Industrious Switch fans have started compiling a fan-made list that tracks which Switch 1 games have seen an improvement - or a decline - in performance when running on their new Nintendo Switch 2 consoles. Read more

Outdated coding practices and memory-unsafe languages like C are putting software, including cryptographic libraries, at risk. Fortunately, memory-safe languages like Rust, along with formal verification tools, are now mature enough to be used at scale, helping prevent issues like crashes, data corruption, flawed implementation, and side-channel attacks. To address these vulnerabilities and improve memory safety, we’re rewriting SymCrypt (opens in new tab)—Microsoft’s open-source cryptographic library—in Rust. We’re also incorporating formal verification methods. SymCrypt is used in Windows, Azure Linux, Xbox, and other platforms. Currently, SymCrypt is primarily written in cross-platform C, with limited use of hardware-specific optimizations through intrinsics (compiler-provided low-level functions) and assembly language (direct processor instructions). It provides a wide range of algorithms, including AES-GCM, SHA, ECDSA, and the more recent post-quantum algorithms ML-KEM and ML-DSA.  Formal verification will confirm that implementations behave as intended and don’t deviate from algorithm specifications, critical for preventing attacks. We’ll also analyze compiled code to detect side-channel leaks caused by timing or hardware-level behavior. Proving Rust program properties with Aeneas Program verification is the process of proving that a piece of code will always satisfy a given property, no matter the input. Rust’s type system profoundly improves the prospects for program verification by providing strong ownership guarantees, by construction, using a discipline known as “aliasing xor mutability”. For example, reasoning about C code often requires proving that two non-const pointers are live and non-overlapping, a property that can depend on external client code. In contrast, Rust’s type system guarantees this property for any two mutably borrowed references. As a result, new tools have emerged specifically for verifying Rust code. We chose Aeneas (opens in new tab) because it helps provide a clean separation between code and proofs. Developed by Microsoft Azure Research in partnership with Inria, the French National Institute for Research in Digital Science and Technology, Aeneas connects to proof assistants like Lean (opens in new tab), allowing us to draw on a large body of mathematical proofs—especially valuable given the mathematical nature of cryptographic algorithms—and benefit from Lean’s active user community. Compiling Rust to C supports backward compatibility   We recognize that switching to Rust isn’t feasible for all use cases, so we’ll continue to support, extend, and certify C-based APIs as long as users need them. Users won’t see any changes, as Rust runs underneath the existing C APIs. Some users compile our C code directly and may rely on specific toolchains or compiler features that complicate the adoption of Rust code. To address this, we will use Eurydice (opens in new tab), a Rust-to-C compiler developed by Microsoft Azure Research, to replace handwritten C code with C generated from formally verified Rust. Eurydice (opens in new tab) compiles directly from Rust’s MIR intermediate language, and the resulting C code will be checked into the SymCrypt repository alongside the original Rust source code. As more users adopt Rust, we’ll continue supporting this compilation path for those who build SymCrypt from source code but aren’t ready to use the Rust compiler. In the long term, we hope to transition users to either use precompiled SymCrypt binaries (via C or Rust APIs), or compile from source code in Rust, at which point the Rust-to-C compilation path will no longer be needed. Microsoft research podcast Ideas: AI and democracy with Madeleine Daepp and Robert Osazuwa Ness As the “biggest election year in history” comes to an end, researchers Madeleine Daepp and Robert Osazuwa Ness and Democracy Forward GM Ginny Badanes discuss AI’s impact on democracy, including the tech’s use in Taiwan and India. Listen now Opens in a new tab Timing analysis with Revizor  Even software that has been verified for functional correctness can remain vulnerable to low-level security threats, such as side channels caused by timing leaks or speculative execution. These threats operate at the hardware level and can leak private information, such as memory load addresses, branch targets, or division operands, even when the source code is provably correct.  To address this, we’re extending Revizor (opens in new tab), a tool developed by Microsoft Azure Research, to more effectively analyze SymCrypt binaries. Revizor models microarchitectural leakage and uses fuzzing techniques to systematically uncover instructions that may expose private information through known hardware-level effects.   Earlier cryptographic libraries relied on constant-time programming to avoid operations on secret data. However, recent research has shown that this alone is insufficient with today’s CPUs, where every new optimization may open a new side channel.  By analyzing binary code for specific compilers and platforms, our extended Revizor tool enables deeper scrutiny of vulnerabilities that aren’t visible in the source code. Verified Rust implementations begin with ML-KEM This long-term effort is in alignment with the Microsoft Secure Future Initiative and brings together experts across Microsoft, building on decades of Microsoft Research investment in program verification and security tooling. A preliminary version of ML-KEM in Rust is now available on the preview feature/verifiedcrypto (opens in new tab) branch of the SymCrypt repository. We encourage users to try the Rust build and share feedback (opens in new tab). Looking ahead, we plan to support direct use of the same cryptographic library in Rust without requiring C bindings.  Over the coming months, we plan to rewrite, verify, and ship several algorithms in Rust as part of SymCrypt. As our investment in Rust deepens, we expect to gain new insights into how to best leverage the language for high-assurance cryptographic implementations with low-level optimizations.  As performance is key to scalability and sustainability, we’re holding new implementations to a high bar using our benchmarking tools to match or exceed existing systems. Looking forward  This is a pivotal moment for high-assurance software. Microsoft’s investment in Rust and formal verification presents a rare opportunity to advance one of our key libraries. We’re excited to scale this work and ultimately deliver an industrial-grade, Rust-based, FIPS-certified cryptographic library. Opens in a new tab

06/12 — 2025 by abduzeedo Explore "100 Beste Plakate 24," a stunning yearbook by Tristesse and Slanted Publishers. Dive into cutting-edge editorial design and visual identity. Design enthusiasts, get ready to dive into the latest from the German-speaking design scene. The "100 Beste Plakate 24" yearbook offers a compelling showcase of contemporary graphic design. It's more than just a collection; it's a deep exploration of visual identity and editorial design. This yearbook, published by Slanted Publishers and edited by 100 beste Plakate e. V. and Fons Hickmann, is a testament to the power of impactful poster design. The design studio Tristesse from Basel took the reins for the overall concept, delivering a fresh and cheeky aesthetic that makes the "100 best posters" feel like leading actors on a vibrant stage. Their in-house approach to layout, typography, and photography truly shines. Unpacking the Visuals The book's format (17×24 cm) and 256 pages allow for large-format images, providing ample space to appreciate each poster's intricate details. It includes detailed credits, content descriptions, and creation contexts. This commitment to detail in the editorial design elevates the reading experience. One notable example within the yearbook is the "To-Do: Diplome 24" poster campaign by Atelier HKB. Designed under Marco Matti's project management, this series features twelve motifs for the Bern University of the Arts graduation events. These posters highlight effective graphic design and visual communication. Another standout is the "Rettungsplakate" by klotz-studio für gestaltung. These "rescue posters," printed on actual rescue blankets, address homelessness in Germany. The raw, impactful visual approach paired with a tangible medium demonstrates powerful design with a purpose. Beyond the Imagery Beyond the stunning visuals, the yearbook offers insightful essays and interviews on current poster design trends. The introductory section features jury members, their works, and statements on the selection process, alongside forewords from the association president and jury chair. This editorial content offers valuable context and insights into the evolving landscape of graphic design. The book’s concept playfully questions the seriousness and benevolence of the honorary certificates awarded to the winning designers. This subtle irony adds a unique layer to the publication, transforming it from a mere compilation into a thoughtful commentary on the design world itself. It's an inspiring showcase of the cutting edge of contemporary graphic design. The Art of Editorial Design "100 Beste Plakate 24" is a prime example of exceptional editorial design. It's not just about compiling images; it's about curating a narrative. The precise layout, thoughtful typography choices, and the deliberate flow of content all contribute to a cohesive and engaging experience. This book highlights how editorial design can transform a collection of works into a compelling story, inviting readers to delve deeper into each piece. The attention to detail, from the softcover with flaps to the thread-stitching and hot-foil embossing, speaks volumes about the dedication to craftsmanship. This is where illustration, graphic design, and branding converge to create a truly immersive experience. Final Thoughts This yearbook is a must-have for anyone passionate about graphic design and visual identity. It offers a fresh perspective on contemporary poster design, highlighting both aesthetic excellence and social relevance. The detailed insights into the design process and the designers' intentions make it an invaluable resource. Pick up a copy and see how impactful design can be. You can learn more about this incredible work and acquire your copy at slanted.de/product/100-beste-plakate-24. Editorial design artifacts Tags editorial design

ChatGPT is rapidly changing the world. The process is already happening, and it’s only going to accelerate as the technology improves, as more people gain access to it, and as more learn how to use it. What’s shocking is just how many tasks ChatGPT is already capable of managing for you. While the naysayers may still look down their noses at the potential of AI assistants, I’ve been using it to handle all kinds of menial tasks for me. Here are my favorite examples. Further reading: This tiny ChatGPT feature helps me tackle my days more productively Write your emails for you Dave Parrack / Foundry We’ve all been faced with the tricky task of writing an email—whether personal or professional—but not knowing quite how to word it. ChatGPT can do the heavy lifting for you, penning the (hopefully) perfect email based on whatever information you feed it. Let’s assume the email you need to write is of a professional nature, and wording it poorly could negatively affect your career. By directing ChatGPT to write the email with a particular structure, content, and tone of voice, you can give yourself a huge head start. A winning tip for this is to never accept ChatGPT’s first attempt. Always read through it and look for areas of improvement, then request tweaks to ensure you get the best possible email. You can (and should) also rewrite the email in your own voice. Learn more about how ChatGPT coached my colleague to write better emails. Generate itineraries and schedules Dave Parrack / Foundry If you’re going on a trip but you’re the type of person who hates planning trips, then you should utilize ChatGPT’s ability to generate trip itineraries. The results can be customized to the nth degree depending on how much detail and instruction you’re willing to provide. As someone who likes to get away at least once a year but also wants to make the most of every trip, leaning on ChatGPT for an itinerary is essential for me. I’ll provide the location and the kinds of things I want to see and do, then let it handle the rest. Instead of spending days researching everything myself, ChatGPT does 80 percent of it for me. As with all of these tasks, you don’t need to accept ChatGPT’s first effort. Use different prompts to force the AI chatbot to shape the itinerary closer to what you want. You’d be surprised at how many cool ideas you’ll encounter this way—simply nix the ones you don’t like. Break down difficult concepts Dave Parrack / Foundry One of the best tasks to assign to ChatGPT is the explanation of difficult concepts. Ask ChatGPT to explain any concept you can think of and it will deliver more often than not. You can tailor the level of explanation you need, and even have it include visual elements. Let’s say, for example, that a higher-up at work regularly lectures everyone about the importance of networking. But maybe they never go into detail about what they mean, just constantly pushing the why without explaining the what. Well, just ask ChatGPT to explain networking! Okay, most of us know what “networking” is and the concept isn’t very hard to grasp. But you can do this with anything. Ask ChatGPT to explain augmented reality, multi-threaded processing, blockchain, large language models, what have you. It will provide you with a clear and simple breakdown, maybe even with analogies and images. Analyze and make tough decisions Dave Parrack / Foundry We all face tough decisions every so often. The next time you find yourself wrestling with a particularly tough one—and you just can’t decide one way or the other—try asking ChatGPT for guidance and advice. It may sound strange to trust any kind of decision to artificial intelligence, let alone an important one that has you stumped, but doing so actually makes a lot of sense. While human judgment can be clouded by emotions, AI can set that aside and prioritize logic. It should go without saying: you don’t have to accept ChatGPT’s answers. Use the AI to weigh the pros and cons, to help you understand what’s most important to you, and to suggest a direction. Who knows? If you find yourself not liking the answer given, that in itself might clarify what you actually want—and the right answer for you. This is the kind of stuff ChatGPT can do to improve your life. Plan complex projects and strategies Dave Parrack / Foundry Most jobs come with some level of project planning and management. Even I, as a freelance writer, need to plan tasks to get projects completed on time. And that’s where ChatGPT can prove invaluable, breaking projects up into smaller, more manageable parts. ChatGPT needs to know the nature of the project, the end goal, any constraints you may have, and what you have done so far. With that information, it can then break the project up with a step-by-step plan, and break it down further into phases (if required). If ChatGPT doesn’t initially split your project up in a way that suits you, try again. Change up the prompts and make the AI chatbot tune in to exactly what you’re looking for. It takes a bit of back and forth, but it can shorten your planning time from hours to mere minutes. Compile research notes Dave Parrack / Foundry If you need to research a given topic of interest, ChatGPT can save you the hassle of compiling that research. For example, ahead of a trip to Croatia, I wanted to know more about the Croatian War of Independence, so I asked ChatGPT to provide me with a brief summary of the conflict with bullet points to help me understand how it happened. After absorbing all that information, I asked ChatGPT to add a timeline of the major events, further helping me to understand how the conflict played out. ChatGPT then offered to provide me with battle maps and/or summaries, plus profiles of the main players. You can go even deeper with ChatGPT’s Deep Research feature, which is now available to free users, up to 5 Deep Research tasks per month. With Deep Research, ChatGPT conducts multi-step research to generate comprehensive reports (with citations!) based on large amounts of information across the internet. A Deep Research task can take up to 30 minutes to complete, but it’ll save you hours or even days. Summarize articles, meetings, and more Dave Parrack / Foundry There are only so many hours in the day, yet so many new articles published on the web day in and day out. When you come across extra-long reads, it can be helpful to run them through ChatGPT for a quick summary. Then, if the summary is lacking in any way, you can go back and plow through the article proper. As an example, I ran one of my own PCWorld articles (where I compared Bluesky and Threads as alternatives to X) through ChatGPT, which provided a brief summary of my points and broke down the best X alternative based on my reasons given. Interestingly, it also pulled elements from other articles. (Hmph.) If you don’t want that, you can tell ChatGPT to limit its summary to the contents of the link. This is a great trick to use for other long-form, text-heavy content that you just don’t have the time to crunch through. Think transcripts for interviews, lectures, videos, and Zoom meetings. The only caveat is to never share private details with ChatGPT, like company-specific data that’s protected by NDAs and the like. Create Q&A flashcards for learning Dave Parrack / Foundry Flashcards can be extremely useful for drilling a lot of information into your brain, such as when studying for an exam, onboarding in a new role, prepping for an interview, etc. And with ChatGPT, you no longer have to painstakingly create those flashcards yourself. All you have to do is tell the AI the details of what you’re studying. You can specify the format (such as Q&A or multiple choice), as well as various other elements. You can also choose to keep things broad or target specific sub-topics or concepts you want to focus on. You can even upload your own notes for ChatGPT to reference. You can also use Google’s NotebookLM app in a similar way. Provide interview practice Dave Parrack / Foundry Whether you’re a first-time jobseeker or have plenty of experience under your belt, it’s always a good idea to practice for your interviews when making career moves. Years ago, you might’ve had to ask a friend or family member to act as your mock interviewer. These days, ChatGPT can do it for you—and do it more effectively. Inform ChatGPT of the job title, industry, and level of position you’re interviewing for, what kind of interview it’ll be (e.g., screener, technical assessment, group/panel, one-on-one with CEO), and anything else you want it to take into consideration. ChatGPT will then conduct a mock interview with you, providing feedback along the way. When I tried this out myself, I was shocked by how capable ChatGPT can be at pretending to be a human in this context. And the feedback it provides for each answer you give is invaluable for knocking off your rough edges and improving your chances of success when you’re interviewed by a real hiring manager. Further reading: Non-gimmicky AI apps I actually use every day

Published June 5, 2025 10:00am EDT close Don’t be so quick to click that Google calendar invite. It could be a hacker’s trap Cybercriminals are sending fake meeting invitations that seem legitimate. NEWYou can now listen to Fox News articles! Americans’ personal data is now spread across more digital platforms than ever. From online shopping habits to fitness tracking logs, personal information ends up in hundreds of company databases. While most people worry about social media leaks or email hacks, a far less visible threat comes from data brokers.I still find it hard to believe that companies like this are allowed to operate with so little legal scrutiny. These firms trade in personal information without our knowledge or consent. What baffles me even more is that they aren’t serious about protecting the one thing that is central to their business model: data. Just last year, we saw news of a massive data breach at a data broker called National Public Data, which exposed 2.7 billion records. And now another data broker, LexisNexis, a major name in the industry, has reported a significant breach that exposed sensitive information from more than 364,000 people. A hacker at work (Kurt "CyberGuy" Knutsson)LexisNexis breach went undetected for months after holiday hackLexisNexis filed a notice with the Maine attorney general revealing that a hacker accessed consumer data through a third-party software development platform. The breach happened on Dec. 25, 2024, but the company only discovered it months later. LexisNexis was alerted on April 1, 2025, by an unnamed individual who claimed to have found sensitive files. It remains unclear whether this person was responsible for the breach or merely came across the exposed data.MASSIVE DATA BREACH EXPOSES 184 MILLION PASSWORDS AND LOGINSA spokesperson for LexisNexis confirmed that the hacker gained access to the company’s GitHub account. This is a platform commonly used by developers to store and collaborate on code. Security guidelines repeatedly warn against storing sensitive information in such repositories; however, mistakes such as exposed access tokens and personal data files continue to occur.The stolen data varies from person to person but includes full names, birthdates, phone numbers, mailing and email addresses, Social Security numbers and driver's license numbers. LexisNexis has not confirmed whether it received any ransom demand or had further contact with the attacker. An individual working on their laptop (Kurt "CyberGuy" Knutsson)Why the LexisNexis hack is a bigger threat than you realizeLexisNexis isn’t a household name for most people, but it plays a major role in how personal data is harvested and used behind the scenes. The company pulls information from a wide range of sources, compiling detailed profiles that help other businesses assess risk and detect fraud. Its clients include banks, insurance companies and government agencies.In 2023, the New York Times reported that several car manufacturers had been sharing driving data with LexisNexis without notifying vehicle owners. That information was then sold to insurance companies, which used it to adjust premiums based on individual driving behavior. The story made one thing clear. LexisNexis has access to a staggering amount of personal detail, even from people who have never willingly engaged with the company.Law enforcement also uses LexisNexis tools to dig up information on suspects. These systems offer access to phone records, home addresses and other historical data. While such tools might assist in investigations, they also highlight a serious issue. When this much sensitive information is concentrated in one place, it becomes a single point of failure. And as the recent breach shows, that failure is no longer hypothetical. A hacker at work (Kurt "CyberGuy" Knutsson)7 expert tips to protect your personal data after a data broker breachKeeping your personal data safe online can feel overwhelming, but a few practical steps can make a big difference in protecting your privacy and reducing your digital footprint. Here are 7 effective ways to take control of your information and keep it out of the wrong hands:1. Remove your data from the internet: The most effective way to take control of your data and avoid data brokers from selling it is to opt for data removal services. While no service promises to remove all your data from the internet, having a removal service is great if you want to constantly monitor and automate the process of removing your information from hundreds of sites continuously over a longer period of time. Check out my top picks for data removal services here.Get a free scan to find out if your personal information is already out on the web.2. Review privacy settings: Take a few minutes to explore the privacy and security settings on the services you use. For example, limit who can see your social media posts, disable unnecessary location-sharing on your phone and consider turning off ad personalization on accounts like Google and Facebook. Most browsers let you block third-party cookies or clear tracking data. The FTC suggests comparing the privacy notices of different sites and apps and choosing ones that let you opt out of sharing when possible.3. Use privacy-friendly tools: Install browser extensions or plugins that block ads and trackers (such as uBlock Origin or Privacy Badger). You might switch to a more private search engine (like DuckDuckGo or Brave) that doesn’t log your queries. Consider using a browser’s "incognito" or private mode when you don’t want your history saved, and regularly clear your cookies and cache. Even small habits, like logging out of accounts when not in use or using a password manager, make you less trackable.GET FOX BUSINESS ON THE GO BY CLICKING HERE4. Beware of phishing links and use strong antivirus software: Scammers may try to get access to your financial details and other important data using phishing links. The best way to safeguard yourself from malicious links is to have antivirus software installed on all your devices. This protection can also alert you to phishing emails and ransomware scams, keeping your personal information and digital assets safe. Get my picks for the best 2025 antivirus protection winners for your Windows, Mac, Android and iOS devices.5. Be cautious with personal data: Think twice before sharing extra details. Don’t fill out online surveys or quizzes that ask for personal or financial information unless you trust the source. Create separate email addresses for sign-ups (so marketing emails don’t go to your main inbox). Only download apps from official stores and check app permissions.6. Opt out of data broker lists: Many data brokers offer ways to opt out or delete your information, though it can be a tedious process. For example, there are sites like Privacy Rights Clearinghouse or the Whitepages opt-out page that list popular brokers and their opt-out procedures. The FTC’s consumer guide, "Your Guide to Protecting Your Privacy Online," includes tips on opting out of targeted ads and removing yourself from people-search databases. Keep in mind you may have to repeat this every few months.7. Be wary of mailbox communications: Bad actors may also try to scam you through snail mail. The data leak gives them access to your address. They may impersonate people or brands you know and use themes that require urgent attention, such as missed deliveries, account suspensions and security alerts.Kurt’s key takeawayFor many, the LexisNexis breach may be the first time they realize just how much of their data is in circulation. Unlike a social media platform or a bank, there is no clear customer relationship with a data broker, and that makes it harder to demand transparency. This incident should prompt serious discussion around what kind of oversight is necessary in industries that operate in the shadows. A more informed public and stronger regulation may be the only things standing between personal data and permanent exposure.CLICK HERE TO GET THE FOX NEWS APPShould companies be allowed to sell your personal information without your consent? Let us know by writing us atCyberguy.com/Contact.For more of my tech tips and security alerts, subscribe to my free CyberGuy Report Newsletter by heading to Cyberguy.com/Newsletter.Ask Kurt a question or let us know what stories you'd like us to cover.Follow Kurt on his social channels:Answers to the most-asked CyberGuy questions:New from Kurt:Copyright 2025 CyberGuy.com. All rights reserved. Kurt "CyberGuy" Knutsson is an award-winning tech journalist who has a deep love of technology, gear and gadgets that make life better with his contributions for Fox News & FOX Business beginning mornings on "FOX & Friends." Got a tech question? Get Kurt’s free CyberGuy Newsletter, share your voice, a story idea or comment at CyberGuy.com.

Last year, China saw a boom in foundation models, the do-everything large language models that underpin the AI revolution. This year, the focus has shifted to AI agents—systems that are less about responding to users’ queries and more about autonomously accomplishing things for them.  There are now a host of Chinese startups building these general-purpose digital tools, which can answer emails, browse the internet to plan vacations, and even design an interactive website. Many of these have emerged in just the last two months, following in the footsteps of Manus—a general AI agent that sparked weeks of social media frenzy for invite codes after its limited-release launch in early March.  These emerging AI agents aren’t large language models themselves. Instead, they’re built on top of them, using a workflow-based structure designed to get things done. A lot of these systems also introduce a different way of interacting with AI. Rather than just chatting back and forth with users, they are optimized for managing and executing multistep tasks—booking flights, managing schedules, conducting research—by using external tools and remembering instructions.  China could take the lead on building these kinds of agents. The country’s tightly integrated app ecosystems, rapid product cycles, and digitally fluent user base could provide a favorable environment for embedding AI into daily life.  For now, its leading AI agent startups are focusing their attention on the global market, because the best Western models don’t operate inside China’s firewalls. But that could change soon: Tech giants like ByteDance and Tencent are preparing their own AI agents that could bake automation directly into their native super-apps, pulling data from their vast ecosystem of programs that dominate many aspects of daily life in the country.  As the race to define what a useful AI agent looks like unfolds, a mix of ambitious startups and entrenched tech giants are now testing how these tools might actually work in practice—and for whom. Set the standard It’s been a whirlwind few months for Manus, which was developed by the Wuhan-based startup Butterfly Effect. The company raised $75 million in a funding round led by the US venture capital firm Benchmark, took the product on an ambitious global roadshow, and hired dozens of new employees.  Even before registration opened to the public in May, Manus had become a reference point for what a broad, consumer‑oriented AI agent should accomplish. Rather than handling narrow chores for businesses, this “general” agent is designed to be able to help with everyday tasks like trip planning, stock comparison, or your kid’s school project.  Unlike previous AI agents, Manus uses a browser-based sandbox that lets users supervise the agent like an intern, watching in real time as it scrolls through web pages, reads articles, or codes actions. It also proactively asks clarifying questions, supports long-term memory that would serve as context for future tasks. “Manus represents a promising product experience for AI agents,” says Ang Li, cofounder and CEO of Simular, a startup based in Palo Alto, California, that’s building computer use agents, AI agents that control a virtual computer. “I believe Chinese startups have a huge advantage when it comes to designing consumer products, thanks to cutthroat domestic competition that leads to fast execution and greater attention to product details.” In the case of Manus, the competition is moving fast. Two of the most buzzy follow‑ups, Genspark and Flowith, for example, are already boasting benchmark scores that match or edge past Manus’s.  Genspark, led by former Baidu executives Eric Jing and Kay Zhu, links many small “super agents” through what it calls multi‑component prompting. The agent can switch among several large language models, accepts both images and text, and carries out tasks from making slide decks to placing phone calls. Whereas Manus relies heavily on Browser Use, a popular open-source product that lets agents operate a web browser in a virtual window like a human, Genspark directly integrates with a wide array of tools and APIs. Launched in April, the company says that it already has over 5 million users and over $36 million in yearly revenue. Flowith, the work of a young team that first grabbed public attention in April 2025 at a developer event hosted by the popular social media app Xiaohongshu, takes a different tack. Marketed as an “infinite agent,” it opens on a blank canvas where each question becomes a node on a branching map. Users can backtrack, take new branches, and store results in personal or sharable “knowledge gardens”—a design that feels more like project management software (think Notion) than a typical chat interface. Every inquiry or task builds its own mind-map-like graph, encouraging a more nonlinear and creative interaction with AI. Flowith’s core agent, NEO, runs in the cloud and can perform scheduled tasks like sending emails and compiling files. The founders want the app to be a “knowledge marketbase”, and aims to tap into the social aspect of AI with the aspiration of becoming “the OnlyFans of AI knowledge creators”. What they also share with Manus is the global ambition. Both Genspark and Flowith have stated that their primary focus is the international market. A global address Startups like Manus, Genspark, and Flowith—though founded by Chinese entrepreneurs—could blend seamlessly into the global tech scene and compete effectively abroad. Founders, investors, and analysts that MIT Technology Review has spoken to believe Chinese companies are moving fast, executing well, and quickly coming up with new products.  Money reinforces the pull to launch overseas. Customers there pay more, and there are plenty to go around. “You can price in USD, and with the exchange rate that’s a sevenfold multiplier,” Manus cofounder Xiao Hong quipped on a podcast. “Even if we’re only operating at 10% power because of cultural differences overseas, we’ll still make more than in China.” But creating the same functionality in China is a challenge. Major US AI companies including OpenAI and Anthropic have opted out of mainland China because of geopolitical risks and challenges with regulatory compliance. Their absence initially created a black market as users resorted to VPNs and third-party mirrors to access tools like ChatGPT and Claude. That vacuum has since been filled by a new wave of Chinese chatbots—DeepSeek, Doubao, Kimi—but the appetite for foreign models hasn’t gone away.  Manus, for example, uses Anthropic’s Claude Sonnet—widely considered the top model for agentic tasks. Manus cofounder Zhang Tao has repeatedly praised Claude’s ability to juggle tools, remember contexts, and hold multi‑round conversations—all crucial for turning chatty software into an effective executive assistant. But the company’s use of Sonnet has made its agent functionally unusable inside China without a VPN. If you open Manus from a mainland IP address, you’ll see a notice explaining that the team is “working on integrating Qwen’s model,” a special local version that is built on top of Alibaba’s open-source model.  An engineer overseeing ByteDance’s work on developing an agent, who spoke to MIT Technology Review anonymously to avoid sanction, said that the absence of Claude Sonnet models “limits everything we do in China.” DeepSeek’s open models, he added, still hallucinate too often and lack training on real‑world workflows. Developers we spoke with rank Alibaba’s Qwen series as the best domestic alternative, yet most say that switching to Qwen knocks performance down a notch. Jiaxin Pei, a postdoctoral researcher at Stanford’s Institute for Human‑Centered AI, thinks that gap will close: “Building agentic capabilities in base LLMs has become a key focus for many LLM builders, and once people realize the value of this, it will only be a matter of time.” For now, Manus is doubling down on audiences it can already serve. In a written response, the company said its “primary focus is overseas expansion,” noting that new offices in San Francisco, Singapore, and Tokyo have opened in the past month. A super‑app approach Although the concept of AI agents is still relatively new, the consumer-facing AI app market in China is already crowded with major tech players. DeepSeek remains the most widely used, while ByteDance’s Doubao and Moonshot’s Kimi have also become household names. However, most of these apps are still optimized for chat and entertainment rather than task execution. This gap in the local market has pushed China’s big tech firms to roll out their own user-facing agents, though early versions remain uneven in quality and rough around the edges.  ByteDance is testing Coze Space, an AI agent based on its own Doubao model family that lets users toggle between “plan” and “execute” modes, so they can either directly guide the agent’s actions or step back and watch it work autonomously. It connects up to 14 popular apps, including GitHub, Notion, and the company’s own Lark office suite. Early reviews say the tool can feel clunky and has a high failure rate, but it clearly aims to match what Manus offers. Meanwhile, Zhipu AI has released a free agent called AutoGLM Rumination, built on its proprietary ChatGLM models. Shanghai‑based Minimax has launched Minimax Agent. Both products look almost identical to Manus and demo basic tasks such as building a simple website, planning a trip, making a small Flash game, or running quick data analysis. Despite the limited usability of most general AI agents launched within China, big companies have plans to change that. During a May 15 earnings call, Tencent president Liu Zhiping teased an agent that would weave automation directly into China’s most ubiquitous app, WeChat.  Considered the original super-app, WeChat already handles messaging, mobile payments, news, and millions of mini‑programs that act like embedded apps. These programs give Tencent, its developer, access to data from millions of services that pervade everyday life in China, an advantage most competitors can only envy. Historically, China’s consumer internet has splintered into competing walled gardens—share a Taobao link in WeChat and it resolves as plaintext, not a preview card. Unlike the more interoperable Western internet, China’s tech giants have long resisted integration with one another, choosing to wage platform war at the expense of a seamless user experience. But the use of mini‑programs has given WeChat unprecedented reach across services that once resisted interoperability, from gym bookings to grocery orders. An agent able to roam that ecosystem could bypass the integration headaches dogging independent startups. Alibaba, the e-commerce giant behind the Qwen model series, has been a front-runner in China’s AI race but has been slower to release consumer-facing products. Even though Qwen was the most downloaded open-source model on Hugging Face in 2024, it didn’t power a dedicated chatbot app until early 2025. In March, Alibaba rebranded its cloud storage and search app Quark into an all-in-one AI search tool. By June, Quark had introduced DeepResearch—a new mode that marks its most agent-like effort to date.  ByteDance and Alibaba did not reply to MIT Technology Review’s request for comments. “Historically, Chinese tech products tend to pursue the all-in-one, super-app approach, and the latest Chinese AI agents reflect just that,” says Li of Simular, who previously worked at Google DeepMind on AI-enabled work automation. “In contrast, AI agents in the US are more focused on serving specific verticals.” Pei, the researcher at Stanford, says that existing tech giants could have a huge advantage in bringing the vision of general AI agents to life—especially those with built-in integration across services. “The customer-facing AI agent market is still very early, with tons of problems like authentication and liability,” he says. “But companies that already operate across a wide range of services have a natural advantage in deploying agents at scale.”

Replicube A game by Walaber Entertainment LLCDev snapshot: Godot 4.5 dev 5By: Thaddeus Crews2 June 2025Pre-releaseBrrr… Do you feel that? That’s the cold front of the feature freeze just around the corner. It’s not upon us just yet, but this is likely to be our final development snapshot of the 4.5 release cycle. As we enter the home stretch of new features, bugs are naturally going to follow suit, meaning bug reports and feedback will be especially important for a smooth beta timeframe.Jump to the Downloads section, and give it a spin right now, or continue reading to learn more about improvements in this release. You can also try the Web editor or the Android editor for this release. If you are interested in the latter, please request to join our testing group to get access to pre-release builds.The cover illustration is from Replicube, a programming puzzle game where you write code to recreate voxelized objects. It is developed by Walaber Entertainment LLC (Bluesky, Twitter). You can get the game on Steam.HighlightsIn case you missed them, see the 4.5 dev 1, 4.5 dev 2, 4.5 dev 3, and 4.5 dev 4 release notes for an overview of some key features which were already in those snapshots, and are therefore still available for testing in dev 5.Native visionOS supportNormally, our featured highlights in these development blogs come from long-time contributors. This makes sense of course, as it’s generally those users that have the familiarity necessary for major changes or additions that are commonly used for these highlights. That’s why it might surprise you to hear that visionOS support comes to us from Ricardo Sanchez-Saez, whose pull request GH-105628 is his very first contribution to the engine! It might not surprise you to hear that Ricardo is part of the visionOS engineering team at Apple, which certainly helps get his foot in the door, but that still makes visionOS the first officially-supported platform integration in about a decade.For those unfamiliar, visionOS is Apple’s XR environment. We’re no strangers to XR as a concept (see our recent XR blogpost highlighting the latest Godot XR Game Jam), but XR platforms are as distinct from one another as traditional platforms. visionOS users have expressed a strong interest in integrating with our ever-growing XR community, and now we can make that happen. See you all in the next XR Game Jam!GDScript: Abstract classesWhile the Godot Engine utilizes abstract classes—a class that cannot be directly instantiated—frequently, this was only ever supported internally. Thanks to the efforts of Aaron Franke, this paradigm is now available to GDScript users (GH-67777). Now if a user wants to introduce their own abstract class, they merely need to declare it via the new abstract keyword:abstract class_name MyAbstract extends Node The purpose of an abstract class is to create a baseline for other classes to derive from:class_name ExtendsMyAbstract extends MyAbstract Shader bakerFrom the technical gurus behind implementing ubershaders, Darío Samo and Pedro J. Estébanez bring us another miracle of rendering via GH-102552: shader baker exporting. This is an optional feature that can be enabled at export time to speed up shader compilation massively. This feature works with ubershaders automatically without any work from the user. Using shader baking is strongly recommended when targeting Apple devices or D3D12 since it makes the biggest difference there (over 20× decrease in load times in the TPS demo)!Before:After:However, it comes with tradeoffs:Export time will be much longer.Build size will be much larger since the baked shaders can take up a lot of space.We have removed several MoltenVK bug workarounds from the Forward+ shader, therefore we no longer guarantee support for the Forward+ renderer on Intel Macs. If you are targeting Intel Macs, you should use the Mobile or Compatibility renderers.Baking for Vulkan can be done from any device, but baking for D3D12 needs to be done from a Windows device and baking for Apple .metallib requires a Metal compiler (macOS with Xcode / Command Line Tools installed).Web: WebAssembly SIMD supportAs you might recall, Godot 4.0 initially released under the assumption that multi-threaded web support would become the standard, and only supported that format for web builds. This assumption unfortunately proved to be wishful thinking, and was reverted in 4.3 by allowing for single-threaded builds once more. However, this doesn’t mean that these single-threaded environments are inherently incapable of parallel processing; it just requires alternative implementations. One such implementation, SIMD, is a perfect candidate thanks to its support across all major browsers. To that end, web-wiz Adam Scott has taken to integrating this implementation for our web builds by default (GH-106319).Inline color pickersWhile it’s always been possible to see what kind of variable is assigned to an exported color in the inspector, some users have expressed a keen interest in allowing for this functionality within the script editor itself. This is because it would mean seeing what kind of color is represented by a variable without it needing to be exposed, as well as making it more intuitive at a glance as to what color a name or code corresponds to. Koliur Rahman has blessed us with this quality-of-life goodness, which adds an inline color picker GH-105724. Now no matter where the color is declared, users will be able to immediately and intuitively know what is actually represented in a non-intrusive manner.Rendering goodiesThe renderer got a fair amount of love this snapshot; not from any one PR, but rather a multitude of community members bringing some long-awaited features to light. Raymond DiDonato helped SMAA 1x make its transition from addon to fully-fledged engine feature (GH-102330). Capry brings bent normal maps to further enhance specular occlusion and indirect lighting (GH-89988). Our very own Clay John converted our Compatibility backend to use a fragment shader copy instead of a blit copy, working around common sample rate issues on mobile devices (GH-106267). More technical information on these rendering changes can be found in their associated PRs.SMAA comparison:OffOnBent normal map comparison:BeforeAfterAnd more!There are too many exciting changes to list them all here, but here’s a curated selection:Animation: Add alphabetical sorting to Animation Player (GH-103584).Animation: Add animation filtering to animation editor (GH-103130).Audio: Implement seek operation for Theora video files, improve multi-channel audio resampling (GH-102360).Core: Add --scene command line argument (GH-105302).Core: Overhaul resource duplication (GH-100673).Core: Use Grisu2 algorithm in String::num_scientific to fix serializing (GH-98750).Editor: Add “Quick Load” button to EditorResourcePicker (GH-104490).Editor: Add PROPERTY_HINT_INPUT_NAME for use with @export_custom to allow using input actions (GH-96611).Editor: Add named EditorScripts to the command palette (GH-99318).GUI: Add file sort to FileDialog (GH-105723).I18n: Add translation preview in editor (GH-96921).Import: Add Channel Remap settings to ResourceImporterTexture (GH-99676).Physics: Improve performance with non-monitoring areas when using Jolt Physics (GH-106490).Porting: Android: Add export option for custom theme attributes (GH-106724).Porting: Android: Add support for 16 KB page sizes, update to NDK r28b (GH-106358).Porting: Android: Remove the gradle_build/compress_native_libraries export option (GH-106359).Porting: Web: Use actual PThread pool size for get_default_thread_pool_size() (GH-104458).Porting: Windows/macOS/Linux: Use SSE 4.2 as a baseline when compiling Godot (GH-59595).Rendering: Add new StandardMaterial properties to allow users to control FPS-style objects (hands, weapons, tools close to the camera) (GH-93142).Rendering: FTI - Optimize SceneTree traversal (GH-106244).Changelog109 contributors submitted 252 fixes for this release. See our interactive changelog for the complete list of changes since the previous 4.5-dev4 snapshot.This release is built from commit 64b09905c.DownloadsGodot is downloading...Godot exists thanks to donations from people like you. Help us continue our work:Make a DonationStandard build includes support for GDScript and GDExtension..NET build (marked as mono) includes support for C#, as well as GDScript and GDExtension.While engine maintainers try their best to ensure that each preview snapshot and release candidate is stable, this is by definition a pre-release piece of software. Be sure to make frequent backups, or use a version control system such as Git, to preserve your projects in case of corruption or data loss.Known issuesWindows executables (both the editor and export templates) have been signed with an expired certificate. You may see warnings from Windows Defender’s SmartScreen when running this version, or outright be prevented from running the executables with a double-click (GH-106373). Running Godot from the command line can circumvent this. We will soon have a renewed certificate which will be used for future builds.With every release, we accept that there are going to be various issues, which have already been reported but haven’t been fixed yet. See the GitHub issue tracker for a complete list of known bugs.Bug reportsAs a tester, we encourage you to open bug reports if you experience issues with this release. Please check the existing issues on GitHub first, using the search function with relevant keywords, to ensure that the bug you experience is not already known.In particular, any change that would cause a regression in your projects is very important to report (e.g. if something that worked fine in previous 4.x releases, but no longer works in this snapshot).SupportGodot is a non-profit, open source game engine developed by hundreds of contributors on their free time, as well as a handful of part and full-time developers hired thanks to generous donations from the Godot community. A big thank you to everyone who has contributed their time or their financial support to the project!If you’d like to support the project financially and help us secure our future hires, you can do so using the Godot Development Fund.Donate now

While it doesn't have quite the same prominence as ChatGPT or Google Gemini, the Claude AI bot developed by Anthropic continues to improve and innovate. Brand new Claude 4 models are now available, promising upgrades in coding, reasoning, precision, and the ability to manage long-running tasks independently.There are two new models, Claude Opus 4 and Claude Sonnet 4, and Anthropic says they're both "setting new standards" for what you can expect from AI. Coding is a big focus, and the models are said to have achieved the highest scores to date on two widely used AI coding benchmarking tools, SWE-bench and Terminal-bench. Claude 4 models can actually work for hours on projects without any user input, Anthropic says. The updated models are better at handling more steps across more complex tasks, debugging their own work, and solving tricky problems along the way. They should also follow user instructions more exactly, and create end results that look better and work more reliably. Anthropic quotes partners such as GitHub, Cursor, and Rakuten in explaining how much of a step forward these models are.Away from code generation and analysis, the models also bring with them extended thinking, the ability to work on multiple tasks in parallel, and improved memory. They're better at integrating web searches as needed, and to check for supporting information and make sure they're on the right track with their answers. New AI model launches usually come with benchmark charts showing improvements—and this one is no different. Credit: Anthropic Also new are "thinking summaries" that give more insight into how Claude 4 has reached its conclusions, and an "extended thinking" feature, launching in beta, that lets you force the AI bot to take more time mulling over its responses. Anthropic is now making its Claude Code suite of tools available more generally as well, another step towards agentic AI that can work autonomously, without continuous help from flesh and blood users. In a demo video, Claude 4 models are shown compiling research papers from the web, putting together an online ordering system, and extracting information from documents to create actionable tasks.Claude 4 is available now (but you'll need to pay for the more advanced model)The Claude Sonnet 4 model, which is faster and doesn't have quite the same capacity in terms of thinking, coding, and memory, is available now to all Claude users. The more advanced Claude Opus 4, which also includes extra tools and integrations, is available to users on any of Anthropic's paid subscriptions.The path to releasing these Claude 4 models wasn't all smooth: Anthropic says its safety advice partner warned against releasing earlier versions of the models because of their tendency to "'scheme' and deceive." Those issues have now been worked out, apparently, but it's a reminder that as AI models get increasingly powerful, they also need to come with improved guardrails and safety features attached. The new models are available inside Claude now. Credit: Lifehacker I'm not really a coder, so I can't comment with any real authority on the primary upgrades included with Claude 4, but I have been able to test out the extended reasoning and thinking capabilities of Claude Sonnet 4 and Claude Opus 4. These capabilities aren't easy to quantify or measure, but all the responses I got were well written and well presented, and as far as I could tell provided accurate information, with online citations.To be honest, I'm always a bit stuck when it comes to how to make full use of AI chatbots and their latest upgrades. They can definitely save time when running certain web searches and researching topics online, but I don't fully trust the results, or AI's ability to decide what is relevant and what isn't—I'd still much rather do the reading and summarizing myself, even if it's slower. There's a new Extended Thinking Mode you can make use of. Credit: Lifehacker Maybe I need to start a coding project and see how far I can get on vibes alone. I did ask Claude Opus 4 to build me a simple HTML time tracker I could run in a browser tab, to make sure I wasn't spending too much time distracted during the day. It did the job in a couple of minutes, and produced something that worked well, closely matching the instructions I gave. While it functioned fine, Claude 4 reported a couple of errors along the way, which of course I didn't understand—I guess I can ask the AI about them.Anthropic isn't the only AI company with new models to tout. At Google I/O 2025 earlier this week, the company unveiled improved coding assistance and thought summaries in Gemini, following on from the announcement of its best AI models yet a few weeks ago. OpenAI, meanwhile, has been testing its GPT-4.5 model since February, touting improvements in coding and problem solving.

As Unity’s Scriptable Render Pipeline (SRP)’s available feature set continues to grow, so does the amount of shader variants being processed and compiled at build time. Alongside ongoing support for additional graphics APIs and an ever-growing selection of target platforms, the SRP’s improvements continue to expand.Shaders are compiled and cached after an initial (“clean”) build, thus accelerating further incremental (“warm”) builds. While clean builds usually take the longest, lengthy warm build times can be a common pain point during project development and iteration.To address this problem, Unity’s Shader Management team has been hard at work to provide meaningful and scalable solutions. This has resulted in significantly reduced shader build times and runtime memory usage for projects created using Unity 2021 LTS and later versions.To read more about these new optimizations, including affected versions, backports, and figures from our internal testing, skip directly to the sections covering shader variantprefiltering and dynamic shader loading. At the end of this blog post, we also address our future plans to further refine shader variant management as a whole – across project authoring, build, and runtime.Before delving into the exciting improvements made to Unity’s shader system, let’s also take the opportunity to quickly review the concepts of conditional shader compilation, shader variants, and shader variant stripping.Conditional shader features enable developers and artists to conveniently control and alter a shader’s functionality using scripts, material settings, as well as project and graphics settings. Such conditional features serve to simplify project authoring, allowing projects to efficiently scale by minimizing the number of shaders you’ll have to author and maintain.Conditional shader features can be implemented in different ways:Static (compile-time) branchingShader variants compilationDynamic (runtime) branchingWhile static branching avoids branching-related shader execution overhead at runtime, it’s evaluated and locked at compilation time and does not provide runtime control. Shader variant compilation, meanwhile, is a form of static branching that provides additional runtime control. This works by compiling a unique shader program (variant) for every possible combination of static branches, in order to maintain optimal GPU performance at runtime.Such variants are created by conditionally declaring and evaluating shader functionality through shader_feature and multi_compile shader keywords. The correct shader variants are loaded at run time based on active keywords and runtime settings. Declaring and evaluating additional shader keywords can lead to an increase in build time, file size, and runtime memory usage.At the same time, dynamic (uniform-based) branching entirely avoids the overhead of shader variant compilation, resulting in faster builds and both reduced file size and memory usage. This can bring forth smoother and faster iteration during development.On the other hand, dynamic branching can have a strong impact on shader execution performance based on the shader’s complexity and the target device. Asymmetric branches, where one side of the branch is much more complex than the other, can negatively impact performance. This is because the execution of a simpler path can still incur the performance penalties of the more complex path.When introducing conditional shader features in your own shaders, these approaches and trade-offs should be kept in mind. For more detailed information, see the shader conditionals, shader branching, and shader variants documentation.To mitigate the increase in shader processing and compilation time, shader variant stripping is utilized. It aims to exclude unnecessary shader variants from compilation based on factors such as:Materials included and keywords enabledProject and Render Pipeline settingsScriptable strippingWhen enumerating shader variants, the Editor will automatically filter out any keywords declared with shader_feature that are not enabled by materials referenced and included in the build. As a result, these keywords will not generate any additional variants.For example, if the Clear Coat material property is not enabled by any material using the Complex Lit URP Shader, all shader variants that implement the Clear Coat functionality will safely be stripped at build time.In the meantime, multi_compile keywords prompt developers and players to freely control the shader’s functionality at runtime based on available Player settings and scripts. The flip side is that such keywords cannot automatically be stripped by the Editor to the same degree as shader_feature keywords. That’s why they generally produce a larger number of variants.Scriptable stripping is a C# API that lets you exclude shader variants from compilation during build time via keywords and combinations not required at runtime. The render pipelines utilize scriptable stripping in order to strip unnecessary variants according to the project’s Render Pipeline settings and Quality Assets included in the build. Low quality High quality Variant multiplier Main Light/Cast Shadows: Off On 2x Main Light/Cast Shadows: On On 1x Main Light/Cast Shadows: Off Off 1xIn order to maximize the effects of the Editor’s shader variant stripping, we recommend disabling all graphics-related features and Render Pipeline settings not utilized at runtime. Please refer to the official documentation for more on shader variant stripping.Shader variant stripping greatly reduces the amount of compiled shader variants, based on factors like the Render Pipeline Quality Assets in the build. However, stripping is currently performed at the end of the shader processing stage. Simply enumerating all the possible variants can still take a long time, regardless of compilation.In order to reduce the shader variant processing (and project build) times, we are now introducing a significant optimization to the engine’s built-in shader variant stripping. With shader variant prefiltering, both clean and warm build times are significantly reduced.The optimization works by introducing the early exclusion of multi_compile keywords, according to Prefiltering Attributes driven by Render Pipeline settings. This decreases the amount of variants being enumerated for potential stripping and compilation, which in turn, reduces shader processing time – with warm build times reduced byup to 90% in the most drastic examples.Shader variant prefiltering first landed in 2023.1.0a14, and has been backported to 2022.2.0b15 and 2021.3.15f1.Variant prefiltering also helps cut down initial/clean build times by applying the same principle.Historically, the Unity runtime would front-load all shader objects from disk to CPU memory during scene and resource load. In most cases, a built project and scene includes many more shader variants than needed at any given moment during the application’s runtime. For projects using a large amount of shaders, this often results in high shader memory usage at runtime.Dynamic shader loading addresses the issue by providing refined user control over shader loading behavior and memory usage. This optimization facilitates the streaming of shader data chunks into memory, as well as the eviction of shader data that is no longer needed at runtime, based on a user controlled memory budget. This allows you to significantly reduce shader memory usage on platforms with limited memory budgets.New Shader Variant Loading Settings are now accessible from the Editor’s Player Settings. Use them to override the maximum number of shader chunks loaded and per-shader chunk size (MB).With the following C# API now available, you can override the Shader Variant Loading Settings using Editor scripts, such as:PlayerSettings.SetDefaultShaderChunkCount and PlayerSettings.SetDefaultShaderChunkSizeInMB to override the project’s default shader loading settingsPlayerSettings.SetShaderChunkCountForPlatform and PlayerSettings.SetShaderChunkSizeInMBForPlatformto override these settings on a per-platform basisYou can also override the maximum amount of loaded shader chunks at runtime using the C# API via Shader.maximumChunksOverride. This enables you to override the shader memory budget based on factors such as the total available system and graphics memory queried at runtime.Dynamic shader loading landed in 2023.1.0a11 and has been backported to 2022.2.0b10, 2022.1.21f1,and 2021.3.12f. In the case of the Universal Render Pipeline (URP)’s Boat Attack, we observed a78.8% reduction in runtime memory usage for shaders, from 315 MiB (default) to 66.8 MiB (dynamic loading). You can read more about this optimization in the official announcement.Beyond the critical changes mentioned above, we are working to enhance the Universal Render Pipeline’s shader variant generation and stripping. We’re also investigating additional improvements to Unity’s shader variant management at large. The ultimate goal is to facilitate the engine’s increasing feature set, while ensuring minimal shader build and runtime overhead.Some of our ongoing investigations involve the deduplication of shader resources across similar variants, as well as overall improvements to the shader keywords and Shader Variant Collection APIs. The aim is to provide more flexibility and control over shader variant processing and runtime performance.Looking ahead, we are also exploring the possibility of in-Editor tooling for shader variant tracing and analysis to provide the following details on shader variant usage:Which shaders and keywords produce the most variants?Which variants are compiled but unused at runtime?Which variants are stripped but requested at runtime?Your feedback has been instrumental so far as it helps us prioritize the most meaningful solutions. Please check out our public roadmap to vote on the features that best suit your needs. If there are additional changes you’d like to see, feel free to submit a feature request, or contact the team directly in this shader forum.