It’s easy to look at President Donald Trump’s second term and conclude that the less power and reach the federal government has, the better. After all, a smaller government might provide Trump or someone like him with fewer opportunities to disrupt people’s lives, leaving America less vulnerable to the whims of an aspiring autocrat. Weaker law-enforcement agencies could lack the capacity to enforce draconian policies. The president would have less say in how universities like Columbia conduct their business if they weren’t so dependent on federal funding. And he would have fewer resources to fundamentally change the American way of life.Trump’s presidency has the potential to reshape an age-old debate between the left and the right: Is it better to have a big government or a small one? The left, which has long advocated for bigger government as a solution to society’s problems, might be inclined to think that in the age of Trump, a strong government may be too risky. Say the United States had a single-payer universal health care system, for example. As my colleague Kelsey Piper pointed out, the government would have a lot of power to decide what sorts of medical treatments should and shouldn’t be covered, and certain forms of care that the right doesn’t support — like abortion or transgender health — would likely get cut when they’re in power. That’s certainly a valid concern. But the dangers Trump poses do not ultimately make the case for a small or weak government because the principal problem with the Trump presidency is not that he or the federal government has too much power. It’s that there’s not enough oversight.Reducing the power of the government wouldn’t necessarily protect us. In fact, “making government smaller” is one of the ways that Trump might be consolidating power.First things first: What is “big government”?When Americans are polled about how they feel about “big government” programs — policies like universal health care, Social Security, welfare for the poor — the majority of people tend to support them. Nearly two-thirds of Americans believe the government should be responsible for ensuring everyone has health coverage. But when you ask Americans whether they support “big government” in the abstract, a solid majority say they view it as a threat.That might sound like a story of contradictions. But it also makes sense because “big government” can have many different meanings. It can be a police state that surveils its citizens, an expansive regulatory state that establishes and enforces rules for the private sector, a social welfare state that directly provides a decent standard of living for everyone, or some combination of the three. In the United States, the debate over “big government” can also include arguments about federalism, or how much power the federal government should have over states. All these distinctions complicate the debate over the size of government: Because while someone might support a robust welfare system, they might simultaneously be opposed to being governed by a surveillance state or having the federal government involved in state and local affairs.As much as Americans like to fantasize about small government, the reality is that the wealthiest economies in the world have all been a product of big government, and the United States is no exception. That form of government includes providing a baseline social safety net, funding basic services, and regulating commerce. It also includes a government that has the capacity to enforce its rules and regulations.A robust state that caters to the needs of its people, that is able to respond quickly in times of crisis, is essential. Take the Covid-19 pandemic. The US government, under both the Trump and Biden administrations, was able to inject trillions of dollars into the economy to avert a sustained economic downturn. As a result, people were able to withstand the economic shocks, and poverty actually declined. Stripping the state of the basic powers it needs to improve the lives of its citizens will only make it less effective and erode people’s faith in it as a central institution, making people less likely to participate in the democratic process, comply with government policies, or even accept election outcomes.A constrained government does not mean a small governmentBut what happens when the people in power have no respect for democracy? The argument for a weaker and smaller government often suggests that a smaller government would be more constrained in the harm it can cause, while big government is more unrestrained. In this case, the argument is that if the US had a smaller government, then Trump could not effectively use the power of the state — by, say, deploying federal law enforcement agencies or withholding federal funds — to deport thousands of immigrants, bully universities, and assault fundamental rights like the freedom of speech. But advocating for bigger government does not mean you believe in handing the state unlimited power to do as it pleases. Ultimately, the most important way to constrain government has less to do with its size and scope and more to do with its checks and balances. In fact, one of the biggest checks on Trump’s power so far has been the structure of the US government, not its size. Trump’s most dangerous examples of overreach — his attempts to conduct mass deportations, eliminate birthright citizenship, and revoke student visas and green cards based on political views — have been an example of how proper oversight has the potential to limit government overreach. To be sure, Trump’s policies have already upended people’s lives, chilled speech, and undermined the principle of due process. But while Trump has pushed through some of his agenda, he hasn’t been able to deliver at the scale he promised. But that’s not because the federal government lacks the capacity to do those things. It’s because we have three equal branches of government, and the judicial branch, for all of its shortcomings in the Trump era, is still doing its most basic job to keep the executive branch in check. Reforms should include more oversight, not shrinking governmentThe biggest lesson from Trump’s first term was that America’s system of checks and balances — rules and regulations, norms, and the separate branches of government — wasn’t strong enough. As it turned out, a lot of potential oversight mechanisms did not have enough teeth to meaningfully restrain the president from abusing his power. Trump incited an assault on the US Capitol in an effort to overturn the 2020 election, and Congress ultimately failed in its duty to convict him for his actions. Twice, impeachment was shown to be a useless tool to keep a president in check.But again that’s a problem of oversight, not of the size and power of government. Still, oversight mechanisms need to be baked into big government programs to insulate them from petty politics or volatile changes from one administration to the next. Take the example of the hypothetical single-payer universal health care system. Laws dictating which treatments should be covered should be designed to ensure that changes to them aren’t dictated by the president alone, but through some degree of consensus that involves regulatory boards, Congress, and the courts. Ultimately, social programs should have mechanisms that allow for change so that laws don’t become outdated, as they do now. And while it’s impossible to guarantee that those changes will always be good, the current system of employer-sponsored health insurance is hardly a stable alternative.By contrast, shrinking government in the way that Republicans often talk about only makes people more vulnerable. Bigger governments — and more bureaucracy — can also insulate public institutions from the whims of an erratic president. For instance, Trump has tried to shutter the Consumer Financial Protection Bureau (CFPB), a regulatory agency that gets in the way of his and his allies’ business. This assault allows Trump to serve his own interests by pleasing his donors.In other words, Trump is currently trying to make government smaller — by shrinking or eliminating agencies that get in his way — to consolidate power. “Despite Donald Trump’s rhetoric about the size or inefficiency of government, what he has done is eradicate agencies that directly served people,” said Julie Margetta Morgan, president of the Century Foundation who served as an associate director at the CFPB. “He may use the language of ‘government inefficiency’ to accomplish his goals, but I think what we’re seeing is that the goals are in fact to open up more lanes for big businesses to run roughshod over the American people.” The problem for small-government advocates is that the alternative to big government is not just small government. It’s also big business because fewer services, rules, and regulations open up the door to privatization and monopolization. And while the government, however big, has to answer to the public, businesses are far less accountable. One example of how business can replace government programs is the Republicans’ effort to overhaul student loan programs in the latest reconciliation bill the House passed, which includes eliminating subsidized loans and limiting the amount of aid students receive. The idea is that if students can’t get enough federal loans to cover the cost of school, they’ll turn to private lenders instead. “It’s not only cutting Pell Grants and the affordability of student loan programs in order to fund tax cuts to the wealthy, but it’s also creating a gap where [private lenders] are all too happy to come in,” Margetta Morgan said. “This is the small government alternative: It’s cutting back on programs that provided direct services for people — that made their lives better and more affordable — and replacing it with companies that will use that gap as an opportunity for extraction and, in some cases, for predatory services.”Even with flawed oversight, a bigger and more powerful government is still preferable because it can address people’s most basic needs, whereas small government and the privatization of public services often lead to worse outcomes.So while small government might sound like a nice alternative when would-be tyrants rise to power, the alternative to big government would only be more corrosive to democracy, consolidating power in the hands of even fewer people (and businesses). And ultimately, there’s one big way for Trump to succeed at destroying democracy, and that’s not by expanding government but by eliminating the parts of government that get in his way.See More:

The European Commission has published its full Digital Markets Act (DMA) decision against Apple, and it’s far, far worse than anybody expected. The Commission, the executive arm of the European Union, has accepted absolutely none of Apple’s arguments against being fined, and the decision threatens yet more existential damage to the company. Apple isn’t winning the argument, and, right or wrong, the decision has fangs. Huge fines, big threats Europe announced in April that it would fine Apple an eye-popping €500 million for noncompliance with the DMA, giving Apple 60 days to comply with its decision. One month later, the Commission published the full ruling against Apple, which details that changes the company made to its App Store rules did not go far enough to bring it into compliance. The decision warns that Apple is subject to additional periodic fines in the future if it fails to comply with the Commission’s strict interpretation of the DMA, no matter how inherently punitive some of its demands may be. (Can anyone else spell “tariffs”?) We’ll know soon enough if there are to be wider consequences to Europe’s demands. Apple now has 30 days to fully comply with the DMA (in Europe’s opinion) or face additional fines. The act itself came into force in November 2022 and began to be implemented against companies defined as ‘gatekeepers’ in 2023. The intention is to stop Apple and others from using their market position to impose anticompetitive limitations on developers.  Who is steering? The big bugbear relates to Apple’s anti-steering restrictions, which prevent developers from telling customers they can purchase services outside the App Store. The DMA demands that Apple let developers offer this option, which Apple does, but Europe argues that the limitations the company makes on doing so are not in compliance with the law. Europe also says Apple’s existing restrictions, fees, and technical limitations undermine the effectiveness of the DMA. That seems to mean Apple cannot charge a commission and cannot warn users of the consequences they face when shopping outside the App Store.  The Commission even plays dumb to the potential significance of permitting developers to link out to any website from within their apps, rather than being constrained to approved (and secure) sites. It says Apple has provided insufficient justification for this restriction and also wants Apple to remove messages warning users when they are about to make a transaction outside the App Store.  That’s going to be particularly pleasing to fraudsters, who may now attempt to create fake payment portals that look like reputable ones. Apple prevented $2 billion in fraud last year, the company has confirmed. Perhaps once the first big frauds take place, the EU may catch up to the online risks we all know exist. While I understand the original aim of Europe’s Digital Markets Act, the demands the Commission is making of Apple appear to go far beyond the original objective, which was to open up Apple’s platforms to competition.  The decisions now open Apple’s platform up to competitors.  There is a difference between the two, and, as described, it means Apple must now create and manage its platforms while permitting competitors to profit from those platforms at little or no cost. Apple rejects Europe Apple will fight in Europe.  “There is nothing in the 70-page decision released today that justifies the European Commission’s targeted actions against Apple, which threaten the privacy and security of our users in Europe and force us to give away our technology for free,” the company said. “Their decision and unprecedented fine came after the Commission continuously moved the goalposts on compliance, and repeatedly blocked Apple’s months-long efforts to implement a new solution. The decision is bad for innovation, bad for competition, bad for our products, and bad for users. While we appeal, we’ll continue engaging with the Commission to advocate on behalf of our European customers.” When the fine was initially revealed, the company also said:  “Today’s announcements are yet another example of the European Commission unfairly targeting Apple in a series of decisions that are bad for the privacy and security of our users, bad for products, and force us to give away our technology for free. We have spent hundreds of thousands of engineering hours and made dozens of changes to comply with this law, none of which our users have asked for. Despite countless meetings, the Commission continues to move the goal posts every step of the way.” My take?  Far from saving Europe’s tech industry, the manner in which the DMA is being applied will make the region even less relevant. Lacking a significant platform of its own, Europe’s approach will reduce choice and increase insecurity. As the clear first target of the DMA, Apple will inevitably be forced to increase prices, charge developers more for access to its developer tools, and will I think simply stop selling some products and services in Europe, rather than threaten customer security. We know it can do this because it has done so before. Fundamentally, of course, the big question remains unaddressed: How much profit it is legitimate to make on any product or service? I imagine the European Commission doesn’t want to go near a question as fundamental to capitalist wealth extraction as that. Can you imagine the collapse in executive bonuses that would follow a decision to define what the maximum profit made in any business transaction should be? Lobbyists across the political spectrum would be appalled — that extra profit pays for their meals. Looking to the extent to which the current application of the DMA seems to favor Apple’s biggest competitors, I can’t help but imagine it’s been paying for a few European meals already. Nice work, if you can get it.  You can follow me on social media! Join me on BlueSky,  LinkedIn, Mastodon, and MeWe.

Microsoft and Google are taking distinctive approaches with AI agents in their productivity suites, and enterprises need to account for the differences when formulating digital labor strategies, analysts said. In recent months, both companies have announced a dizzying array of new agents aimed at extracting value from corporate documents and maximizing efficiency. The tech giants have dropped numerous hints about where they’re headed with AI agents in their respective office suites, Microsoft 365 and Google Workspace. Microsoft is reshaping its Copilot assistant as a series of tools to create, tap into, and act on insights at individual and organizational levels. The Microsoft 365 roadmap lists hundreds of specialized AI tools under development to automate work for functions such as HR and accounting. The company is also developing smaller AI models to carry out specific functions. Google is going the opposite way, with its large-language model Gemini at the heart of Workspace. Google offers tools that include Gems for workers to create simple custom agents that automate tasks such as customer service, and Agentspace in Google Cloud to build more complex custom agents for collaboration and workflow management. At the recent Google I/O developer conference, the company added real-time speech translation to Google Meet. “For both, the goal is to bring usable and practical productivity and efficiency capabilities to work tools,” said Liz Miller, vice president and principal analyst at Constellation Research. But the differing AI agent strategies are heavily rooted in each company’s philosophical approaches to productivity. Although Microsoft has long encouraged customers to move from its traditional “perpetual-license” Office suite to the Microsoft 365 subscription-based model, M365 notably retains the familiar desktop apps. Google Workspace, on the other hand, has always been cloud-based. Microsoft users are typically a bit more tethered to traditional enterprise work styles, while Google has always been the “cloud-first darling for smaller organizations that still crave real-time collaboration,” Miller said. When it comes to the generative AI models being integrated into the two office suites, “Google’s Gemini models are beating out the models being deployed by Microsoft,” Miller said. “But as Microsoft expands its model ‘inventory’ in use across M365, this could change.” Microsoft has an advantage, as many desktop users live in Outlook or Word. The intelligence Copilot can bring from CRM software is readily available, while that integration is more complex in the cloud-native Google Workspace. “Microsoft still has an edge in a foundational understanding of work and the capacity to extend Copilot connections across applications as expansive as the Office suite through to Dynamics, giving AI a greater opportunity to be present in the spaces and presentation layers where workers enjoy working,” Miller said. Microsoft’s Copilot Agents and Google’s Gems and Agentspace are in their early stages, but there have been positive developments, said J.P. Gownder, a vice president and principal analyst on Forrester’s Future of Work team. Microsoft recently adopted Google’s A2A protocol, which makes it easier for users of both productivity suites to collaborate and unlock value from stagnant data sitting on other platforms. “That should be a win for interoperability,” Gownder said. But most companies that are Microsoft shops have years or decades of digital assets that hold them back from considering Google, he said. For example, Excel macros, pivot tables, and customizations cannot be easily or automatically migrated to Google Sheets, he said. “As early as this market is, I don’t think it’s fair to rank either player — Microsoft or Google — as being the leader; both of them are constructing new ecosystems to support the growth of agentic AI,” Gownder said. Most Microsoft Office users have moved to M365, but AI is helping Google is making inroads into larger organizations, especially among enterprises that are newer and less oriented toward legacy Microsoft products, said Jack Gold, principal analyst at J. Gold Associates. Technologies like A2A blur the line between on-premises and cloud productivity. As a result, “Google Workspace is no longer perceived as inferior, as it had been in the past,” Gold said. And for budget-constrained enterprises, the value of AI agent features is not the only consideration. “There is also the cost equation at work here, as Google seems to have a much more transparent cost structure than Microsoft with all of its user classes and discounts,” Gold said. Microsoft does not include Copilot in its M365 subscriptions, which vary in price depending on the type of customer. The Copilot business subscriptions range from $30 per user per month for M365 Copilot to $200 per month for 25,000 messages for Copilot Studio, which is also available under a pay-as-you-go model. Google has flat subscription pricing for Workspace, starting at $14 per user per month for business plans with Gemini included.

With just days to go before WWDC, the consensus is that Apple will unveil a big, visionOS-inspired redesign across its operating systems. And while some might be dreading a repeat of the iOS 7 announcement from a decade ago, it’s been long enough that many readers might not remember (or may have never even seen) what that overhaul actually looked like. So here’s a quick refresher on what happened, and why this year will likely (I mean, hopefully?) be different. The years between 2011 and 2013 were pretty busy at Apple. Following Steve Jobs’ passing, Apple fired Scott Forstall (then SVP of iOS Software) over the botched release of Apple Maps. That left a gap in software design leadership, which was filled by Jony Ive, who also led hardware design. Soon after, rumors began swirling that he was planning a major visual overhaul of the entire system. Flat In the run-up to WWDC 2013, the Wall Street Journal reported that Ive had been working on “a more ‘flat design’ that is starker and simpler,” a sharp departure from the great skeuomorphic visuals of the time (think linen textures, paper-like folders, glass effects, and yes, Corinthian leather). Some time after that, 9to5Mac exclusively shared mockups of the redesign, which had been leaked to Mark Gurman. It was chaos. I vividly remember thinking it was reckless to publish such unfairly primitive sketches of what would certainly be a more polished overhaul. After weeks of intense debate and fierce expectations that the rumors had been wrong, Apple introduced iOS 7: In the years that followed, Apple scaled back its over-flattening of the system, evolving toward what we have today. Now, that’s about to change once again. Why iOS 26 probably won’t be like iOS 7 Currently, most reports tend to agree that the redesign will be deeply influenced by the visual language of visionOS, with its translucent layers, depth effects, and soft glassy textures. And even if you’re like me and you’ve never worn an Apple Vision Pro, chances are you’ve seen what visionOS looks like. Apple has already laid the groundwork, so the change won’t be such a jarring surprise, like with iOS 7. And from a design perspective, speaking as someone who’s worked in graphic design for over two decades, the best move Apple could make is exactly what’s been reported: updating all systems at once. If you’ve ever had to adapt interfaces and key visuals to multiple concepts, such as wide, narrow, square, rectangular, big, small, etc., you know that with every new aspect ratio, you become a little more familiar and more comfortable with each individual element. By starting out with the virtually boundless, unconstrained environment of visionOS, then increasingly moving to smaller interfaces across macOS, iPadOS, iOS, and watchOS, every decision informs past and future visual adaptations. In other words, a redesign this broad can be iterative in both directions. Will it be beautiful? That’s subjective. Even iOS 7 had a handful of defenders. But one thing is certain: Apple’s design team knows how much this moment matters. This is the biggest task they’ve been given since Ive left the company, and they are well aware of the contentious history of iOS design updates. The mere fact that the new design hasn’t leaked yet points to the absence of dissidents inside the team, and considering how close we are to the announcement, that’s already a victory in itself. Add 9to5Mac to your Google News feed.  FTC: We use income earning auto affiliate links. More.You’re reading 9to5Mac — experts who break news about Apple and its surrounding ecosystem, day after day. Be sure to check out our homepage for all the latest news, and follow 9to5Mac on Twitter, Facebook, and LinkedIn to stay in the loop. Don’t know where to start? Check out our exclusive stories, reviews, how-tos, and subscribe to our YouTube channel

With millions of businesses now using Amazon Web Services (AWS) for their cloud computing needs, it’s become a vital consideration for IT security teams and professionals. As such, AWS offers a broad range of cyber security tools to secure AWS-based tech stacks. They cover areas such as data privacy, access management, configuration management, threat detection, network security, vulnerability management, regulatory compliance and so much more.  Along with being broad in scope, AWS security tools are also highly scalable and flexible. Therefore, they’re ideal for high-growth organisations facing a fast-expanding and increasingly sophisticated cyber threat landscape. On the downside, they can be complex to use, don’t always integrate well with multi-cloud environments, and become outdated and expensive quickly. These challenges underscore the importance of continual learning and effective cost management in the cyber security suite. One of the best things AWS offers cyber security professionals is a centralised view of all their different virtual environments, including patch management, vulnerability scanning and incident response, to achieve “smoother operations”, according to Richard LaTulip, field chief information security officer at cyber threat intelligence platform Recorded Future. Specifically, he says tools like AWS CloudTrail and AWS Config allow cyber security teams to accelerate access management, anomaly detection and real-time policy compliance, and that risk orchestration is also possible thanks to AWS’s support for specialist platforms such as Recorded Future.  This sentiment is echoed by Crystal Morin, cyber security strategist at container security firm Sysdig, who describes AWS CloudTrail and AWS GuardDuty as “the bedrock” for organisations with a multi- or hybrid cloud environment.  She says these tools offer “great insight” into cloud environment activity that can be used to identify issues affecting corporate systems, better understand them and ultimately determine their location for prompt removal.  Having made tons of cloud security deployments for Fortune 200 companies in his previous role as global AWS security lead at consulting giant Accenture, Shaan Mulchandani, founder and CEO of cloud security firm HTCD, knows a thing or two about AWS’s cyber security advantages.  Mulchandani says AWS implementations helped these companies secure their baseline configurations, streamline C-suite IT approvals to speed up AWS migration, eliminate manual post-migration security steps and seamlessly scale environments containing thousands of workloads. “I continue to help executives at organisations architect, deploy and maximise outcomes using AWS-native tools,” he adds. As a senior threat researcher at cyber intelligence platform EclecticIQ, Arda Büyükkaya uses AWS tools to scale threat behaviour analysis, develop secure malware analysis environments, and automate threat intelligence data collection and processing.  Calling AWS an “invaluable” threat analysis resource, he says the platform has made it a lot easier to roll out isolated research environments. “AWS’s scalability enables us to process large volumes of threat data efficiently, whilst their security services help maintain the integrity of our research infrastructure,” Büyükkaya tells Computer Weekly. At log management and security analytics software company Graylog, AWS usage happens across myriad teams. One of these is led by EMEA and UK lead Ross Brewer. His department is securing and protecting customer instances using tools like AWS GuardDuty, AWS Security Hub, AWS Config, AWS CloudTrail, AWS Web Application Firewall (WAF), AWS Inspector and AWS Identity and Access Management (IAM).  Its IT and application security department also relies on security logs provided by AWS GuardDuty and AWS CloudTrail to spot anomalies affecting customer instances. Brewer says the log tracking and monitoring abilities of these tools have been invaluable for security, compliance and risk management. “We haven’t had any issues with our desired implementations,” he adds. Cyber law attorney and entrepreneur Andrew Rossow is another firm believer in AWS as a cyber security tool. He thinks its strongest aspect is the centralised security management it offers for monitoring threats, responding to incidents and ensuring regulatory compliance, and describes the usage of this unified, data-rich dashboard as the “difference between proactive defence and costly damage control” for small businesses with limited resources.  But Rossow believes this platform’s secret sauce is its underlying artificial intelligence (AI) and machine learning models, which power background threat tracking, and automatically alert users to security issues, data leaks and suspicious activity. These abilities, he says, allow cyber security professionals to “stay ahead of potential crises”. Another area where Rossow thinks AWS excels is its integration with regulatory frameworks such as the California Consumer Privacy Act, the General Data Protection Regulation and the Payment Card Industry Data Security Standard. He explains that AWS Config and AWS Security Hub offer configuration and resource auditing to ensure business activities and best practices meet such industry standards. “This not only protects our clients, but also shields us from the legal and reputational fallout of non-compliance,” adds Rossow. AWS tools provide cyber security teams with “measurable value”, argues Shivraj Borade, senior analyst at management consulting firm Everest Group. He says GuardDuty is powerful for real-time monitoring, AWS Config for security posture management and IAM Access Analyzer for privilege sprawl prevention. “What makes these tools powerful is their interoperability, enabling a scalable and cohesive security architecture,” says Borade. Although AWS is a valuable tool for cyber security professionals, Borade emphasises that it’s “not without limitations”. He says the platform’s lack of depth and flexibility means it isn’t always suitable for modelling complex cyber security threats or handling specific compliance issues. Rather, cyber security professionals should use AWS as a foundational element of their wider tech stack.  Using the AWS Security Hub as an example, Borade says it can effectively serve the purpose of an “aggregation layer”. But he warns that incorrect configurations often result in alert fatigue, meaning people can become oblivious to notifications when repeatedly spammed with them.  Borade also warns of misconfigurations arising from teams’ lack of understanding of how cloud technology works. Consequently, he urges cyber security teams to “embed cloud-native security into the DevSecOps lifecycle” and “invest in continuous cross-functional training”. For Morin, the biggest challenge of using AWS as a security tool is that it’s constrained by best practice gaps around areas like workload protection, vulnerability management, identity management and threat detection. She says one classic example is the difficulty cyber security teams face when monitoring access permissions granted over time, leaving organisations with large IT environments dangerously exposed.  Using multiple AWS security tools also increases the attack surface for cyber criminals to exploit. Morin warns that hackers may look for “visibility gaps” by sifting through different AWS planes, helping them “mask their activities” and “effectively bypass detection”. To stay one step ahead of cyber crooks, she advises organisations to invest in runtime solutions alongside AWS-native tools. These will provide real-time security insights. Technical and cost issues may also impact AWS implementations in cyber security departments, warns Mulchandani. For instance, Amazon Macie may be able to create inventories for all object versions across different buckets, but Mulchandani says this creates a “mountain of medium-severity findings” to decipher. “Without strict scoping, licence costs and analyst time balloon,” he adds. “Costs can also increase when an organisation requires a new AWS launch that isn’t available in their region and they subsequently invest in a temporary solution from a different vendor. For those new to using AWS security tools, Morin says an important first step is to understand the cloud security shared responsibility model. She explains that the user is responsible for securing their deployments, correctly configuring them and closing any security visibility gaps. AWS, on the other hand, must ensure the underlying infrastructure provided is safe to use.  As part of the users’ role in this model, she says they should enable logging and alerts for AWS tools and services used in their organisation. What’s also key is detailing standard organisational operating behaviour in a security baseline. This, she claims, will let organisations tell suspicious user actions apart from normal ones. Many tried-and-tested best practices can be found in professional benchmarks such as the AWS Well-Architected framework and the Center of Internet Security’s Benchmark for AWS. “Make use of the work of those who have been fighting the good fight,” says Morin. Finally, she urges anyone working in cloud security to remember that real-time operations are essential. Runtime security can help by protecting all running applications and data from the latest cyber security threats, many of which are preventable through automated processes.  Starting small is a good idea, too. Mulchandani recommends that AWS newbies begin with AWS tooling, and if any gaps persist, they can then look for third-party offerings. “Do not try to procure and integrate 20-plus external tools upfront as this will cause numerous architectural, security and cost challenges,” he says. With the rapid pace of innovation across the AWS ecosystem, Borade urges anyone using this platform to stay up-to-date with the latest releases by participating in certification programmes, attending re:Inforce sessions and tracking the latest release notes from AWS. In the future, he expects automation, AI-fuelled insights, “tighter” third-party integrations, and identity orchestration and policy-as-code frameworks to dominate the AWS cyber security ecosystem.  On the whole, understanding the AWS platform and its role in cloud security is a vital skill for cyber security professionals. And AWS certainly offers some great tools for managing the biggest risks impacting its popular cloud platform. But cyber security professionals looking to leverage AWS in their day-to-day roles must be willing to get to grips with some complex tools, keep up-to-date with the latest releases in the vast AWS ecosystem and ensure their department budget can accommodate spiralling AWS costs. Read more about AWS An AWS tech stack can aid business growth and facilitate efficient operations, but misconfigurations have become all too common and stall this progress. The AWS Summit in London saw the public cloud giant appoint itself to take on the task of skilling up hundreds of thousands of UK people in using AI technologies. Amazon Web Services debuts new Outposts racks and servers that extend its infrastructure to the edge to support network intensive workloads and cloud radio access applications.

The web is beginning to part ways with third-party cookies, a technology it once heavily relied on. Introduced in 1994 by Netscape to support features like virtual shopping carts, cookies have long been a staple of web functionality. However, concerns over privacy and security have led to a concerted effort to eliminate them. The World Wide Web Consortium Technical Architecture Group (W3C TAG) has been vocal in advocating for the complete removal of third-party cookies from the web platform. Major browsers (Chrome, Safari, Firefox, and Edge) are responding by phasing them out, though the transition is gradual. While this shift enhances user privacy, it also disrupts legitimate functionalities that rely on third-party cookies, such as single sign-on (SSO), fraud prevention, and embedded services. And because there is still no universal ban in place and many essential web features continue to depend on these cookies, developers must detect when third-party cookies are blocked so that applications can respond gracefully. Don’t Let Silent Failures Win: Why Cookie Detection Still Matters Yes, the ideal solution is to move away from third-party cookies altogether and redesign our integrations using privacy-first, purpose-built alternatives as soon as possible. But in reality, that migration can take months or even years, especially for legacy systems or third-party vendors. Meanwhile, users are already browsing with third-party cookies disabled and often have no idea that anything is missing. Imagine a travel booking platform that embeds an iframe from a third-party partner to display live train or flight schedules. This embedded service uses a cookie on its own domain to authenticate the user and personalize content, like showing saved trips or loyalty rewards. But when the browser blocks third-party cookies, the iframe cannot access that data. Instead of a seamless experience, the user sees an error, a blank screen, or a login prompt that doesn’t work. And while your team is still planning a long-term integration overhaul, this is already happening to real users. They don’t see a cookie policy; they just see a broken booking flow. Detecting third-party cookie blocking isn’t just good technical hygiene but a frontline defense for user experience. Why It’s Hard To Tell If Third-Party Cookies Are Blocked Detecting whether third-party cookies are supported isn’t as simple as calling navigator.cookieEnabled. Even a well-intentioned check like this one may look safe, but it still won’t tell you what you actually need to know: // DOES NOT detect third-party cookie blocking function areCookiesEnabled() { if (navigator.cookieEnabled === false) { return false; } try { document.cookie = "test_cookie=1; SameSite=None; Secure"; const hasCookie = document.cookie.includes("test_cookie=1"); document.cookie = "test_cookie=; Max-Age=0; SameSite=None; Secure"; return hasCookie; } catch (e) { return false; } } This function only confirms that cookies work in the current (first-party) context. It says nothing about third-party scenarios, like an iframe on another domain. Worse, it’s misleading: in some browsers, navigator.cookieEnabled may still return true inside a third-party iframe even when cookies are blocked. Others might behave differently, leading to inconsistent and unreliable detection. These cross-browser inconsistencies — combined with the limitations of document.cookie — make it clear that there is no shortcut for detection. To truly detect third-party cookie blocking, we need to understand how different browsers actually behave in embedded third-party contexts. How Modern Browsers Handle Third-Party Cookies The behavior of modern browsers directly affects which detection methods will work and which ones silently fail. Safari: Full Third-Party Cookie Blocking Since version 13.1, Safari blocks all third-party cookies by default, with no exceptions, even if the user previously interacted with the embedded domain. This policy is part of Intelligent Tracking Prevention (ITP). For embedded content (such as an SSO iframe) that requires cookie access, Safari exposes the Storage Access API, which requires a user gesture to grant storage permission. As a result, a test for third-party cookie support will nearly always fail in Safari unless the iframe explicitly requests access via this API. Firefox: Cookie Partitioning By Design Firefox’s Total Cookie Protection isolates cookies on a per-site basis. Third-party cookies can still be set and read, but they are partitioned by the top-level site, meaning a cookie set by the same third-party on siteA.com and siteB.com is stored separately and cannot be shared. As of Firefox 102, this behavior is enabled by default in the Standard (default) mode of Enhanced Tracking Protection. Unlike the Strict mode — which blocks third-party cookies entirely, similar to Safari — the Standard mode does not block them outright. Instead, it neutralizes their tracking capability by isolating them per site. As a result, even if a test shows that a third-party cookie was successfully set, it may be useless for cross-site logins or shared sessions due to this partitioning. Detection logic needs to account for that. Chrome: From Deprecation Plans To Privacy Sandbox (And Industry Pushback) Chromium-based browsers still allow third-party cookies by default — but the story is changing. Starting with Chrome 80, third-party cookies must be explicitly marked with SameSite=None; Secure, or they will be rejected. In January 2020, Google announced their intention to phase out third-party cookies by 2022. However, the timeline was updated multiple times, first in June 2021 when the company pushed the rollout to begin in mid-2023 and conclude by the end of that year. Additional postponements followed in July 2022, December 2023, and April 2024. In July 2024, Google has clarified that there is no plan to unilaterally deprecate third-party cookies or force users into a new model without consent. Instead, Chrome is shifting to a user-choice interface that will allow individuals to decide whether to block or allow third-party cookies globally. This change was influenced in part by substantial pushback from the advertising industry, as well as ongoing regulatory oversight, including scrutiny by the UK Competition and Markets Authority (CMA) into Google’s Privacy Sandbox initiative. The CMA confirmed in a 2025 update that there is no intention to force a deprecation or trigger automatic prompts for cookie blocking. As for now, third-party cookies remain enabled by default in Chrome. The new user-facing controls and the broader Privacy Sandbox ecosystem are still in various stages of experimentation and limited rollout. Edge (Chromium-Based): Tracker-Focused Blocking With User Configurability Edge (which is a Chromium-based browser) shares Chrome’s handling of third-party cookies, including the SameSite=None; Secure requirement. Additionally, Edge introduces Tracking Prevention modes: Basic, Balanced (default), and Strict. In Balanced mode, it blocks known third-party trackers using Microsoft’s maintained list but allows many third-party cookies that are not classified as trackers. Strict mode blocks more resource loads than Balanced, which may result in some websites not behaving as expected. Other Browsers: What About Them? Privacy-focused browsers, like Brave, block third-party cookies by default as part of their strong anti-tracking stance. Internet Explorer (IE) 11 allowed third-party cookies depending on user privacy settings and the presence of Platform for Privacy Preferences (P3P) headers. However, IE usage is now negligible. Notably, the default “Medium” privacy setting in IE could block third-party cookies unless a valid P3P policy was present. Older versions of Safari had partial third-party cookie restrictions (such as “Allow from websites I visit”), but, as mentioned before, this was replaced with full blocking via ITP. As of 2025, all major browsers either block or isolate third-party cookies by default, with the exception of Chrome, which still allows them in standard browsing mode pending the rollout of its new user-choice model. To account for these variations, your detection strategy must be grounded in real-world testing — specifically by reproducing a genuine third-party context such as loading your script within an iframe on a cross-origin domain — rather than relying on browser names or versions. Overview Of Detection Techniques Over the years, many techniques have been used to detect third-party cookie blocking. Most are unreliable or obsolete. Here’s a quick walkthrough of what doesn’t work (and why) and what does. Basic JavaScript API Checks (Misleading) As mentioned earlier, the navigator.cookieEnabled or setting document.cookie on the main page doesn’t reflect cross-site cookie status: In third-party iframes, navigator.cookieEnabled often returns true even when cookies are blocked. Setting document.cookie in the parent doesn’t test the third-party context. These checks are first-party only. Avoid using them for detection. Storage Hacks Via localStorage (Obsolete) Previously, some developers inferred cookie support by checking if window.localStorage worked inside a third-party iframe — which is especially useful against older Safari versions that blocked all third-party storage. Modern browsers often allow localStorage even when cookies are blocked. This leads to false positives and is no longer reliable. Server-Assisted Cookie Probe (Heavyweight) One classic method involves setting a cookie from a third-party domain via HTTP and then checking if it comes back: Load a script/image from a third-party server that sets a cookie. Immediately load another resource, and the server checks whether the cookie was sent. This works, but it: Requires custom server-side logic, Depends on HTTP caching, response headers, and cookie attributes (SameSite=None; Secure), and Adds development and infrastructure complexity. While this is technically valid, it is not suitable for a front-end-only approach, which is our focus here. Storage Access API (Supplemental Signal) The document.hasStorageAccess() method allows embedded third-party content to check if it has access to unpartitioned cookies: ChromeSupports hasStorageAccess() and requestStorageAccess() starting from version 119. Additionally, hasUnpartitionedCookieAccess() is available as an alias for hasStorageAccess() from version 125 onwards. FirefoxSupports both hasStorageAccess() and requestStorageAccess() methods. SafariSupports the Storage Access API. However, access must always be triggered by a user interaction. For example, even calling requestStorageAccess() without a direct user gesture (like a click) is ignored. Chrome and Firefox also support the API, and in those browsers, it may work automatically or based on browser heuristics or site engagement. This API is particularly useful for detecting scenarios where cookies are present but partitioned (e.g., Firefox’s Total Cookie Protection), as it helps determine if the iframe has unrestricted cookie access. But for now, it’s still best used as a supplemental signal, rather than a standalone check. iFrame + postMessage (Best Practice) Despite the existence of the Storage Access API, at the time of writing, this remains the most reliable and browser-compatible method: Embed a hidden iframe from a third-party domain. Inside the iframe, attempt to set a test cookie. Use window.postMessage to report success or failure to the parent. This approach works across all major browsers (when properly configured), requires no server (kind of, more on that next), and simulates a real-world third-party scenario. We’ll implement this step-by-step next. Bonus: Sec-Fetch-Storage-Access Chrome (starting in version 133) is introducing Sec-Fetch-Storage-Access, an HTTP request header sent with cross-site requests to indicate whether the iframe has access to unpartitioned cookies. This header is only visible to servers and cannot be accessed via JavaScript. It’s useful for back-end analytics but not applicable for client-side cookie detection. As of May 2025, this feature is only implemented in Chrome and is not supported by other browsers. However, it’s still good to know that it’s part of the evolving ecosystem. Step-by-Step: Detecting Third-Party Cookies Via iFrame So, what did I mean when I said that the last method we looked at “requires no server”? While this method doesn’t require any back-end logic (like server-set cookies or response inspection), it does require access to a separate domain — or at least a cross-site subdomain — to simulate a third-party environment. This means the following: You must serve the test page from a different domain or public subdomain, e.g., example.com and cookietest.example.com, The domain needs HTTPS (for SameSite=None; Secure cookies to work), and You’ll need to host a simple static file (the test page), even if no server code is involved. Once that’s set up, the rest of the logic is fully client-side. Step 1: Create A Cookie Test Page (On A Third-Party Domain) Minimal version (e.g., https://cookietest.example.com/cookie-check.html): <!DOCTYPE html> <html> <body> <script> document.cookie = "thirdparty_test=1; SameSite=None; Secure; Path=/;"; const cookieFound = document.cookie.includes("thirdparty_test=1"); const sendResult = (status) => window.parent?.postMessage(status, "*"); if (cookieFound && document.hasStorageAccess instanceof Function) { document.hasStorageAccess().then((hasAccess) => { sendResult(hasAccess ? "TP_COOKIE_SUPPORTED" : "TP_COOKIE_BLOCKED"); }).catch(() => sendResult("TP_COOKIE_BLOCKED")); } else { sendResult(cookieFound ? "TP_COOKIE_SUPPORTED" : "TP_COOKIE_BLOCKED"); } </script> </body> </html> Make sure the page is served over HTTPS, and the cookie uses SameSite=None; Secure. Without these attributes, modern browsers will silently reject it. Step 2: Embed The iFrame And Listen For The Result On your main page: function checkThirdPartyCookies() { return new Promise((resolve) => { const iframe = document.createElement('iframe'); iframe.style.display = 'none'; iframe.src = "https://cookietest.example.com/cookie-check.html"; // your subdomain document.body.appendChild(iframe); let resolved = false; const cleanup = (result, timedOut = false) => { if (resolved) return; resolved = true; window.removeEventListener('message', onMessage); iframe.remove(); resolve({ thirdPartyCookiesEnabled: result, timedOut }); }; const onMessage = (event) => { if (["TP_COOKIE_SUPPORTED", "TP_COOKIE_BLOCKED"].includes(event.data)) { cleanup(event.data === "TP_COOKIE_SUPPORTED", false); } }; window.addEventListener('message', onMessage); setTimeout(() => cleanup(false, true), 1000); }); } Example usage: checkThirdPartyCookies().then(({ thirdPartyCookiesEnabled, timedOut }) => { if (!thirdPartyCookiesEnabled) { someCookiesBlockedCallback(); // Third-party cookies are blocked. if (timedOut) { // No response received (iframe possibly blocked). // Optional fallback UX goes here. someCookiesBlockedTimeoutCallback(); }; } }); Step 3: Enhance Detection With The Storage Access API In Safari, even when third-party cookies are blocked, users can manually grant access through the Storage Access API — but only in response to a user gesture. Here’s how you could implement that in your iframe test page: <button id="enable-cookies">This embedded content requires cookie access. Click below to continue.</button> <script> document.getElementById('enable-cookies')?.addEventListener('click', async () => { if (document.requestStorageAccess && typeof document.requestStorageAccess === 'function') { try { const granted = await document.requestStorageAccess(); if (granted !== false) { window.parent.postMessage("TP_STORAGE_ACCESS_GRANTED", "*"); } else { window.parent.postMessage("TP_STORAGE_ACCESS_DENIED", "*"); } } catch (e) { window.parent.postMessage("TP_STORAGE_ACCESS_FAILED", "*"); } } }); </script> Then, on the parent page, you can listen for this message and retry detection if needed: // Inside the same onMessage listener from before: if (event.data === "TP_STORAGE_ACCESS_GRANTED") { // Optionally: retry the cookie test, or reload iframe logic checkThirdPartyCookies().then(handleResultAgain); } (Bonus) A Purely Client-Side Fallback (Not Perfect, But Sometimes Necessary) In some situations, you might not have access to a second domain or can’t host third-party content under your control. That makes the iframe method unfeasible. When that’s the case, your best option is to combine multiple signals — basic cookie checks, hasStorageAccess(), localStorage fallbacks, and maybe even passive indicators like load failures or timeouts — to infer whether third-party cookies are likely blocked. The important caveat: This will never be 100% accurate. But, in constrained environments, “better something than nothing” may still improve the UX. Here’s a basic example: async function inferCookieSupportFallback() { let hasCookieAPI = navigator.cookieEnabled; let canSetCookie = false; let hasStorageAccess = false; try { document.cookie = "testfallback=1; SameSite=None; Secure; Path=/;"; canSetCookie = document.cookie.includes("test_fallback=1"); document.cookie = "test_fallback=; Max-Age=0; Path=/;"; } catch (_) { canSetCookie = false; } if (typeof document.hasStorageAccess === "function") { try { hasStorageAccess = await document.hasStorageAccess(); } catch (_) {} } return { inferredThirdPartyCookies: hasCookieAPI && canSetCookie && hasStorageAccess, raw: { hasCookieAPI, canSetCookie, hasStorageAccess } }; } Example usage: inferCookieSupportFallback().then(({ inferredThirdPartyCookies }) => { if (inferredThirdPartyCookies) { console.log("Cookies likely supported. Likely, yes."); } else { console.warn("Cookies may be blocked or partitioned."); // You could inform the user or adjust behavior accordingly } }); Use this fallback when: You’re building a JavaScript-only widget embedded on unknown sites, You don’t control a second domain (or the team refuses to add one), or You just need some visibility into user-side behavior (e.g., debugging UX issues). Don’t rely on it for security-critical logic (e.g., auth gating)! But it may help tailor the user experience, surface warnings, or decide whether to attempt a fallback SSO flow. Again, it’s better to have something rather than nothing. Fallback Strategies When Third-Party Cookies Are Blocked Detecting blocked cookies is only half the battle. Once you know they’re unavailable, what can you do? Here are some practical options that might be useful for you: Redirect-Based Flows For auth-related flows, switch from embedded iframes to top-level redirects. Let the user authenticate directly on the identity provider's site, then redirect back. It works in all browsers, but the UX might be less seamless. Request Storage Access Prompt the user using requestStorageAccess() after a clear UI gesture (Safari requires this). Use this to re-enable cookies without leaving the page. Token-Based Communication Pass session info directly from parent to iframe via: postMessage (with required origin); Query params (e.g., signed JWT in iframe URL). This avoids reliance on cookies entirely but requires coordination between both sides: // Parent const iframe = document.getElementById('my-iframe'); iframe.onload = () => { const token = getAccessTokenSomehow(); // JWT or anything else iframe.contentWindow.postMessage( { type: 'AUTH_TOKEN', token }, 'https://iframe.example.com' // Set the correct origin! ); }; // iframe window.addEventListener('message', (event) => { if (event.origin !== 'https://parent.example.com') return; const { type, token } = event.data; if (type === 'AUTH_TOKEN') { validateAndUseToken(token); // process JWT, init session, etc } }); Partitioned Cookies (CHIPS) Chrome (since version 114) and other Chromium-based browsers now support cookies with the Partitioned attribute (known as CHIPS), allowing per-top-site cookie isolation. This is useful for widgets like chat or embedded forms where cross-site identity isn’t needed. Note: Firefox and Safari don’t support the Partitioned cookie attribute. Firefox enforces cookie partitioning by default using a different mechanism (Total Cookie Protection), while Safari blocks third-party cookies entirely. But be careful, as they are treated as “blocked” by basic detection. Refine your logic if needed. Final Thought: Transparency, Transition, And The Path Forward Third-party cookies are disappearing, albeit gradually and unevenly. Until the transition is complete, your job as a developer is to bridge the gap between technical limitations and real-world user experience. That means: Keep an eye on the standards.APIs like FedCM and Privacy Sandbox features (Topics, Attribution Reporting, Fenced Frames) are reshaping how we handle identity and analytics without relying on cross-site cookies. Combine detection with graceful fallback.Whether it’s offering a redirect flow, using requestStorageAccess(), or falling back to token-based messaging — every small UX improvement adds up. Inform your users.Users shouldn't be left wondering why something worked in one browser but silently broke in another. Don’t let them feel like they did something wrong — just help them move forward. A clear, friendly message can prevent this confusion. The good news? You don’t need a perfect solution today, just a resilient one. By detecting issues early and handling them thoughtfully, you protect both your users and your future architecture, one cookie-less browser at a time. And as seen with Chrome’s pivot away from automatic deprecation, the transition is not always linear. Industry feedback, regulatory oversight, and evolving technical realities continue to shape the time and the solutions. And don’t forget: having something is better than nothing.

Photo courtesy of the UK Department for Business and Trade In late April, the Regulatory Organizations of Architecture in Canada (ROAC) and the Architects Registration Board (ARB) in the U.K. signed a Mutual Recognition Agreement (MRA) for architects. The MRA, which officially came into effect on May 14th, is significant—it streamlines the process for architects in Canada to have their license recognised in the U.K. and vice versa. When you consider that architects in the U.K, and Canada already traded nearly $300m (£160m) in architectural services in 2023, it’s clear that the agreement is building from a position of strength. But going forward, the MRA will swing the doors wide open for architects to work internationally, fostering collaboration and building upon existing professional partnerships by acknowledging the validity of the education, experience, and examination requirements within each country, and allowing for simplified registration. From a qualitative perspective, there are many reasons the signing of the MRA is a positive step: It should allow for cross fertilization of materials into the supply chain, and awareness of a broader range of practices in adjacent fields that can better support how we define policy, code, design, procurement, and construction of projects. For instance, in the U.K., pre-zoning creates mostly as of right planning and development conditions that quietly allow for complex but successful public and private partnerships to take place. With increasing global challenges related to environmental and social sustainability, architecture as a profession needs to broaden access to different engagement techniques, material innovations, policy considerations, and construction methodologies by freely trading best practices and ideas between nations. This includes Canada’s emerging rediscovery of mass timber construction. We can learn more from each other in how we approach the critically important exercise of making architectural education more accessible to all socio-economic backgrounds. Bringing a greater diversity of lived experiences into the profession can further elevate the role and relevance of architecture in contributing to enhancing quality of life for everyone. The diversity of project teams and internal opportunities can create a sense of unity, which in turn can boost motivation, productivity, and improved design and project outcomes. Broader perspectives can lead to more bespoke solutions to design challenges; this isn’t just about creating better solution to problems but identifying issues sooner at all stages of the design process. It’s important to also recognize that the MRA signed between two countries also reveals the steps that we need to take here in Canada regarding interprovincial and interterritorial recognition. One could argue that it’s now easier for an Ontario-based architect to practice in London, Cardiff, or Edinburgh than in Saskatchewan. Above all, the MRA is a powerful symbolic gesture about the important role that architecture and design should have in an increasingly connected world. As a profession, we should be identifying and sharing solutions to the common challenges that we face, fostering a greater breadth of design approaches, as well as a broader range of technologies and innovation—no matter the geography. From an individual perspective, I am very familiar with several U.K. licensed architects now based here in Canada (including myself). The opportunity to be licensed formalizes the knowledge, competencies, and skills that these architects already have. It’s a marker of their commitment to Canada and represents an opportunity to maximize their contributions to the architectural community. Conversely, I have worked in Canada long enough to see the breadth of talent and cultural distinctions that makes Canadian architecture highly exportable—not just to the U.K., but to other geographies. The agreement could help enable Canadian firms of all sizes to push their talents to the limits where that might otherwise be constrained in the local context. It’s this latter point that I am most optimistic about. We should be inspired. We should be proud. We should be welcoming. And above all, we should be ambitious. There are few places on earth that have experienced the kind of pro-longed, consistent growth that Canadian cities have these past 25 years. Every year, 120,000 new people call the Toronto region alone home. There has been a cost, as the city has struggled to accommodate the rapid rate of change. A recent downturn in market conditions is foreshadowing an even greater supply crunch in the years to come. Now is the time that we should be importing ideas from around the world to help tackle the challenges we face here at home. At the same time, Canadian firms have well over two decades of experience dealing with growth at a tremendous scale. The idea of a housing crisis is not contained to North America. Cities around the world are facing similar conditions. There is a wealth of expertise in this country that is well earned, and worthy of exporting. It should not be lost on anyone that the Mutual Recognition Agreement between the UK and Canada has come as calls for protectionism are broadcast loudly. That we have, instead, chosen to embrace collaboration and open opportunity speaks to the moment and the need for even more of Canada on the world stage. British architect Ossie Airewele is a Senior Associate at BDP Quadrangle in Toronto. As a thought leader at BDP, he encourages meaningful conversations about the future of inclusive design, leads key residential and mixed-use projects, and is driving the North American expansion of the studio. The post Op Ed: The value of an architectural reciprocity agreement between Canada-UK appeared first on Canadian Architect.

Would you believe Google really wants to sell you on its AI? This week, we dive into the news from Google I/O 2025 with Engadget's Karissa Bell. We discuss how Gemini is headed to even more places, as well as Karissa's brief hands-on with Google's prototype XR glasses. It seems like Google is trying a bit harder now than it did with Google Glass and its defunct Daydream VR platform. But will the company end up giving up again, or does it really have a shot against Meta and Apple? Subscribe! iTunes Spotify Pocket Casts Stitcher Google Podcasts Topics Lots of AI and a little XR: Highlights from Google I/O 2025 – 1:15 OpenAI buys Jony Ive’s design company for $6.6B, in an all equity deal – 29:27 Fujifilm’s $850 X Half could be the perfect retro camera for the social media age – 39:42 Sesame Street is moving from HBO to Netflix – 44:09 Cuts to IMLS will lead to headaches accessing content on apps like Libby and Hoopla – 45:49 Listener Mail: Should I replace my Chromebook with a Mac or PC Laptop? – 48:33 Pop culture picks – 52:22 Credits  Hosts: Devindra Hardawar and Karissa BellProducer: Ben EllmanMusic: Dale North and Terrence O'Brien Transcript Devindra: [00:00:00] What's up, internet and welcome back to the Engadget Podcast. I'm Senior Editor Devindra Hardawar. I'm joined this morning by Senior Writer Karissa Bell. Hello, Karissa. Karissa: Hello. Good morning. Devindra: Good morning. And also podcast producer Ben Elman. Hey Ben, I'm muted my dang self. Hello. Hello, Ben. Good morning. It's been a busy week, like it's one of those weeks where. Three major conferences happened all at once and a varying like relevance to us. Google IO is the big one. We'll be talking about that with Karissa who was there and got to demo Google's XR glasses, but also Computex was happening. That's over in Taipei and we got a lot of news from that to, we'll mention some of those things. Also, Microsoft build happened and I feel like this was the less least relevant build to us ever. I got one bit of news I can mention there. That's pretty much it. It's been a crazy hectic week for us over at Eng Gadget. As always, if you're enjoying the show, please be free to subscribe to us on iTunes or your podcast catcher of choice. Leave us a review on iTunes, drop us email at podcast@enggadget.com. [00:01:00] Those emails, by the way, if you ask a good question, it could end up being part of our Ask Engadget section, so that's something we're starting out. I have another good one. I'll be throwing to asking Eng gadgets soon. So send us your emails podcast@enggadget.com, Google io. It's all about ai, isn't it? I feel like Karissa, we were watching the keynote for this thing and it felt like it went on and on of the thing about the things, like we all pretty much expect more about Gemini ai, more about their newer models a bit about xr. Can you give me, what's your overall impression of IO at this point? Karissa: Yeah, it's interesting because I've been covering IO long enough that I remember back when it used to be Android. And then there'd be like that little section at the end about, AI and some of the other stuff. And now it's completely reversed where it's entirely AI and basically no Android to the point where they had a whole separate event with their typical Android stuff the week before. So it didn't have to go through and talk about any of yeah, the mobile things. Devindra: That was just like a live stream that was just like a chill, live stream. No real [00:02:00] effort put into it. Whereas this is the whole show. They had a, who was it? But they had TOIs. TOIs, yeah. They had actual music which is something a lot of these folks do at keynotes. It's actually really disconcerting to see cool musicians taking the corporate gig and performing at one of these things. I think, it was like 20 13, 20 14, maybe the Intel one, IDF or something. But the weekend was there. Just trying to jam to all these nerds and it was sad, but yeah. How was the experience Karissa like actually going there? Karissa: Yeah, it was good. That keynote is always kind of a slog. Just, live blogging for our almost two hours straight, just constant is it's a lot. I did like the music. Towa was very chill. It was a nice way to start much. I preferred it over the crazy loop daddy set we got last year. If anyone remembers that. Devindra: Yeah. Ben: Yeah. Oh, I remember that. Mark Rub was at audio. That was so weird. Devindra: Yeah. Yeah, it was a little intense. Cool. So what are some of the highlights? Like there, there's a bunch of stuff. If you go look on, on the site on Engadget, we [00:03:00] have rounded up like all the major news and that includes a couple of things like hey, AI mode, chat bot coming to search. That's cool. We got more, I think the thing a lot of people were looking at was like Project Astra and where that's gonna be going. And that is the sort of universal AI assistant where you could hold your phone up and just ask it questions about the world. We got another demo video about that. Which again, the actual utility of it, I'm weirded out by. There was also one video where they were just like I'm gonna be dumb. I'm gonna pretend I'm very stupid and ask ask Astro, what is this tall building in front of me. And it was like a fire hydrant or something. It was like some piece of street thing. It was not a really well done demo. Do you have any thoughts about that, Krista? Does that seem more compelling to you now or is it the same as what we saw last year? Karissa: I think what was interesting to me about it was that we saw Astro last year and like that, I think there was a lot of excitement around that, but it wasn't really entirely clear where that. Project is going. They've said it's like an experimental research thing. And then, I feel like this year they really laid out that they want to [00:04:00] bring all that stuff to Gemini. Astra is sort of their place to like tinker with this and, get all this stuff working. But like their end game is putting this into Gemini. You can already see it a little bit in Gemini Live, which is like their multimodal feature where you can do some. Version of what ASRA can do. And so that was interesting. They're saying, we want Gemini to be this universal AI assistant. They didn't use the word a GI or anything like that. But I think it's pretty clear where they're going and like what their ambition is they want this to be, an all seeing, all knowing AI assistant that can help you with anything is what they're trying to sell it as. Devindra: It is weird, like we're watching the demo video and it's a guy trying to fix his bike and he is pointing his phone at like the bike and asking questions like which, which particular, I don't know. It's which particular nut do I need for this tightening thing and it's giving him good advice. It's pointing to things on YouTube. I. I don't know how useful this will actually be. This kind of goes to part of the stuff we're seeing with AI too, of just like offloading [00:05:00] some of the grunt work of human intelligence because you can do this right now, people have been YouTubing to fix things forever. YouTube has become this like information repository of just fix it stuff or home plumbing or whatever. And now it's just like you'll be able to talk to your phone. It'll direct you right to those videos or. Extract the actual instructions from those. That's cool. I feel like that's among the more useful things, more useful than like putting Gemini right into Chrome, which is another thing they're talking about, and I don't know how useful that is other than. They wanna push AI in front of us, just like Microsoft wants to push copilot in front of us at all times. Ben: What is a situation where you would have a question about your Chrome tabs? Like I'm not one of those people that has 15 chrome tabs open at any given time, and I know that I am. Yeah, I know. Wait, you're saying that like it's a high. Like it's high. Yeah, no I know. So I have a abnormally low number of chrome tabs open, but can you still come up [00:06:00] with an idea of why you would ask Gemini anything about your own tabs open? Hopefully you have them organized. At least Karissa: they should. A few examples of like online shopping, like maybe you have. Two tabs of two different products open. And you can say Devindra: exactly, Karissa: ask Gemini to like, compare the reviews. Or they use like the example of a recipe video, a recipe blog. And maybe, you wanna make some kind of modification, make the recipe gluten free. And you could ask Gemini Hey, make this how would I make this gluten free? But I think you're right, like it's not exactly clear. You can already just open a new tab and go to Gemini and ask it. Something. So they're just trying to reduce Devindra: friction. I think that's the main thing. Like just the less you have to think about it, the more it's in your face. You can just always always just jump right to it. It's hey, you can Google search from any your UL bar, your location bar in any browser. We've just grown to use that, but that didn't used to be the case. I remember there used to be a separate Google field. Some browsers and it wasn't always there in every browser too. They did announce some new models. We [00:07:00] saw there's Gemini 2.5 Pro. There's a deep think reasoning model. There's also a flash model that they announced for smaller devices. Did they show any good demos of the reasoning stuff? Because I that's essentially slower AI processing to hopefully get you better answers with fewer flaws. Did they actually show how that worked? Karissa. Karissa: I only saw what we all saw during the keynote and I think it's, we've seen a few other AI companies do something similar where you can see it think like its reasoning process. Yeah. And see it do that in real time. But I think it's a bit unclear exactly what that's gonna look like. Devindra: Watching a video, oh, Gemini can simulate nature simulate light. Simulate puzzles, term images into code. Ben: I feel like the big thing, yeah. A lot of this stuff is from DeepMind, right? This is DeepMind an alphabet company. Devindra: DeepMind and Alphabet company. There is Deep mind. This is deep Think and don't confuse this with deep seek, which is that the Chinese AI company, and they [00:08:00] clearly knew what they were doing when they call it that thing. Deep seek. But no, yeah, that is, this is partially stuff coming out of DeepMind. DeepMind, a company which Google has been like doing stuff with for a while. And we just have not really seen much out of it. So I guess Gemini and all their AI processes are a way to do that. We also saw something that got a lot of people, we saw Ben: Nobel Prize from them. Come on. Devindra: Hey, we did see that. What does that mean? What is that even worth anymore? That's an open question. They also showed off. A new video tool called Flow, which I think got a lot of people intrigued because it's using a new VO three model. So an updated version of what they've had for video effects for a while. And the results look good. Like the video looks higher quality. Humans look more realistic. There have been. The interesting thing about VO three is it can also do synchronized audio to actually produce audio and dialogue for people too. So people have been uploading videos around this stuff online at this point, and you have to [00:09:00] subscribe to the crazy high end. Version of Google's subscription to even test out this thing at this point that is the AI Ultra plan that costs $250 a month. But I saw something of yeah, here's a pretend tour of a make believe car show. And it was just people spouting random facts. So yeah, I like EVs. I would like an ev. And then it looks realistic. They sound synchronized like you could. I think this is a normal person. Then they just kinda start laughing at the end for no reason. Like weird little things. It's if you see a sociopath, try to pretend to be a human for a little bit. There's real Patrick Bateman vibes from a lot of those things, so I don't know. It's fun. It's cool. I think there's, so didn't we Ben: announce that they also had a tool to help you figure out whether or not a video was generated by flow? They did announce that Devindra: too. Ben: I've yeah, go ahead. Go Karissa: ahead. Yeah. The synth id, they've been working on that for a while. They talked about it last year at io. That's like their digital watermarking technology. And the funny thing about this is [00:10:00] their whole, the whole concept of AI watermarking is you put like these like invisible watermarks into AI generated content. You might, you couldn't just. See it, just watching this content. But you can go to this website now and basically like double check. If it has one of these watermarks, which is on one hand it's. I think it's important that they do this work, but I also just wonder how many people are gonna see a video and think I wonder what kind of AI is in this. Let me go to this other website and like double check it like that. Just, Ben: yeah. The people who are most likely to immediately believe it are the, also the least likely to go to the website and be like, I would like to double check Devindra: this. It doesn't matter because most people will not do it and the damage will be done. Just having super hyper realistic, AI video, they can, you can essentially make anything happen. It's funny that the big bad AI bad guy in the new Mission Impossible movies, the entity, one of the main things it does is oh, we don't know what's true anymore because the entity can just cr fabricate reality at whim. We're just doing that. [00:11:00] We're just doing that for, I don't know, for fun. I feel like this is a thing we should see in all AI video tools. This doesn't really answer the problem, answer the question that everyone's having though. It's what is the point of these tools? Because it does devalue filmmaking, it devalues people using actual actors or using, going out and actually shooting something. Did Google make a better pitch for why you would use Flow Karissa or how it would fit into like actual filmmaking? Karissa: I'm not sure they did. They showed that goofy Darren Aronofsky trailer for some woman who was trying to like, make a movie about her own birth, and it was like seemed like they was trying to be in the style of some sort of like psychological thriller, but it just, I don't know, it just felt really weird to me. I was I was just like, what are we watching? This doesn't, what are we watching? Yeah. Ben: Was there any like good backstory about why she was doing that either or was it just Hey, we're doing something really weird? Karissa: No, she was just oh I wonder, you know what? I wanna tell the story of my own birth and Okay. Ben: [00:12:00] Okay, but why is your relate birth more? Listen its like every, I need more details. Why is your birth more important? It's, everybody wants lots of babies. Write I memoir like one of three ways or something. Devindra: Yeah, it's about everybody who wants to write a memoir. It's kinda the same thing. Kinda that same naval ga thing. The project's just called ancestral. I'm gonna play a bit of a trailer here. I remember seeing this, it reminds me of that footage I dunno if you guys remember seeing, look who's talking for the very first time or something, or those movies where they, they showed a lot of things about how babies are made. And as a kid I was like, how'd they make that, how'd that get done? They're doing that now with AI video and ancestral this whole project. It is kinda sad because Aronofsky is one of my, like one of my favorite directors when he is on, he has made some of my favorite films, but also he's a guy who has admittedly stolen ideas and concepts from people like Satoshi kh as specific framing of scenes and things like that. In Requa for a Dream are in some cones movies as well. So [00:13:00] I guess it's to be expected, but it is. Sad because Hollywood as a whole, the union certainly do not like AI video. There was a story about James Earl Jones' voice being used as Darth Vader. In Fortnite. In Fortnite. In Fortnite, yeah. Which is something we knew was gonna happen because Disney licensed the rights to his voice before he died from his estate. He went in and recorded lines to at least create a better simulation of his voice. But people are going out there making that Darth Vader swear and say bad things in Fortnite and the WGA or is it sag? It's probably sag but sad. Like the unions are pissed off about this because they do not know this was happening ahead of time and they're worried about what this could mean for the future of AI talent. Flow looks interesting. I keep seeing play people play with it. I made a couple videos asked it to make Hey, show me three cats living in Brooklyn with a view of the Manhattan skyline or something. And it, it did that, but the apartment it rendered didn't look fully real. [00:14:00] It had like weird heating things all around. And also apparently. If you just subscribe to the basic plan to get access to flow, you can use flow, but that's using the VO two model. So older AI model. To get VO three again, you have to pay $250 a month. So maybe that'll come down in price eventually. But we shall see. The thing I really want to talk with you about Krisa is like, what the heck is happening with Android xr? And that is a weird project for them because I was writing up the news and they announced like a few things. They were like, Hey we have a new developer released to help you build Android XR apps. But it wasn't until the actual a IO show. That they showed off more of what they were actually thinking about. And you got to test out a pair of prototype Google XR glasses powered by Android xr. Can you tell me about that experience and just how does it differ from the other XR things you've seen from who is it from Several, look, you've seen Metas Meta, you saw one from Snap, right? Meta Karissa: I've seen Snap. Yeah. Yeah. I've seen the X reel. Yeah, some of the other smaller [00:15:00] companies I got to see at CES. Yeah, that was like a bit of a surprise. I know that they've been talking about Android XR for a while. I feel like it's been a little, more in the background. So they brought out these, these glasses and, the first thing that I noticed about them was like, they were actually pretty small and like normal looking compared to, met Orion or like the snap spectacles. Like these were very thin which was cool. But the display was only on one side. It was only on one lens. They called it like a monocular display. So there's one lens on one side. So it's basically just like a little window, very small field of view. Devindra: We could see it in, if you go to the picture on top of Chris's hands on piece, you can see the frame out. Of what that lens would be. Yeah. Karissa: Yeah. And I noticed even when we were watching that, that demo video that they did on stage, that like the field of view looked very small. It was even smaller than Snaps, which is 35 degrees like this. I would, if I had to guess, I'd say it's maybe like around 20. They wouldn't say what it was. They said, this is a prototype. We don't wanna say the way I thought about it, the way [00:16:00] I compared it to my piece was like the front screwing on a foldable phone, so it's you can get notifications and you can like glance at things, but it's not fully immersive ar it's not, surrounding your space and like really cha changing your reality, in the way that like snap and and meta are trying to do later when I was driving home, I realized it actually was reminded me like a better comparison might be the heads up display in your car. Speaker: Yeah. Yeah. Karissa: If you have a car that has that little hu where you can see how fast you're going and directions and stuff like that. Devindra: That's what Google Glass was doing too, right? Because that was a little thing off to the side of your revision that was never a full takeover. Your vision type of thing. Karissa: Yeah. It's funny, that's what our editor Aaron said when he was editing my piece, he was like, oh, this sounds like Google Glass. And I'm like, no, it actually, it's, it is better than that. These are like normal looking glasses. The, I tried Google Glass many years ago. Like the Fidelity was better. Actually I was thinking. It feels like a happy medium almost between, meta ray bands and like full ar Yeah, like I, I've had a meta ray band glasses [00:17:00] for a long time and people always ask me, like when I show it to someone, they're like, oh, that's so cool. And then they go, but you can see stuff, right? There's a display and I'm like. No. These are just, glasses with the speaker. And I feel like this might be like a good kind of InBetween thing because you have a little bit of display, but they still look like glasses. They're not bulky 'cause they're not trying to do too much. One thing I really liked is that when you take a photo, you actually get a little preview of that image that like floats onto the screen, which was really cool because it's hard to figure out how to frame pictures when you are taking using glasses camera on your smart glasses. So I think there's some interesting ideas, but it's very early. Obviously they want like Gemini to be a big part of it. The Gemini stuff. Was busted in my demo. Devindra: You also said they don't plan on selling these are like purely, hey, this is what could be a thing. But they're not selling these specific glasses, right? Karissa: Yeah, these specific ones are like, this is a research prototype. But they did also announce a partnership with Warby Parker and another glasses company. So I think it's like you can see them trying to take a meta approach here, which [00:18:00] actually would be pretty smart to say let's partner with. A known company that makes glasses, they're already popular. We can give them our, our tech expertise. They can make the glasses look good and, maybe we'll get something down the line. I actually heard a rumor that. Prototype was manufactured by Samsung. They wouldn't say Devindra: Of course it's Sam, Samsung wants to be all over this. Samsung is the one building their the full on Android XR headset, which is a sort of like vision Pro copycat, like it is Mohan. Yeah. Moan. It is displays with the pass through camera. That should be coming later this year. Go ahead Ben. Ben: Yeah. Question for Karissa. When Sergey brand was talking about Google Glass, did that happen before or after the big demo for the Google XR glasses? Karissa: That was after. That was at the end of the day. He was a surprise guest in this fireside chat with the DeepMind, CEO. And yeah, it was, we were all wondering about that. 'cause we all, dev probably remembers this very well the, when Google Glass came out and cereal and skydive [00:19:00] wearing them into io. Yeah. Speaker: Yep. Karissa: And then, now for him to come back and say we made a lot of mistakes with that product and. Ben: But was it mistakes or was it just the fact that like technology was not there yet because he was talking about like consumer electronic supply chain, blah, blah, blah, blah, blah. Devindra: He's right that the tech has caught up with what the vision of what they wanted to do, but also I think he fundamentally misread like people will see you looking like the goddamn borg and want to destroy you. They want you will turn into Captain Picard and be like, I must destroy whoever is wearing Google Glass because this looks like an alien trying to take over my civilization. And the thing that meta did right, that you've seen Karissa, is that make 'em look like normal glasses and Yeah, but nobody will knows, Ben: Karissa does not look entirely human in this picture either. Karissa: Yes. But listen from, if you see 'em straight on, they don't, they look transparent. That was I used that photo because I was trying to. Devindra: You get the angle, show The display. Karissa: Yeah. Devindra: [00:20:00] Yeah. There's another one like you. This looks normal. This looks totally normal. The glasses themselves look like, they look like typical hipster glasses. Like they're not like a super big frame around them. You're they look like the arms seem big. The arms seem wider than a typical pair of glasses, but you wouldn't know that 'cause it's covered in your hair. A lot of people won't notice glasses, arms as much. Ben: Yeah, Devindra: that is cool. The issue Ben: still is that all of these frames are so chunky. And it's because you need to hide all of the internals and everything, but you're not gonna get like the beautiful, like thin Japanese like titanium anytime soon. No, because this stuff needs to shrink way more. Devindra: This stuff that's not, those the kind of frames they are. I will say I had a meeting with the one of the I believe the CEO of X reel who. Came not, I did talk to them at c so they, they had like a lot of ideas about that. I talked to the the head of space top, which is [00:21:00] the, that's the company that was doing the sort of AR laptop thing. And then they gave up on that idea because AI PCs have the nmps that they need to do that stuff. And they're all in on the idea that, more people will want to use these sorts of glasses. Maybe not all the time, but for specific use cases. Something that co covers your field of vision more. Could be a great thing when you sit down at your desk. I could see people doing this. I could see people getting these glasses. I don't know if it's gonna be good for society, right? It feels when Bluetooth headsets were first popping up and everybody hated those people, and you're like, oh, we must shun this person from society. This one, you can't quite see the screen. So you can pretend to be a normal human and then have this like augmented ability next to you. If they can hide that, if they can actually hide the fact that you have a display on your glasses that would help people like me who are face blind and I walk around I don't, I know this person. I've seen them before. What is their name? What is their name? I could see that being useful. Ben: On the other side of it [00:22:00] though, if you have one standard look for glasses like this, then you know, oh, this person is, I. Also interacting with like information and stuff that's like popping up in front of their eyes. It's a universal signifier, just like having a big pair of headphones is Devindra: I think you will see people looking off to the distance. Krisa, did you notice that your eye line was moving away from people you were talking to while you were wearing these? Karissa: Yeah, and that was also one of the issues that I had was that the. Actual, like display was like, was it like didn't quite render right? Where I'm not a farsighted person, but I actually had to look farther off in the distance to actually get it to like my eyes to focus on it. And I asked 'em about that and they're like, oh it's a prototype. It's not quite dialed in. They weren't calibrating these things to your eyeballs. Like the way when I did the Meta Orion demo, they have to take these specific measurements because there's eye tracking and all these things and this, didn't have any of that. There. Yeah, there definitely was. You're, somebody's talking to you, but you're looking over here. Devindra: That's not great. That's [00:23:00] not great for society. You're having a conversation with people. I like how they're framing this oh yes, you can be more connected with reality. 'cause you don't have a phone in front of your face, except you always have another display in front of your face, which nobody else can see, and you're gonna look like an alien walking around. They showed some videos of people using it for like street navigation. Which I kinda like. You're in a new city, you'll see the arrows and where to turn and stuff. That's useful. But there is this, there was one that was really overwrought. It was a couple dancing at Sunset, and the guy is take a picture of this beautiful moment of the sun peeking through behind, my lady friend. And it just felt like that's what you wanna do in that moment. You wanna talk to your virtual assistant while you should be enjoying the fact that you are having this beautiful dancing evening, which nobody will ever actually have. So that's the whole thing. I will say my overall thoughts on this stuff, like just looking at this, the stuff they showed before they actually showed us the glasses, it doesn't feel like Google is actually that far in terms of making this a reality. Karissa the, like I'm comparing it to. Where Meta [00:24:00] is right now, and even where Apple is right now, like when Apple showed us the vision Pro. We were able to sit down and I had a 30 minute demo of that thing working, and I saw the vision of what they were doing and they thought a lot about how this was. How long was your demo with this thing? Karissa: I was in the room with them for about five minutes and I had them on for about three minutes myself. That's not a demo. That's not a demo. Ben: Oh, goodness. So all of these pictures were taken in the same 90 seconds? Yes. Yeah. God. That's amazing. Devindra: It's amazing you were able to capture these impressions, Karissa. Yeah, Karissa: I will say that they did apparently have a demo in December, a press event in December where people got to see these things for a lot longer, but it was, they could not shoot them at all. We, a lot of us were wondering if that was why it was so constrained. They only had one room, there's hundreds of people basically lining up to try these out. And they're like very strict. You got five minutes, somebody's in there like after a couple minutes, rushing you out, and we're like, okay. Like Devindra: They clearly only have a handful of these. That's like the main reason this is happening. I am, this is the company, that did Google Glass and that was too [00:25:00] early and also maybe too ambitious. But also don't forget, Google Cardboard, which was this that was a fun little project of getting phone-based vr happening. Daydream vr, which was their self-contained headset, which was cool. That was when Samsung was doing the thing with Meta as well, or with Oculus at the time. So and they gave up on those things. Completely. And Google's not a company I trust with consumer Hardaware in general. So I am. Don't think there is a huge future in Android xr, but they wanna be there. They wanna be where Meta is and where Apple is and we shall see. Anything else you wanna add about io, Karissa? Karissa: No, just that AI. A i a ai Devindra: a I didn't AI ao, A IAO a IO starline. The thing that was a, like weird 3D rendering teleconferencing video that is becoming a real thing that's turning to Google Beam video. But it's gonna be an enterprise thing. They're teaming up with AI to, with HP to bring a scaled down version of that two businesses. I don't think we'll love or see That's one of those things where it's oh, this exists [00:26:00] in some corporate offices who will pay $50,000 for this thing, but. I don't, normal people will never interact with this thing, so it practically just does not exist. So we shall see. Anyway, stay tuned for, we're gonna have more demos of the Gemini stuff. We'll be looking at the new models, and certainly Chris and I will be looking hard at Android XR and wherever the heck that's going. Let's quickly move on to other news. And I just wanna say there were other events, Compex, we wrote up a couple, a whole bunch of laptops. A MD announced a cheaper radio on graphics card. Go check out our stories on that stuff. Build. I wrote one, I got a 70 page book of news from Microsoft about build and 99% of that news just does not apply to us because Build is so fully a developer coding conference. Hey, there's more more copilot stuff. There's a copilot app coming to 360 [00:27:00] fi subscribers, and that's cool, but not super interesting. I would say the big thing that happened this week and that surprised a lot of us is the news that OpenAI has bought. Johnny i's design startup for six and a half billion. Dollars. This is a wild story, which is also paired with a weird picture. It looks like they're getting married. It looks like they're announcing their engagement over here because Johnny, ive is just leaning into him. Their heads are touching a little bit. It's so adorable. You're not showing Ben: the full website though. The full website has like a script font. It literally looks, yeah, like something from the knot. Devindra: It Is it? Yeah. Let's look at here. Sam and Johnny introduced io. This is an extraordinary moment. Computers are now seeing, thinking, understanding, please come to our ceremony at this coffee shop. For some reason, they also yeah, so they produced this coffee shop video to really show this thing off and, it is wild to me. Let me pull this up over here. Ben: While we're doing that. Karissa, what do you [00:28:00] have to say about this? Karissa: I don't, I'm trying to remember, so I know this is Johnny Ives like AI because he also has like the love from, which is still Devindra: this is love from, this is, so he is, let me get the specifics of the deal out here. Yeah. As part of the deal Ive and his design studio love form. Is it love form or love form? Love form. Yeah. Love form are gonna be joining are gonna work independently of open ai. But Scott Cannon Evans Hanky and Ang Tan who co-founded io. This is another io. I hate these. Yeah, so IO is his AI. Karissa: Focused design thing. And then love form is like his design Devindra: studio thing. Karissa: Sure. Yeah. I'm just, he Devindra: has two design things. Karissa: I'm trying to remember what they've done. I remember there was like a story about they made like a really expensive jacket with some weird buttons or something like Devindra: Yep. I do remember that. Karissa: I was just trying to back my brain of what Johnny Iiv has really done in his post Apple life. I feel like we haven't, he's made Devindra: billions of dollars courses. What's happened? Yes. [00:29:00] Because he is now still an independent man. Clearly he's an independent contractor, but love like the other side of io. Which includes those folks. They will become open AI employees alongside 50 other engineers, designers, and researchers. They're gonna be working on AI Hardaware. It seems like Johnny, I will come in with like ideas, but he, this is not quite a marriage. He's not quite committing. He's just taking the money and being like, Ew, you can have part of my AI startup for six and a half billion dollars. Ben: Let us know your taxes. It's all equity though, so this is all paper money. Six and a half billion dollars. Of like open AI's like crazy, their crazy valuation who knows how act, how much it's actually going to be worth. But all these people are going to sell a huge chunk of stock as soon as open AI goes public anyway. So it's still gonna be an enormous amount of money. Devindra: Lemme, let me see here, the latest thing. Open OpenAI has raised 57.9 billion of funding over 11 rounds. [00:30:00] Good Lord. Yeah. Yeah. So anyway, a big chunk of that is going to, to this thing because I think what happened is that Sam Altman wants to, he clearly just wants to be Steve Jobs. I think that's what's happening here. And go, I, all of you go look at the video, the announcement video for this thing, because it is one of the weirdest things I've seen. It is. Johnny I have walking through San Francisco, Sam Altman, walking through San Francisco with his hands in his pockets. There's a whole lot of setup to these guys meeting in a coffee shop, and then they sit there at the coffee shop like normal human beings, and then have an announcement video talking to nobody. They're just talking to the middle of the coffee bar. I don't know who they're addressing. Sometimes they refer to each other and sometimes they refer to camera, but they're never looking at the camera. This is just a really wild thing. Also. Yet, another thing that makes me believe, I don't think Sam Altman is is a real human boy. I think there is actually something robotic about this man, because I can't see him actually perform in real life [00:31:00] what they're gonna do. They reference vagaries, that's all. It's, we don't know what exactly is happening. There is a quote. From Johnny Ive, and he says, quote, the responsibility that Sam shares is honestly beyond my comprehension end quote. Responsibility of what? Just building this like giant AI thing. Sam Alman For humanity. Yeah, for humanity. Like just unlocking expertise everywhere. Sam Altman says he is. He has some sort of AI device and it's changed his life. We don't know what it is. We dunno what they're actually working on. They announced nothing here. But Johnny Ive is very happy because he has just made billions of dollars. He's not getting all of that money, but he, I think he's very pleased with this arrangement. And Sam Malman seems pleased that, oh, the guy who who designed the iPhone and the MacBook can now work for me. And Johnny, I also says the work here at Open AI is the best work he's ever done. Sure. You'd say that. Sure. By the way. Karissa: Sure. What do you think Apple thinks about all this? Devindra: Yeah, Karissa: their AI [00:32:00] program is flailing and like their, star designer who, granted is not, separated from Apple a while ago, but is now teaming up with Sam Altman for some future computing AI Hardaware where like they can't even get AI Siri to work. That must be like a gut punch for folks maybe on the other side of it though. Yeah, I Ben: don't think it's sour grapes to say. Are they going into the like. Friend, like friend isn't even out yet, but like the humane pin? Yes. Or any of the other like AI sidekick sort of things like that has already crashed and burned spectacularly twice. Devindra: I think Apple is, maybe have dodged a bullet here because I, the only reason Johnny and I just working on this thing is because he OpenAI had put some money into left Formm or IO years ago too. So they already had some sort of collaboration and he's just okay, people are interested in the ai. What sort of like beautiful AI device can I buy? The thing is. [00:33:00] Johnny Ive unchecked as a designer, leads to maddening things like the magic mouse, the charges from the bottom butterfly Karissa: keyboard, Devindra: any butterfly keyboard. Yeah, that's beautiful, but not exactly functional. I've always worked best when he Johnny, ive always worked best when I. He had the opposing force of somebody like a Steve Jobs who could be like, no, this idea is crazy. Or reign it in or be more functional. Steve Jobs not a great dude in many respects, but the very least, like he was able to hone into product ideas and think about how humans use products a lot. I don't think Johnny, ive on his own can do that. I don't think Sam Altman can do that because this man can barely sit and have a cup of coffee together. Like a human being. So I, whatever this is. I honestly, Chris, I feel like Apple has dodged a bullet because this is jumping into the AI gadget trend. Apple just needs to get the software right, because they have the devices, right? We are wearing, we're wearing Apple watches. People have iPhones, people have MacBooks. What they need to do, solidify the infrastructure the AI [00:34:00] smarts between all those devices. They don't need to go out and sell a whole new device. This just feels like opening AI is a new company and they can try to make an AI device a thing. I don't think it's super compelling, but let us know listeners, if any of this, listen to this chat of them talking about nothing. Unlocking human greatness, unlocking expertise just through ai, through some AI gadget. I don't quite buy it. I think it's kind of garbage, but yeah. Ben: Anything else you guys wanna say about this? This is coming from the same guy who, when he was asked in an interview what college students should study, he said Resilience. Karissa: Yeah. I just think all these companies want. To make the thing that's the next iPhone. Yes. They can all just stop being relying on Apple. It's the thing that Mark Zuckerberg has with all of their like Hardaware projects, which by the way, there was one of the stories said that Johnny I thing has been maybe working on some kind of. Head earbuds with cameras on them, which sounded [00:35:00] very similar to a thing that meta has been rumored about meta for a long time. And and also Apple, Devindra: like there, there were rumors about AirPods with head with Karissa: cameras. Yeah. And everyone's just I think trying to like, make the thing that's like not an iPhone that will replace our iPhones, but good luck to them, good, good Devindra: luck to that because I think that is coming from a fundamentally broken, like it's a broken purpose. The whole reason doing that is just try to outdo the iPhone. I was thinking about this, how many companies like Apple that was printing money with iPods would just be like, Hey we actually have a new thing and this will entirely kill our iPod business. This new thing will destroy the existing business that is working so well for us. Not many companies do that. That's the innovator's dilemma that comes back and bites companies in the butt. That's why Sony held off so long on jumping into flat screen TVs because they were the world's leader in CRTs, in Trinitron, and they're like, we're good. We're good into the nineties. And then they completely lost the TV business. That's why Toyota was so slow to EVs, because they're like, hybrids are good to us. Hybrids are great. We don't need an EV for a very long time. And then they released an EV that [00:36:00] we, where the wheels fell off. So it comes for everybody. I dunno. I don't believe in these devices. Let's talk about something that could be cool. Something that is a little unrealistic, I think, but, for a certain aesthetic it is cool. Fujifilm announced the X half. Today it is an $850 digital camera with an analog film aesthetic. It shoots in a three by four portrait aspect ratio. That's Inax mini ratio. It looks like an old school Fuji camera. This thing is pretty wild because the screen it's only making those portrait videos. One of the key selling points is that it can replicate some film some things you get from film there's a light leak simulation for when you like Overexpose film A little bit, a ation, and that's something Ben: that Fujifilm is known for. Devindra: Yes. They love that. They love these simulation modes. This is such a social media kid camera, especially for the people who cannot afford the $2,000 Fuji films, compact cameras. [00:37:00] Wow. Even the Ben: screen is do you wanna take some vertical photographs for your social media? Because vertical video has completely won. Devindra: You can't, and it can take video, but it is just, it is a simplistic living little device. It has that, what do you call that? It's that latch that you hit to wind film. It has that, so you can put it into a film photograph mode where you don't see anything on the screen. You have to use the viewfinder. To take pictures and it starts a countdown. You could tell it to do like a film, real number of pictures, and you have to click through to hit, take your next picture. It's the winder, it's, you can wind to the next picture. You can combine two portrait photos together. It's really cool. It's really cute. It's really unrealistic I think for a lot of folks, but. Hey, social media kits like influencers, the people who love to shoot stuff for social media and vertical video. This could be a really cool little device. I don't, what do you guys think about this? Karissa: You know what this reminds me of? Do you remember like in the early Instagram days when there was all these [00:38:00] apps, like hip, systematic where they tried to emulate like film aesthetics? And some of them would do these same things where like you would take the picture but you couldn't see it right away. 'cause it had to develop. And they even had a light leak thing. And I'm like, now we've come full circle where the camera companies are basically like yeah. Taking or like just doing their own. Spin on that, but Devindra: it only took them 15 years to really jump on this trend. But yes, everybody was trying to emulate classic cameras and foodie was like, oh, you want things that cost more but do less. Got it. That's the foodie film X half. And I think this thing will be a huge success. What you're talking about krisa, there is a mode where it's just yeah. You won't see the picture immediately. It has to develop in our app and then you will see it eventually. That's cool honestly, like I love this. I would not, I love it. I would not want it to be my main camera, but I would love to have something like this to play around when you could just be a little creative and pretend to be a street photographer for a little bit. Oh man. This would be huge in Brooklyn. I can just, Ben: Tom Rogers says cute, but stupid tech. I think that's [00:39:00] the perfect summary. Devindra: But this is, and I would say this compared to the AI thing, which is just like. What is this device? What are you gonna do with it? It feels like a lot of nothing in bakery. Whereas this is a thing you hold, it takes cool pictures and you share it with your friends. It is such a precise thing, even though it's very expensive for what it is. I would say if you're intrigued by this, you can get cheap compact cameras, get used cameras. I only ever buy refurbished cameras. You don't necessarily need this, but, oh man, very, but having a Karissa: Fuji film camera is a status symbol anyway. So I don't know. This is it's eight 50 still seems like a little steep for a little toy camera, basically. But also I'm like I see that. I'm like, Ooh, that looks nice. Devindra: Yeah. It's funny the power shots that kids are into now from like the two thousands those used to cost like 200 to 300 bucks and I thought, oh, that is a big investment in camera. Then I stepped up to the Sony murals, which were like 500 to 600 or so. I'm like, okay, this is a bigger step up than even that. Most people would be better off with a [00:40:00] muralist, but also those things are bigger than this tiny little pocket camera. I dunno. I'm really I think it's, I'm enamored with this whole thing. Also briefly in other news we saw that apparently Netflix is the one that is jumping out to save Sesame Street and it's going to, Sesame Street will air on Netflix and PBS simultaneously. That's a good, that's a good thing because there was previously a delay when HBO was in charge. Oh really? Yeah. They would get the new episodes and there was like, I forget how long the delay actually was, but it would be a while before new stuff hit PBS. This is just Hey, I don't love that so much of our entertainment and pop culture it, we are now relying on streamers for everything and the big media companies are just disappointing us, but. This is a good move. I think Sesame Street should stick around, especially with federal funding being killed left and right for public media like this. This is a good thing. Sesame Street is still good. My kids love it. When my son starts leaning into like his Blippy era, I. I just [00:41:00] kinda slowly tune that out. Here's some Sesame Street. I got him into PeeWee's Playhouse, which is the original Blippy. I'm like, yes, let's go back to the source. Because Peewee was a good dude. He's really, and that show still holds up. That show is so much fun. Like a great introduction to camp for kids. Great. In introduction to like also. Diverse neighborhoods, just Sesame Street as well. Peewee was, or mr. Rogers was doing Ben: it before. I think everyone, Devindra: Mr. Rogers was doing it really well too. But Peewee was always something special because PeeWee's Wild, Peewee, Lawrence Fishburn was on Peewee. There, there's just a lot of cool stuff happening there. Looking back at it now as an adult, it is a strange thing. To watch, but anyway, great to hear that Sesame Street is back. Another thing, not so quick. Ben: Yeah, let me do this one. Go ahead, if I may. Go ahead. So if you have any trouble getting audio books on Libby or Hoopla or any of the other interlibrary loan systems that you can like access on your phone or iPad any tablet. That's [00:42:00] because of the US government because a while ago the Trump administration passed yet another executive order saying that they wanted to cut a bunch of funding to the Institute of Museum and Library Services, the IMLS, and they're the ones who help circulate big quotation marks there just because it's digital files, all of these things from interlibrary loans. So you can, get your audio books that you want. The crazy thing about this is that the IMLS was created in 1996 by a Republican controlled Congress. What's the deal here, guys? There's no waste, fraud and abuse, but if you have problems getting audio books, you can tell a friend or if anybody's complaining about why their, library selection went down. By a lot on Libby recently, now you have the answer. Devindra: It is truly sad. A lot of what's happening is just to reduce access to information because hey, a well-formed population is [00:43:00] dangerous to anybody in charge, right? Terrible news. Let's move on to stuff from that's happening around in gadget. I wanna quickly shout out that Sam Rutherford has reviewed the ACEs RG flow Z 13. This is the sort of like surface like device. That's cool. This is the rise in pro Max chip. Sam seems to like it, so that's, it's a cool thing. Not exactly stealthy. He gave it a 79, which is right below. The threshold we have for recommending new products because this thing is expensive. You're paying a lot of money to get, essentially get a gaming tablet. But I tested out cs. It is cool that it actually worked for a certain type of person with too much money and who just needs the lightest gaming thing possible. I could see it being compelling. Let's see, what is the starting price? $2,100. $2,100 for a gaming tablet. Sam says it costs the same or more as a comparable RRG Zes G 14 with a real RTX 50 70. That is a great laptop. The RRGs Zes G 14, we have praised that laptop so much. So this is not [00:44:00] really meant for anybody ACEs lifts to do these experiments. They're getting there, they're getting there in terms of creating a gaming tablet, but not quite something I'd recommend for everybody at this point. All right. We have a quick email from a listener too. Thank you for sending this in, Jake Thompson. If you wanna send us an email, e podcast in gadget.com, and again, your emails may head into our Asking Gadget section. Jake asks. He's a real estate agent in need of a new laptop. He uses a Chromebook right now and it meets every need he has. Everything they do is web-based, but should they consider alternatives to a premium com Chromebook for their next computer, he says he doesn't mind spending $750 or more if he can get something lightweight, trustworthy with a solid battery life. What would we consider in the search? I would point to, I immediately point to Jake, to our laptop guides because literally everything we mention, the MacBook Air. The Asis [00:45:00] Zen book, S 14, even the Dell Xbs 13 would be not much more than that price. I think more useful than a premium Chromebook because I think the idea of a premium Chromebook is a, is insanity. I don't know why you're spending so much money for a thing that can only do web apps, cheap Chromebooks, mid-range Chromebooks fine, $500 or less. Great. But if you're spending that much money and you want something that's more reliable, that you could do more with, even if everything you're doing is web-based, there may be other things you wanna do. MacBook Windows laptop. There is so much more you can unlock there. Little bit, a little bit of gaming, a little bit of media creation. I don't know, Karissa. Ben, do you have any thoughts on this? What would you recommend or do, would you guys be fine with the Chromebook? Karissa: I like Chromebooks. I thought my first thought, and maybe this is like too out there, but would an iPad Pro fit that fit those requirements? 'cause you can do a lot with an iPad Pro. You Devindra: can do a lot that's actually great battery, Karissa: lightweight, lots of apps. If most everything he's doing is web based, there's. You can probably use iPad apps. Devindra: That's actually a good point. Karissa you can [00:46:00] do a lot with an iPad and iPad Pro does start at around this price too. So it would be much lighter and thinner than a laptop. Especially if you could do a lot of web stuff. I feel like there are some web things that don't always run well in an iPad form. Safari and iPad doesn't support like everything you'd expect from a web-based site. Like I think if you. There are things we use like we use Video Ninja to record podcasts and that's using web RTC. Sometimes there are things like zencaster, something you have to use, apps to go use those things because I, iOS, iPad OS is so locked down. Multitasking isn't great on iPad os. But yeah, if you're not actually doing that much and you just want a nice. Media device. An iPad is a good option too. Alright, thank you so much Jake Thompson. That's a good one too because I wanna hear about people moving on from Chromebooks. 'cause they, send us more emails at podcast@enggadget.com for sure. Let's just skip right past what we're working on 'cause we're all busy. We're all busy with stuff unless you wanna mention anything. Chris, anything you're working on at the moment? Karissa: The only thing I wanna flag is that [00:47:00] we are rapidly approaching another TikTok sale or ban. Deadline Yes. Next month. Speaker: Sure. Karissa: Been a while since we heard anything about that, but, I'm sure they're hard at work on trying to hammer out this deal. Ben: Okay. But that's actually more relevant because they just figured out maybe the tariff situation and the tariff was the thing that spoiled the first deal. So we'll see what happens like at the beginning of July, yeah. I think Karissa: The deadline's the 19th of June Ben: oh, at the beginning of June. Sorry. Karissa: Yeah, so it's. It's pretty close. And yeah, there has been not much that I've heard on that front. So Devindra: this is where we are. We're just like walking to one broken negotiation after another for the next couple years. Anything you wanna mention, pop culture related krisa that is taking your mind off of our broken world. Karissa: So this is a weird one, but I have been, my husband loves Stargate, and we have been for years through, wait, the movie, the TV shows, Stargate [00:48:00] SG one. Oh Devindra: God. And I'm yeah. Just on the Karissa: last few episodes now in the end game portion of that show. So that has been I spent years like making fun of this and like making fun of him for watching it, but that show's Devindra: ridiculously bad, but yeah. Yeah. Karissa: Everything is so bad now that it's, actually just a nice. Yeah. Distraction to just watch something like so silly. Devindra: That's heartwarming actually, because it is a throwback to when things were simpler. You could just make dumb TV shows and they would last for 24 episodes per season. My for how Ben: many seasons too, Devindra: Karissa? Karissa: 10 seasons. Devindra: You just go on forever. Yeah. My local or lamb and rice place, my local place that does essentially New York streetcar style food, they placed Arga SG one. Every time I'm in there and I'm sitting there watching, I was like, how did we survive with this? How did we watch this show? It's because we just didn't have that much. We were desperate for for genre of fiction, but okay, that's heartwarming Krisa. Have you guys done Farscape? No. Have you seen Farscape? 'cause Farscape is very, is a very similar type of [00:49:00] show, but it has Jim Henson puppets and it has better writing. I love Jim Henson. It's very cool. Okay. It's it's also, it's unlike Stargate. It also dares to be like I don't know, sexy and violent too. Stargate always felt too campy to me. But Farscape was great. I bought that for $15. On iTunes, so that was a deal. I dunno if that deal is still there, but the entire series plus the the post series stuff is all out there. Shout out to Farscape. Shout out to Stargate SG one Simpler times. I'll just really briefly run down a few things and or season two finished over the last week. Incredible stuff. As I said in my initial review, it is really cool to people see people watching this thing and just being blown away by it. And I will say the show. Brought me to tears at the end, and I did not expect that. I did not expect that because we know this guy's gonna die. This is, we know his fate and yet it still means so much and it's so well written and the show is a phenomenon. Chris, I'd recommend it to you when you guys are recovering from Stargate SG one loss and or is fantastic. I also checked out a bit of murderbot the [00:50:00] Apple TV plus adaptation of the Martha Wells books. It's fine. It is weirdly I would say it is funny and entertaining because Alexander Skarsgard is a fun person to watch in in genre fiction. But it also feels like this could be funnier, this could be better produced. Like you could be doing more with this material and it feels like just lazy at times too. But it's a fine distraction if you are into like half-baked sci-fi. So I don't know. Another recommendation for Stargate SG one Levers, Karissa Final Destination Bloodlines. I reviewed over at the film Cast and I love this franchise. It is so cool to see it coming back after 15 years. This movie is incredible. Like this movie is great. If you understand the final destination formula, it's even better because it plays with your expectations of the franchise. I love a horror franchise where there's no, no definable villain. You're just trying to escape death. There's some great setups here. This is a great time at the movies. Get your popcorn. Just go enjoy the wonderfully creative kills. [00:51:00] And shout out to the Zap lapovsky and Adam B. Stein who. Apparently we're listening to my other podcast, and now we're making good movies. So that's always fun thing to see Mount Destination Bloodlines a much better film. The Mission Impossible, the Final Reckoning. My review of that is on the website now too. You can read that in a gadget. Ben: Thanks everybody for listening. Our theme music is by Game Composer Dale North. Our outro music is by our former managing editor, Terrence O'Brien. The podcast is produced by me. Ben Elman. You can find Karissa online at Karissa: Karissa b on threads Blue Sky, and sometimes still X. Ben: Unfortunately, you can find Dendra online Devindra: At dendra on Blue Sky and also podcast about movies and TV at the film cast@thefilmcast.com. Ben: If you really want to, you can find me. At hey bellman on Blue Sky. Email us at podcast@enggadget.com. Leave us a review on iTunes and subscribe on anything that gets podcasts. That includes [00:52:00] Spotify. This article originally appeared on Engadget at https://www.engadget.com/mobile/smartphones/engadget-podcast-the-ai-and-xr-of-google-io-2025-131552868.html?src=rss

Artificial intelligence is a deep and convoluted world. The scientists who work in this field often rely on jargon and lingo to explain what they’re working on. As a result, we frequently have to use those technical terms in our coverage of the artificial intelligence industry. That’s why we thought it would be helpful to put together a glossary with definitions of some of the most important words and phrases that we use in our articles. We will regularly update this glossary to add new entries as researchers continually uncover novel methods to push the frontier of artificial intelligence while identifying emerging safety risks. AGI Artificial general intelligence, or AGI, is a nebulous term. But it generally refers to AI that’s more capable than the average human at many, if not most, tasks. OpenAI CEO Sam Altman recently described AGI as the “equivalent of a median human that you could hire as a co-worker.” Meanwhile, OpenAI’s charter defines AGI as “highly autonomous systems that outperform humans at most economically valuable work.” Google DeepMind’s understanding differs slightly from these two definitions; the lab views AGI as “AI that’s at least as capable as humans at most cognitive tasks.” Confused? Not to worry — so are experts at the forefront of AI research. AI agent An AI agent refers to a tool that uses AI technologies to perform a series of tasks on your behalf — beyond what a more basic AI chatbot could do — such as filing expenses, booking tickets or a table at a restaurant, or even writing and maintaining code. However, as we’ve explained before, there are lots of moving pieces in this emergent space, so “AI agent” might mean different things to different people. Infrastructure is also still being built out to deliver on its envisaged capabilities. But the basic concept implies an autonomous system that may draw on multiple AI systems to carry out multistep tasks. Chain of thought Given a simple question, a human brain can answer without even thinking too much about it — things like “which animal is taller, a giraffe or a cat?” But in many cases, you often need a pen and paper to come up with the right answer because there are intermediary steps. For instance, if a farmer has chickens and cows, and together they have 40 heads and 120 legs, you might need to write down a simple equation to come up with the answer (20 chickens and 20 cows). In an AI context, chain-of-thought reasoning for large language models means breaking down a problem into smaller, intermediate steps to improve the quality of the end result. It usually takes longer to get an answer, but the answer is more likely to be correct, especially in a logic or coding context. Reasoning models are developed from traditional large language models and optimized for chain-of-thought thinking thanks to reinforcement learning. (See: Large language model) Techcrunch event Join us at TechCrunch Sessions: AI Secure your spot for our leading AI industry event with speakers from OpenAI, Anthropic, and Cohere. For a limited time, tickets are just $292 for an entire day of expert talks, workshops, and potent networking. Exhibit at TechCrunch Sessions: AI Secure your spot at TC Sessions: AI and show 1,200+ decision-makers what you’ve built — without the big spend. Available through May 9 or while tables last. Berkeley, CA | June 5 REGISTER NOW Deep learning A subset of self-improving machine learning in which AI algorithms are designed with a multi-layered, artificial neural network (ANN) structure. This allows them to make more complex correlations compared to simpler machine learning-based systems, such as linear models or decision trees. The structure of deep learning algorithms draws inspiration from the interconnected pathways of neurons in the human brain. Deep learning AI models are able to identify important characteristics in data themselves, rather than requiring human engineers to define these features. The structure also supports algorithms that can learn from errors and, through a process of repetition and adjustment, improve their own outputs. However, deep learning systems require a lot of data points to yield good results (millions or more). They also typically take longer to train compared to simpler machine learning algorithms — so development costs tend to be higher. (See: Neural network) Diffusion Diffusion is the tech at the heart of many art-, music-, and text-generating AI models. Inspired by physics, diffusion systems slowly “destroy” the structure of data — e.g. photos, songs, and so on — by adding noise until there’s nothing left. In physics, diffusion is spontaneous and irreversible — sugar diffused in coffee can’t be restored to cube form. But diffusion systems in AI aim to learn a sort of “reverse diffusion” process to restore the destroyed data, gaining the ability to recover the data from noise. Distillation Distillation is a technique used to extract knowledge from a large AI model with a ‘teacher-student’ model. Developers send requests to a teacher model and record the outputs. Answers are sometimes compared with a dataset to see how accurate they are. These outputs are then used to train the student model, which is trained to approximate the teacher’s behavior. Distillation can be used to create a smaller, more efficient model based on a larger model with a minimal distillation loss. This is likely how OpenAI developed GPT-4 Turbo, a faster version of GPT-4. While all AI companies use distillation internally, it may have also been used by some AI companies to catch up with frontier models. Distillation from a competitor usually violates the terms of service of AI API and chat assistants. Fine-tuning This refers to the further training of an AI model to optimize performance for a more specific task or area than was previously a focal point of its training — typically by feeding in new, specialized (i.e., task-oriented) data.  Many AI startups are taking large language models as a starting point to build a commercial product but are vying to amp up utility for a target sector or task by supplementing earlier training cycles with fine-tuning based on their own domain-specific knowledge and expertise. (See: Large language model [LLM]) GAN A GAN, or Generative Adversarial Network, is a type of machine learning framework that underpins some important developments in generative AI when it comes to producing realistic data – including (but not only) deepfake tools. GANs involve the use of a pair of neural networks, one of which draws on its training data to generate an output that is passed to the other model to evaluate. This second, discriminator model thus plays the role of a classifier on the generator’s output – enabling it to improve over time.  The GAN structure is set up as a competition (hence “adversarial”) – with the two models essentially programmed to try to outdo each other: the generator is trying to get its output past the discriminator, while the discriminator is working to spot artificially generated data. This structured contest can optimize AI outputs to be more realistic without the need for additional human intervention. Though GANs work best for narrower applications (such as producing realistic photos or videos), rather than general purpose AI. Hallucination Hallucination is the AI industry’s preferred term for AI models making stuff up – literally generating information that is incorrect. Obviously, it’s a huge problem for AI quality.  Hallucinations produce GenAI outputs that can be misleading and could even lead to real-life risks — with potentially dangerous consequences (think of a health query that returns harmful medical advice). This is why most GenAI tools’ small print now warns users to verify AI-generated answers, even though such disclaimers are usually far less prominent than the information the tools dispense at the touch of a button. The problem of AIs fabricating information is thought to arise as a consequence of gaps in training data. For general purpose GenAI especially — also sometimes known as foundation models — this looks difficult to resolve. There is simply not enough data in existence to train AI models to comprehensively resolve all the questions we could possibly ask. TL;DR: we haven’t invented God (yet).  Hallucinations are contributing to a push towards increasingly specialized and/or vertical AI models — i.e. domain-specific AIs that require narrower expertise – as a way to reduce the likelihood of knowledge gaps and shrink disinformation risks. Inference Inference is the process of running an AI model. It’s setting a model loose to make predictions or draw conclusions from previously-seen data. To be clear, inference can’t happen without training; a model must learn patterns in a set of data before it can effectively extrapolate from this training data. Many types of hardware can perform inference, ranging from smartphone processors to beefy GPUs to custom-designed AI accelerators. But not all of them can run models equally well. Very large models would take ages to make predictions on, say, a laptop versus a cloud server with high-end AI chips. [See: Training] Large language model (LLM) Large language models, or LLMs, are the AI models used by popular AI assistants, such as ChatGPT, Claude, Google’s Gemini, Meta’s AI Llama, Microsoft Copilot, or Mistral’s Le Chat. When you chat with an AI assistant, you interact with a large language model that processes your request directly or with the help of different available tools, such as web browsing or code interpreters. AI assistants and LLMs can have different names. For instance, GPT is OpenAI’s large language model and ChatGPT is the AI assistant product. LLMs are deep neural networks made of billions of numerical parameters (or weights, see below) that learn the relationships between words and phrases and create a representation of language, a sort of multidimensional map of words. These models are created from encoding the patterns they find in billions of books, articles, and transcripts. When you prompt an LLM, the model generates the most likely pattern that fits the prompt. It then evaluates the most probable next word after the last one based on what was said before. Repeat, repeat, and repeat. (See: Neural network) Neural network A neural network refers to the multi-layered algorithmic structure that underpins deep learning — and, more broadly, the whole boom in generative AI tools following the emergence of large language models.  Although the idea of taking inspiration from the densely interconnected pathways of the human brain as a design structure for data processing algorithms dates all the way back to the 1940s, it was the much more recent rise of graphical processing hardware (GPUs) — via the video game industry — that really unlocked the power of this theory. These chips proved well suited to training algorithms with many more layers than was possible in earlier epochs — enabling neural network-based AI systems to achieve far better performance across many domains, including voice recognition, autonomous navigation, and drug discovery. (See: Large language model [LLM]) Training Developing machine learning AIs involves a process known as training. In simple terms, this refers to data being fed in in order that the model can learn from patterns and generate useful outputs. Things can get a bit philosophical at this point in the AI stack — since, pre-training, the mathematical structure that’s used as the starting point for developing a learning system is just a bunch of layers and random numbers. It’s only through training that the AI model really takes shape. Essentially, it’s the process of the system responding to characteristics in the data that enables it to adapt outputs towards a sought-for goal — whether that’s identifying images of cats or producing a haiku on demand. It’s important to note that not all AI requires training. Rules-based AIs that are programmed to follow manually predefined instructions — for example, such as linear chatbots — don’t need to undergo training. However, such AI systems are likely to be more constrained than (well-trained) self-learning systems. Still, training can be expensive because it requires lots of inputs — and, typically, the volumes of inputs required for such models have been trending upwards. Hybrid approaches can sometimes be used to shortcut model development and help manage costs. Such as doing data-driven fine-tuning of a rules-based AI — meaning development requires less data, compute, energy, and algorithmic complexity than if the developer had started building from scratch. [See: Inference] Transfer learning A technique where a previously trained AI model is used as the starting point for developing a new model for a different but typically related task – allowing knowledge gained in previous training cycles to be reapplied.  Transfer learning can drive efficiency savings by shortcutting model development. It can also be useful when data for the task that the model is being developed for is somewhat limited. But it’s important to note that the approach has limitations. Models that rely on transfer learning to gain generalized capabilities will likely require training on additional data in order to perform well in their domain of focus (See: Fine tuning) Weights Weights are core to AI training, as they determine how much importance (or weight) is given to different features (or input variables) in the data used for training the system — thereby shaping the AI model’s output.  Put another way, weights are numerical parameters that define what’s most salient in a dataset for the given training task. They achieve their function by applying multiplication to inputs. Model training typically begins with weights that are randomly assigned, but as the process unfolds, the weights adjust as the model seeks to arrive at an output that more closely matches the target. For example, an AI model for predicting housing prices that’s trained on historical real estate data for a target location could include weights for features such as the number of bedrooms and bathrooms, whether a property is detached or semi-detached, whether it has parking, a garage, and so on.  Ultimately, the weights the model attaches to each of these inputs reflect how much they influence the value of a property, based on the given dataset. Topics

NVIDIA has released Llama Nemotron Nano 4B, an open-source reasoning model designed to deliver strong performance and efficiency across scientific tasks, programming, symbolic math, function calling, and instruction following—while being compact enough for edge deployment. With just 4 billion parameters, it achieves higher accuracy and up to 50% greater throughput than comparable open models with up to 8 billion parameters, according to internal benchmarks. The model is positioned as a practical foundation for deploying language-based AI agents in resource-constrained environments. By focusing on inference efficiency, Llama Nemotron Nano 4B addresses a growing demand for compact models capable of supporting hybrid reasoning and instruction-following tasks outside traditional cloud settings. Model Architecture and Training Stack Nemotron Nano 4B builds upon the Llama 3.1 architecture and shares lineage with NVIDIA’s earlier “Minitron” family. The architecture follows a dense, decoder-only transformer design. The model has been optimized for performance in reasoning-intensive workloads while maintaining a lightweight parameter count. The post-training stack for the model includes multi-stage supervised fine-tuning on curated datasets for mathematics, coding, reasoning tasks, and function calling. In addition to traditional supervised learning, Nemotron Nano 4B has undergone reinforcement learning optimization using Reward-aware Preference Optimization (RPO), a method intended to enhance the model’s utility in chat-based and instruction-following environments. This combination of instruction tuning and reward modeling helps align the model’s outputs more closely with user intent, particularly in multi-turn reasoning scenarios. The training approach reflects NVIDIA’s emphasis on aligning smaller models to practical usage tasks that traditionally require significantly larger parameter sizes. Performance Benchmarks Despite its compact footprint, Nemotron Nano 4B exhibits robust performance in both single-turn and multi-turn reasoning tasks. According to NVIDIA, it provides 50% higher inference throughput compared to similar open-weight models within the 8B parameter range. The model supports a context window of up to 128,000 tokens, which is particularly useful for tasks involving long documents, nested function calls, or multi-hop reasoning chains. While NVIDIA has not disclosed full benchmark tables in the Hugging Face documentation, the model reportedly outperforms other open alternatives in benchmarks across math, code generation, and function calling precision. Its throughput advantage suggests it can serve as a viable default for developers targeting efficient inference pipelines with moderately complex workloads. Edge-Ready Deployment One of the core differentiators of Nemotron Nano 4B is its focus on edge deployment. The model has been explicitly tested and optimized to run efficiently on NVIDIA Jetson platforms and NVIDIA RTX GPUs. This enables real-time reasoning capabilities on low-power embedded devices, including robotics systems, autonomous edge agents, or local developer workstations. For enterprises and research teams concerned with privacy and deployment control, the ability to run advanced reasoning models locally—without relying on cloud inference APIs—can provide both cost savings and greater flexibility. Licensing and Access The model is released under the NVIDIA Open Model License, which permits commercial usage. It is available through Hugging Face at huggingface.co/nvidia/Llama-3.1-Nemotron-Nano-4B-v1.1, with all relevant model weights, configuration files, and tokenizer artifacts openly accessible. The license structure aligns with NVIDIA’s broader strategy of supporting developer ecosystems around its open models. Conclusion Nemotron Nano 4B represents NVIDIA’s continued investment in bringing scalable, practical AI models to a broader development audience—especially those targeting edge or cost-sensitive deployment scenarios. While the field continues to see rapid progress in ultra-large models, compact and efficient models like Nemotron Nano 4B provide a counterbalance, enabling deployment flexibility without compromising too heavily on performance. Check out the Model on Hugging Face. All credit for this research goes to the researchers of this project. Also, feel free to follow us on Twitter and don’t forget to join our 95k+ ML SubReddit and Subscribe to our Newsletter. Asif RazzaqWebsite |  + postsBioAsif Razzaq is the CEO of Marktechpost Media Inc.. As a visionary entrepreneur and engineer, Asif is committed to harnessing the potential of Artificial Intelligence for social good. His most recent endeavor is the launch of an Artificial Intelligence Media Platform, Marktechpost, which stands out for its in-depth coverage of machine learning and deep learning news that is both technically sound and easily understandable by a wide audience. The platform boasts of over 2 million monthly views, illustrating its popularity among audiences.Asif Razzaqhttps://www.marktechpost.com/author/6flvq/A Coding Implementation to Build an AI Agent with Live Python Execution and Automated ValidationAsif Razzaqhttps://www.marktechpost.com/author/6flvq/Step-by-Step Guide to Build a Customizable Multi-Tool AI Agent with LangGraph and Claude for Dynamic Agent CreationAsif Razzaqhttps://www.marktechpost.com/author/6flvq/A Comprehensive Coding Guide to Crafting Advanced Round-Robin Multi-Agent Workflows with Microsoft AutoGenAsif Razzaqhttps://www.marktechpost.com/author/6flvq/Microsoft AI Introduces Magentic-UI: An Open-Source Agent Prototype that Works with People to Complete Complex Tasks that Require Multi-Step Planning and Browser Use