigation framework, making it a practical addition to the modern technical toolkit. Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

A24 and Bandai Namco have named Alex Garland as the director and writer of the live-action Elden Ring adaptation.Read More

StreamElements' latest report shows that Call of Duty: Warzone got a viewership spike thanks to the re-launch of Verdansk.Read More

Omeda Studios announced a full competitive summer calendar for their free-to-play third-person action MOBA Predecessor.Read More

Savvy Games Group has made a lot of news as it has built the newest financial empire in games with acquisitions of companies.Read More

Video game adaptations continue to make a big splash in Hollywood.Read More

PlaySafe ID -- a platform for gamers that keeps cheaters, hackers, bots, and predators out of games -- has raised $1.12 million in funding.Read More

A new benchmark can test how much LLMs become sycophants, and found that GPT-4o was the most sycophantic of the models tested.Read More

Yield Guild Games Association announced that it has expanded into Web3 game publishing with the launch of the causal Degen game LOL Land. With the launch of its first title, LOL Land, Yield Guild Games Association (Yield Guild) has announced YGG Play, a new game publishing unit dedicated to lighthearted, addictive games that embrace the chaos and culture of crypto degeneracy. With YGG Play, the association aims to define a new genre in gaming: Casual Degen. The debut game — a browser-based casual board game featuring members of the Pudgy Penguins community as playable characters — with 116,000 pre-registrations prior to its exclusive launch on Abstract Chain. By playing, collecting points, and unlocking NFTs in LOL Land, players have the chance to win part of a prize pool funded on May 1, valued at $10 million in YGG token rewards. “Over the past four years, we’ve seen and played hundreds of games, and Yield Guild has invested in many of them. Through that experience, we’ve come to know what’s fun, what works and what doesn’t, and that’s led us to develop some very strong opinions on how to best serve the crypto market,” said Gabby Dizon, cofounder of Yield Guild, in a statement. “YGG Play, which is specifically targeting the Casual Degen market, is our expression of everything we’ve learned, starting with our first game, LOL Land.” LOL Land features four thematic boards, each with a unique mechanic: YGG City, Beach Day, Carnival, and Ice World Wonderland — the latter being home to the Pudgy Penguins IP, where players can choose to playable characters based on notable members from Pudgy’s community, including Aaron Teng, Rusk0f, KP, Burnasso, and Orange. Dizon is also a playable game character, and other themes are under development, including Pirate Cove and Parkside Story. The Abstract Chain, the exclusive platform for LOL Land, was created by Igloo Inc., the parent company of Pudgy Penguins. Its consumer-first approach makes it easier for players to get into Web3 games by using chain abstraction and integrating friendlier onboarding mechanisms such as social logins and passkeys. The platform has already succeeded with several launches, including Myriad, Roach Racing Club, and Gacha. The launch also coincides with the YGG token being available on Abstract. “LOL Land is a game built by degens for degens,” said Luca Netz, CEO of Pudgy Penguins, in a statement. “It takes everything we’ve come to love about playing Web3 games and puts it into a casual pick-up-and-play experience. We’re really excited to be a part of YGG’s first game and its exclusive launch on Abstract.”  LOL Land offers two gameplay modes to suit different player preferences. The free mode allows players to enjoy unlimited rolls at no cost, though it does not include YGG token rewards. In the premium mode, players can purchase premium rolls, with points earned through gameplay eligible for redemption in YGG tokens. LOL Land is only available to players 18 years or older and in certain countries. The game is available to play now at LOL.Land.  I asked Dizon a few questions. In a message to GamesBeat, Dizon said YGG is a guild protocol that provides guilds with a suite of tools to organize onchain. “Our core community’s Discord currently has over 112,000 members and our Guild Advancement Program (GAP) has seen over 300,000 community members use our questing platform,” Dizon said. “Since launching on Base in November last year, we’ve had more than 75 Onchain Guilds sign up for our platform’s pilot. We also have a global network of guilds spanning 11 regions, represented by our guild partners YGG Pilipinas (Philippines), W3GG (Southeast Asia), OLA GG (Latin America), KGeN (India, Brazil, Saudi Arabia), SKYGG (South Korea), YGG Japan, YGG China, AMG DAO (Europe), and NYXL (North America).” Asked why YGG decided to make a game, Dizon said, “We decided to launch our game publishing arm because we want to establish the “Casual Degen” category in Web3 games. With our history and our deep connection to the community and its crypto-native, grassroots, play-first culture, game publishing felt like a natural expansion.” Gabby Dizon is cofounder of YGG. YGG has grown into a hub for players to discover new games and contribute their skills to different projects and ecosystems. Dizon added, “We represent highly skilled communities that can bootstrap the launch of a game by connecting them with dedicated players as well as groups that can provide the services these early-stage projects need like community management, marketing and local events, content creation, education, esports, game testing, and more.” And he said, “YGG is uniquely positioned as a publisher because we know what the community looks for in games and we know where these games need help at launch—we hear it directly from the players. So, expanding into publishing and making the first game ourselves made sense.” I asked Dizon how YGG chose the game. He noted that before he got into crypto and cofounded YGG, he was a game developer for over 20 years. “I was actually building casual games for much of my career. It’s nearly five years since we founded YGG, and we’ve played hundreds of Web3 games and invested in many of them,” Dizon said. “Building our ecosystem and running events like the YGG Play Summit has kept us closely aligned with our community’s interests. We understand what the YGG community wants: playing fun games and enjoying the rewards together.” And he added, “Based on our experience, we also knew that the Web3-native game market is still very underserved. These are the people who aren’t necessarily hardcore gamers but are in crypto for fun — or what we call degens. Hardcore Web3 gamers are actually a very small portion of the total Web3 user base. But the “Casual Degens” — that’s people trading memecoins, minting NFTs, or trading on exchanges — they’re a much larger market. And we know exactly how to target these markets.” Yield Guild Games (YGG) is a global collective of gamers, creators, and builders turning play into opportunity. Known for pioneering the Web3 guild model with the rise of the play-to-earn movement in 2020, YGG laid the foundation for a new kind of digital coordination: a network of onchain groups, each organized around their own purpose, and united by shared values of collaboration, progression, and ownership. Pre-registration is currently ongoing at https://www.lol.land/pre-register. GB Daily Stay in the know! Get the latest news in your inbox daily Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

Virtuos rebuild of Oblivion Remastered topped April's sales charts, with Microsoft's PlayStation launches also charting well.Read More

Bowman later edited his tweet and the following one in a thread to read as follows, but it still didn't convince the naysayers.Read More

Someone also appears to have published a full scrape of the Time article online on the news aggregator app Newsbreak.Read More

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Anthropic released Claude Opus 4 and Claude Sonnet 4 today, dramatically raising the bar for what AI can accomplish without human intervention. The company’s flagship Opus 4 model maintained focus on a complex open-source refactoring project for nearly seven hours during testing at Rakuten — a breakthrough that transforms AI from a quick-response tool into a genuine collaborator capable of tackling day-long projects. This marathon performance marks a quantum leap beyond the minutes-long attention spans of previous AI models. The technological implications are profound: AI systems can now handle complex software engineering projects from conception to completion, maintaining context and focus throughout an entire workday. Anthropic claims Claude Opus 4 has achieved a 72.5% score on SWE-bench, a rigorous software engineering benchmark, outperforming OpenAI’s GPT-4.1, which scored 54.6% when it launched in April. The achievement establishes Anthropic as a formidable challenger in the increasingly crowded AI marketplace. Comparative benchmarks show Claude 4 models (left) outperforming competitors across coding and reasoning tasks, with Claude Opus 4 achieving a 72.5% score on the critical SWE-bench test. (Credit: Anthropic) Beyond quick answers: the reasoning revolution transforms AI The AI industry has pivoted dramatically toward reasoning models in 2025. These systems work through problems methodically before responding, simulating human-like thought processes rather than simply pattern-matching against training data. OpenAI initiated this shift with its “o” series last December, followed by Google’s Gemini 2.5 Pro with its experimental “Deep Think” capability. DeepSeek’s R1 model unexpectedly captured market share with its exceptional problem-solving capabilities at a competitive price point. This pivot signals a fundamental evolution in how people use AI. According to Poe’s Spring 2025 AI Model Usage Trends report, reasoning model usage jumped fivefold in just four months, growing from 2% to 10% of all AI interactions. Users increasingly view AI as a thought partner for complex problems rather than a simple question-answering system. The share of reasoning messages surged in early 2025 as new AI models captured user interest. (Credit: Poe) Claude’s new models distinguish themselves by integrating tool use directly into their reasoning process. This simultaneous research-and-reason approach mirrors human cognition more closely than previous systems that gathered information before beginning analysis. The ability to pause, seek data, and incorporate new findings during the reasoning process creates a more natural and effective problem-solving experience. Dual-mode architecture balances speed with depth Anthropic has addressed a persistent friction point in AI user experience with its hybrid approach. Both Claude 4 models offer near-instant responses for straightforward queries and extended thinking for complex problems — eliminating the frustrating delays earlier reasoning models imposed on even simple questions. This dual-mode functionality preserves the snappy interactions users expect while unlocking deeper analytical capabilities when needed. The system dynamically allocates thinking resources based on the complexity of the task, striking a balance that earlier reasoning models failed to achieve. Memory persistence stands as another breakthrough. Claude 4 models can extract key information from documents, create summary files, and maintain this knowledge across sessions when given appropriate permissions. This capability solves the “amnesia problem” that has limited AI’s usefulness in long-running projects where context must be maintained over days or weeks. The technical implementation works similarly to how human experts develop knowledge management systems, with the AI automatically organizing information into structured formats optimized for future retrieval. This approach enables Claude to build an increasingly refined understanding of complex domains over extended interaction periods. The timing of Anthropic’s announcement highlights the accelerating pace of competition in advanced AI. Just five weeks after OpenAI launched its GPT-4.1 family, Anthropic has countered with models that challenge or exceed it in key metrics. Google updated its Gemini 2.5 lineup earlier this month, while Meta recently released its Llama 4 models featuring multimodal capabilities and a 10-million token context window. Each major lab has carved out distinctive strengths in this increasingly specialized marketplace. OpenAI leads in general reasoning and tool integration, Google excels in multimodal understanding, and Anthropic now claims the crown for sustained performance and professional coding applications. The strategic implications for enterprise customers are significant. Organizations now face increasingly complex decisions about which AI systems to deploy for specific use cases, with no single model dominating across all metrics. This fragmentation benefits sophisticated customers who can leverage specialized AI strengths while challenging companies seeking simple, unified solutions. Anthropic has expanded Claude’s integration into development workflows with the general release of Claude Code. The system now supports background tasks via GitHub Actions and integrates natively with VS Code and JetBrains environments, displaying proposed code edits directly in developers’ files. GitHub’s decision to incorporate Claude Sonnet 4 as the base model for a new coding agent in GitHub Copilot delivers significant market validation. This partnership with Microsoft’s development platform suggests large technology companies are diversifying their AI partnerships rather than relying exclusively on single providers. Anthropic has complemented its model releases with new API capabilities for developers: a code execution tool, MCP connector, Files API, and prompt caching for up to an hour. These features enable the creation of more sophisticated AI agents that can persist across complex workflows—essential for enterprise adoption. Transparency challenges emerge as models grow more sophisticated Anthropic’s April research paper, “Reasoning models don’t always say what they think,” revealed concerning patterns in how these systems communicate their thought processes. Their study found Claude 3.7 Sonnet mentioned crucial hints it used to solve problems only 25% of the time — raising significant questions about the transparency of AI reasoning. This research spotlights a growing challenge: as models become more capable, they also become more opaque. The seven-hour autonomous coding session that showcases Claude Opus 4’s endurance also demonstrates how difficult it would be for humans to fully audit such extended reasoning chains. The industry now faces a paradox where increasing capability brings decreasing transparency. Addressing this tension will require new approaches to AI oversight that balance performance with explainability — a challenge Anthropic itself has acknowledged but not yet fully resolved. A future of sustained AI collaboration takes shape Claude Opus 4’s seven-hour autonomous work session offers a glimpse of AI’s future role in knowledge work. As models develop extended focus and improved memory, they increasingly resemble collaborators rather than tools — capable of sustained, complex work with minimal human supervision. This progression points to a profound shift in how organizations will structure knowledge work. Tasks that once required continuous human attention can now be delegated to AI systems that maintain focus and context over hours or even days. The economic and organizational impacts will be substantial, particularly in domains like software development where talent shortages persist and labor costs remain high. As Claude 4 blurs the line between human and machine intelligence, we face a new reality in the workplace. Our challenge is no longer wondering if AI can match human skills, but adapting to a future where our most productive teammates may be digital rather than human. Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

Wildlife Studios and Winners Alliance today announced a major licensing partnership, bringing Naomi Osaka and other stars to the hit mobile game Tennis Clash. Tennis Clash has been played by more than 170 million users worldwide. Through the agreement, Wildlife Studios now holds rights to feature the names, images, and likenesses of select men’s and women’s players, marking the first time officially licensed players will appear in Tennis Clash. Launched in 2019, Tennis Clash is one of the most downloaded mobile sports games globally. Wildlife Studios – founded in 2011 – has released more than 60 games and surpassed 4 billion lifetime downloads across its portfolio. Starting May 23, users can compete as four-time Grand Slam champion Naomi Osaka, who debuts in a custom outfit inspired by her 2024 U.S. Open look. Wildlife Studios will introduce at least three more licensed player characters later this year, with more in development. The integration of real players brings even more realism and depth to the game, allowing fans to experience tennis in a more immersive, authentic, and personal way. Through interactive avatars, the sport’s top stars are connected with Tennis Clash’s massive global audience in ways that reimagine how tennis is played, promoted, and experienced. “Winners Alliance is proud to bring professional tennis players into a platform as dynamic and far-reaching as Tennis Clash,” said Eric Winston, president of Winners Alliance, in a statement. “Wildlife Studios has taken great care to reflect each athlete’s officially licensed identity, and this is just the beginning of what’s possible when gaming and athlete licensing come together at scale.” “We strive to make Tennis Clash as authentic and entertaining as possible, which is why we’re excited to partner with Winners Alliance to welcome some of the sport’s most beloved stars into the game,” said Giovanni Piffer, partnerships lead at Wildlife Studios, in a statement. “We’re eager to see how our players enjoy the new Naomi Osaka character and can’t wait to unveil the additional characters coming soon. Tennis Clash is giving gamers and tennis fans a new way to engage with the sport – closer, more personal, and more interactive than ever before.” Winners Alliance is a global, athlete-centric commercial solution dedicated to creating novel group licensing, sponsorship, partnership, content, investment and event opportunities for world-class athletes collectively. Wildlife is one of the world’s largest mobile gaming publishers. In over ten years, it has launched more than 60 games, including hits such as Sniper 3D, Zooba, Tennis Clash, War Machines and Colorfy. Founded by brothers Victor and Arthur Lazarte, the company’s games have been downloaded more than 4 billion times across the planet. GB Daily Stay in the know! Get the latest news in your inbox daily Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

AppOnBoard is unveiling Quvy, a tool that enables AI to test user acquisition schemes on simulated audiences to speed up a game's growth.Read More

Enchant is launching a new zero-equity accelerator for gaming and AI startups, with applications now open for its three-month program. The accelerator is dedicated to early-stage teams at the crossroads of gaming and artificial intelligence. Founded by veterans of Sony, Supercell, EA, Ubisoft and Playrix, the fully remote program gives founders hands-on product sprints, weekly one-on-one sessions with industry leaders and direct access to a global investor network — while letting every startup keep 100 percent of its cap-table. Enchant backs companies building games, gaming-industry tools, AI products and Web3 experiences. By stripping away office requirements and equity grabs, the accelerator concentrates on rapid product iteration, real-world traction and warm introductions to publishers, VCs and strategic partners, said Eugenii Unegovskiy, cofounder of Enchant, in an interview with GamesBeat. Rolling admission, personal start date Enchant is taking applications for its zero-equity accelerator. The program is an intensive 12-week track that begins the day a team is accepted. Founders apply whenever they’re ready; there is no need to wait for fixed cohorts. From day one, each company follows a milestone roadmap that includes product deep-dives, go-to-market workshops and Demo-Day preparation. “Founders shouldn’t have to wait for a calendar slot,” said Unegovskiy. “Acceptance triggers day-one sprints, live play-tests and build reviews shoulder-to-shoulder with veterans from triple-A studios. It’s real co-building, not classroom mentoring, and it means products hit the market in weeks, not months.” “Zero-equity is a strong benefit, with an even bigger advantage being a clean cap table that helps early gaming and AI teams close future rounds faster,” added Aakash Parameswaran, CEO of Fable Fox, a former Game Designer at Electronic Arts and an expert in Enchant, in a statement. “From refining core loops to designing live-service economies, we work in the trenches so founders stay player-first and ship games investors can trust.” Why zero-equity matters Enchant is taking flat fees for its work. Keeping the cap-table clean at the earliest stage helps founders retain control and stay more attractive for future Seed and Series A rounds, when dilution stakes are far larger. It also lets Enchant operate purely on the startup’s success, aligning incentives around product wins and fundraising milestones. Applications are reviewed on a rolling basis, and accepted teams can start immediately. Unegovskiy said that while running demo days for game startups at Subscrible, he realized something crucial: most early-stage founders don’t fail because the product is weak — they fail because the business around it isn’t ready. No go-to-market, no monetization strategy, no clear idea how much money they need — or why. “That’s why I teamed up with folks from Supercell, Ubisoft, Sony, EA, King, Xsolla, and Oracle — many of whom I’m sure you know or have crossed paths with — along with the creators of Tetris, Angry Birds, and Cut the Rope, to launch Enchant, a new kind of accelerator for gaming and AI startups,” Unegovskiy said. Without cohorts or equity, the company will provide focused support that helps founders move forward. “We help with pitch training, fundraising prep, GTM and monetization strategy, grant and UA support, legal and marketing guidance — plus access to tools, credits, and partner services worth over $500,000 (cloud credits, analytics, dev tools, etc.),” Unegovskiy said. “It’s a simple model: small monthly fee + a success fee if we help a team raise or grow — no early dilution.” The team includes Tim Horton, an experienced gaming executive with 16+ years in leadership, business development, and partnerships within the global gaming industry. Horton helps optimize operations, drive growth, and deliver high-quality projects in mobile gaming, triple-A art, console porting, and game development. Sutton Trout specializes in leadership/strategic partnerships/business development/goal attainment and revenue. He is an action-oriented leader with 20 years of experiences in games and entertainment. He has worked at Supercell, Sony and WB Games. And he helped with complex projects and major launches like Hogwarts Legacy. Also on the team is Andrey Feinberg (Kuznetsov), former CIO at GDEV and Nexters. He has more than 10 years of experience in gaming, tech, and internet industries. He helped closed 40+ deals worth over $800 million, and he is now an independent advisor focused on fundraising, M&A, and strategy for gaming and interactive entertainment companies. Unegovskiy is one of the experts, as he is a six-time founder with experience in game development and AI. Other experts include Amir Shaked, Igor Diev, Jun Qian, Oded Lavie, Martin Hoszowski, Kevin Beardslee, Yosef Warburg, Wes Harris, Elvis Fernandes, Han Lee, Faisal Nazir, Tim Gerritsen, Hovav Shoshan, Igor Bazhanov, Cipto Adiguno, Casey Dickinson, Rosario Basilotta, Aakash Parameswaran, Mariam Japaridze and Eugene Volikov. Enchant expert Aakash Parameswaran, CEO of Fable Fox. By not taking a stake in the company, Unegovskiy said the company focuses on working side-by-side with founders to improve everything from pitch to revenue strategy. It takes a flat monthly fee, plus a success fee if it helps a team raise money or grow. The fee is typically 5% to 10% of the raised amount or a share of revenue growth. “We also reserve the right to invest in future rounds on founder-friendly terms — aligned with long-term support, not short-term ownership,” said Unegovskiy. He also said mentors and advisserss are similiarly aligned to the startup’s success.The focus of games and AI at Enchant. The company just opened applications and are starting to see a diverse mix of early-stage founders — from game studios to AI-tool builders, across multiple regions. A significant share of interest is coming from projects working at the intersection of gaming and AI — blending creative ambition with scalable infrastructure. “We’re seeing strong indicators that the next wave of gaming startups will be deeply technical, globally minded, and looking for business-first support without early dilution,” Unegovskiy said. GB Daily Stay in the know! Get the latest news in your inbox daily Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

Crunchyroll, the global brand for anime, and Discord are teaming up to bring the beloved series My Hero Academia to the Discord platform.Read More

Netmarble announced the launch of action-adventure RPG title Game of Thrones: Kingsroad on mobile and PC platforms.Read More

Logitech G announced the launch of the Logitech G522 gaming headset, a new headset with a focus on gameplay and personal expression.Read More

Mistplay has launched its reward-based user acquisition platform for mobile gamers on the iPhone in the U.S. Since its foundation in 2016, Montreal-based Mistplay has powered the leading rewards-based user acquisition platform for mobile gamers, delivering high-lifetime value (LTV) players at scale through its proven rewarded play model. The company announced its release of Mistplay for iPhone, bringing new dimensions of scale and inventory and offering more opportunities for publishers to acquire, monetize, and nurture loyal, high-value players. “Unveiling Mistplay for iPhone today marks a milestone we’ve long awaited – and one our partners have been asking for” said Tricia Han, CEO of Mistplay, in a statement. “After years of innovating rewarded advertising on Android, we’re thrilled to bring our rewarded play expertise to iOS in a unique way to solve real challenges for advertisers – reaching high-value users in a privacy-centric landscape, driving engagement beyond traditional ads, and scaling LTV. To our partners who’ve been with us on this journey: thank you. The best is yet to come.” Launched in 2016, Mistplay’s loyalty platform has scaled to millions of active users driving hundreds of millions in IAP spend across 9 key markets. With the arrival of Mistplay for iPhone, we’re introducing a powerful new reward system for iPhone users and unlocking new ad inventory for publishers. Tricia Han is CEO of Mistplay. Players can earn rewards for completing in-game events, go head-to-head in Tournaments to increase their earnings, and participate in game trivia and surveys. This approach enables publishers to hit their advertising performance goals and achieve ROI by acquiring high-value, loyal users. The commitment to innovation and development ensures that both Mistplay reward apps across iOS and Android platforms will continue to evolve as we continue to push the rewarded advertising and rewarded play industry forward for players and publishers alike. Mistplay for iPhone is currently available for iPhone devices in the U.S., with a continued global rollout planned for the rest of 2025. GB Daily Stay in the know! Get the latest news in your inbox daily Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

Google co-founder Sergey Brin makes surprise appearance at Google I/O, declaring "Gemini will be the very first AGI," revealing philosophical tensions with DeepMind CEO Hassabis who urges scientific caution in the high-stakes AI race.Read More

Voya Games raised $5 million in funding to support the growth of Craft World, a casual Web3 game.Read More

Support for remote Model Context Protocol servers, integration of image generation and Code Interpreter tools, and upgrades to file search...Read More

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Well-funded French AI model maker Mistral has consistently punched above its weight since its debut of its own powerful open source foundation model in fall 2023 — but it took some criticism among developers on X recently for its last release of a proprietary large language model (LLM) called Medium 3, which some viewed as betraying its open source roots and commitment. (Recall that open source models can be taken and adapted freely by anyone, while proprietary models must be paid for and their customization options are more limited and controlled by the model maker.) But today, Mistral is back and recommitting to the open source AI community, and AI-powered software development in particular, in a big way. The company has teamed up with open source startup All Hands AI, creators of Open Devin to release Devstral, a new open-source language model with 24-million parameters — much smaller than many rivals whose models are in the multibillions, and thus, requiring far less computing power such that it can be run on a laptop — purpose-built for agentic AI development. Unlike traditional LLMs designed for short-form code completions or isolated function generation, Devstral is optimized to act as a full software engineering agent—capable of understanding context across files, navigating large codebases, and resolving real-world issues. The model is now freely available under the permissive Apache 2.0 license, allowing developers and organizations to deploy, modify, and commercialize it without restriction. “We wanted to release something open for the developer and enthusiast community—something they can run locally, privately, and modify as they want,” said Baptiste Rozière, research scientist at Mistral AI. “It’s released under Apache 2.0, so people can do basically whatever they want with it.” Building upon Codestral Devstral represents the next step in Mistral’s growing portfolio of code-focused models, following its earlier success with the Codestral series. First launched in May 2024, Codestral was Mistral’s initial foray into specialized coding LLMs. It was a 22-billion-parameter model trained to handle over 80 programming languages and became well-regarded for its performance in code generation and completion tasks. The model’s popularity and technical strengths led to rapid iterations, including the launch of Codestral-Mamba—an enhanced version built on Mamba architecture—and most recently, Codestral 25.01, which has found adoption among IDE plugin developers and enterprise users looking for high-frequency, low-latency models. The momentum around Codestral helped establish Mistral as a key player in the coding-model ecosystem and laid the foundation for the development of Devstral—extending from fast completions to full-agent task execution. Outperforms larger models on top SWE benchmarks Devstral achieves a score of 46.8% on the SWE-Bench Verified benchmark, a dataset of 500 real-world GitHub issues manually validated for correctness. This places it ahead of all previously released open-source models and ahead of several closed models, including GPT-4.1-mini, which it surpasses by over 20 percentage points. “Right now, it’s by pretty far the best open model for SWE-bench verified and for code agents,” said Rozière. “And it’s also a very small model—only 24 billion parameters—that you can run locally, even on a MacBook.” “Compare Devstral to closed and open models evaluated under any scaffold—we find that Devstral achieves substantially better performance than a number of closed-source alternatives,” wrote Sophia Yang, Ph.D., Head of Developer Relations at Mistral AI, on the social network X. “For example, Devstral surpasses the recent GPT-4.1-mini by over 20%.” The model is finetuned from Mistral Small 3.1 using reinforcement learning and safety alignment techniques. “We started from a very good base model with Mistral’s small tree control, which already performs well,” Rozière said. “Then we specialized it using safety and reinforcement learning techniques to improve its performance on SWE-bench.” Built for the agentic era Devstral is not just a code generation model — it is optimized for integration into agentic frameworks like OpenHands, SWE-Agent, and OpenDevin. These scaffolds allow Devstral to interact with test cases, navigate source files, and execute multi-step tasks across projects. “We’re releasing it with OpenDevin, which is a scaffolding for code agents,” said Rozière. “We build the model, and they build the scaffolding — a set of prompts and tools that the model can use, like a backend for the developer model.” To ensure robustness, the model was tested across diverse repositories and internal workflows. “We were very careful not to overfit to SWE-bench,” Rozière explained. “We trained only on data from repositories that are not cloned from the SWE-bench set and validated the model across different frameworks.” He added that Mistral dogfooded Devstral internally to ensure it generalizes well to new, unseen tasks. Efficient deployment with permissive open license — even for enterprise and commercial projects Devstral’s compact 24B architecture makes it practical for developers to run locally, whether on a single RTX 4090 GPU or a Mac with 32GB of RAM. This makes it appealing for privacy-sensitive use cases and edge deployments. “This model is targeted toward enthusiasts and people who care about running something locally and privately—something they can use even on a plane with no internet,” Rozière said. Beyond performance and portability, its Apache 2.0 license offers a compelling proposition for commercial applications. The license permits unrestricted use, adaptation, and distribution—even for proprietary products—making Devstral a low-friction option for enterprise adoption. Detailed specifications and usage instructions are available on the Devstral-Small-2505 model card on Hugging Face. The model features a 128,000 token context window and uses the Tekken tokenizer with a 131,000 vocabulary. It supports deployment through all major open source platforms including Hugging Face, Ollama, Kaggle, LM Studio, and Unsloth, and works well with libraries such as vLLM, Transformers, and Mistral Inference. Available via API or locally Devstral is accessible via Mistral’s Le Platforme API (application programming interface) under the model name devstral-small-2505, with pricing set at $0.10 per million input tokens and $0.30 per million output tokens. For those deploying locally, support for frameworks like OpenHands enables integration with codebases and agentic workflows out of the box. Rozière shared how he incorporates Devstral in his own development flow: “I use it myself. You can ask it to do small tasks, like updating the version of a package or modifying a tokenization script. It finds the right place in your code and makes the changes. It’s really nice to use.” More to come While Devstral is currently released as a research preview, Mistral and All Hands AI are already working on a larger follow-up model with expanded capabilities. “There will always be a gap between smaller and larger models,” Rozière noted, “but we’ve gone a long way in bridging that. These models already perform very strongly, even compared to some larger competitors.” With its performance benchmarks, permissive license, and agentic design, Devstral positions itself not just as a code generation tool—but as a foundational model for building autonomous software engineering systems. Daily insights on business use cases with VB Daily If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI. Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

Banijay Asia and Nodwin Gaming, a division of Nazara Technologies, announced a strategic partnership to revolutionize esports, gaming and youth entertainment in India. This collaboration marks a significant step in blending cutting-edge content with the booming gaming and youth culture ecosystem. As part of this alliance, Banijay Asia will take the helm in producing Playground, Nodwin Gaming’s flagship gaming reality show that has redefined the intersection of gaming, entertainment, and influencer culture in the region. The partnership will also bring new scripted and captive reality shows to daylight, with both companies co-developing new intellectual properties (IPs) and pioneering fresh content concepts across gaming, esports and youth culture. Deepak Dhar, group CEO of Banijay Asia & Endemol Shine India, said in a statement, “This collaboration with Nodwin Gaming is an exciting extension of our vision to create content that resonates with evolving youth audiences. Gaming is not just a sport; it’s a culture, a lifestyle, and a massive content opportunity. With Playground and our upcoming IPs, we aim to redefine engagement and storytelling in this space.” He further added, “I look forward to the creative collaboration with Akshat Rathee, and Joost Roset who we previously worked with, in bringing iconic global IPs to India. This is an invaluable opportunity to create clutter-breaking formats in the youth and gaming space.” With a diverse portfolio of high-impact IPs including the televised esports tournament – BGMI Masters Series (BGMS), Comic Con India, NH7 Weekender, and more, Nodwin Gaming is a key player in the gaming, esports and youth entertainment ecosystem. Talking about the partnership, Akshat Rathee, managing director of Nodwin Gaming, said in a statement, “Youth entertainment is evolving rapidly, and our partnership with Banijay Asia positions us perfectly to lead that change. With Playground, we’ve just begun to explore what’s possible. Backed by Banijay’s global expertise, we’re now set to scale the IP across multiple languages and markets, with the ambition of building an international format.” Rathee added, “At the heart of our approach is our philosophy of ‘timeshare of mindshare’. We’re not just creating content, we’re creating cultural moments. The Rusk creative team will stay actively involved as we expand the universe of Playground, aiming to grow from one show to between five and ten each year. This is just the beginning of a bold new chapter in gaming and youth content.” With gaming emerging as one of the most influential youth-driven industries, this partnership is poised to set new benchmarks in how gaming and entertainment are consumed in India. By leveraging Banijay Asia’s expertise in content creation and Nodwin Gaming’s deep-rooted understanding of gaming communities, the collaboration promises to push creative boundaries and capture the imagination of young audiences nationwide. Banijay Entertainment has strong expertise in the gaming space, via its global distribution arm Banijay Rights, with successful gaming activations leveraging key brands including Peaky Blinders, Black Mirror, Simon’s Cat and MasterChef. GB Daily Stay in the know! Get the latest news in your inbox daily Read our Privacy Policy Thanks for subscribing. Check out more VB newsletters here. An error occured.

Astar Network, a gateway to take blockchain projects to Japan and more, announced a strategic investment from Animoca Brands.Read More