Every seasoned marketing professional I speak to echoes a similar sentiment: something fundamental has shifted, and the old playbooks are no longer effective. This is not just a platform change; it is a psychological one. For the first time, users are moving from searching for information to seeking certainty, a distinction that changes everything.

When Behavior Was Predictable

I remember the era when a celebrity endorsement on television was a near-guarantee for brand loyalty. Fans followed their idols to products; attention created association, which led to purchase. This formula worked consistently for decades until the internet digitized that predictability rather than dismantling it.

Search engines like Google turned discovery into a structured keyword-driven system. Users searched, engines returned ranked results, and businesses at the top won the customer. Core principle: be visible, and you will be chosen. This shaped marketing strategy for nearly thirty years.

What Has Actually Changed

The change is not about which platform wins; it runs deeper – how people make decisions. Celebrity credibility has eroded as consumers understand the commercial ecosystem. An endorsement is seen as a transaction, not a sufficient reason to spend money. Younger consumers, especially Gen Z and late millennials, have moved toward first-hand experience or that of someone relatable. They verify everything.

The online and offline distinction has dissolved. A consumer sees a product in a store and pulls out a phone before buying. They cross-check recommendations from friends. Behaviors that once lived in separate worlds now happen simultaneously, fluidly, and constantly.

What the Research Showed

To test these observations, I conducted an in-person field survey starting in mid-2025 with nearly 500 respondents across diverse age groups and backgrounds. The results confirmed the pattern.

• Among 16–20 year olds, 87% said their primary trust for purchase decisions lies with friends, parents, or teachers.
• In the 21–30 age group, 73% blend peer input with social media, but 96% re-verify suggestions before acting.
• Among 31–40 year olds, 65% exhibit similar verification behavior.
• Even in the 41+ segment, 44% now follow the same pattern, showing the direction of change across all generations.

The common thread: trust is no longer accepted; it is earned and then verified. Consumers of every generation have become active validators, not passive recipients.

Are LLMs an Innovation or a Response to Market Pressure?

Technology history shows a pattern every ten to fifteen years: radio to television, television to internet, internet to search engines, search engines to social media. Each revolution changed buyer behavior. To understand the AI era, the first question is not “how do I optimize for this platform?” but “how has buyer behavior changed, and why?”

The rise of large language models (ChatGPT, Gemini, Perplexity) is a direct response to this psychological shift. These tools did not create the verification instinct; they answered it. Traditional search engines offered lists of options, leaving users to sort through competing claims. LLMs synthesize multiple sources and return structured answers, precisely what consumers needed: faster, more certain decision-making.

Tech giants investing aggressively in LLMs were not purely motivated by innovation. They recognized that their audiences were fragmenting across social media, e-commerce, and other channels. LLMs are a strategic attempt to re-aggregate that audience under a single trusted interface. They are building these tools because remaining passive risks losing the next interface layer of the internet. The stakes are high: if users trust an LLM for purchase decisions, that LLM must remain unbiased. Any perception of commercial favoritism leads users to abandon it for a more neutral tool. The entire value proposition depends on perceived trustworthiness.

What This Means for Brands

The shift from visibility to credibility is profound. In the old paradigm, showing up frequently and loudly enough led to being chosen. Now, showing up is necessary but insufficient. If a brand cannot survive the moment a consumer decides to verify claims through AI, peer networks, reviews, or independent sources, it will not remain a consideration.

Consider how consumers make even small purchase decisions: discover on TikTok or Instagram, search for YouTube reviews, cross-check opinions on Reddit, compare alternatives on Google, and finally ask ChatGPT to summarize the best option. The behavior, not the number of platforms, matters.

Take a procurement manager evaluating CX outsourcing vendors: encounter a shortlist via AI Overview, cross-check on Clutch or G2, look for case studies, scan industry forums, and ask ChatGPT to compare options. A company with verified reviews, documented case studies, and third-party coverage survives that journey; one without does not.

Consumers build confidence through layered verification. For brands, this means thinking less about impression count and more about information integrity: Are claims verifiable? Consistency across website, third-party reviews, forums, AI summaries? Is there enough legitimate information for an LLM to surface the brand accurately? These are infrastructure questions, not just marketing ones.

Most brands still optimize for reach, frequency, and creative impact. The ones pulling ahead are making themselves easy to trust at the exact moment a skeptical consumer looks closer – not by being louder, but by having nothing to hide.

The Deeper Shift

After everything observed in the survey and three decades of market movement, the underlying human need has not changed: people want certainty before they commit. What has changed is the threshold for that certainty and the speed at which they expect to reach it. Search has not become less important; it has become more decisive. Users increasingly seek to reduce uncertainty quickly. If a brand cannot be part of that moment in a way that holds up to scrutiny, in that moment of decision, that brand simply does not exist.

That is a harder problem than getting SEO right, but also a more honest one, because it forces brands to ask not just “how do I get found?” but “do I deserve to be chosen?” In the AI era, that is the only question that matters.

Source: TNW | Artificial-Intelligence News

Key Facts:

• Salesforce closed 29,000 Agentforce deals with $800 million in annual recurring revenue (ARR).
• Stock fell nearly 21% in 2025 and another 30% in 2026, amid the broader SaaSpocalypse selloff.
• Showcase demos from Williams-Sonoma, UChicago Medicine, and SharkNinja were works in progress, not live deployments.
• Revenue growth slowed from roughly 25% a few years ago to about 10% in fiscal 2026.
• Consumption-based pricing (Agentforce) uses 'agentic work units' rather than per-seat licences.

Salesforce has a problem that no amount of marketing can fix. The company has built its entire narrative around Agentforce, its AI agent platform, and the numbers look impressive on paper: 29,000 deals closed, $800 million in annual recurring revenue, and a roadmap that promises to replace entire categories of human work. But Wall Street is not buying it, and the gap between what Salesforce shows on stage and what customers actually use keeps getting wider.

The stock tells the story. Salesforce shares fell nearly 21 per cent in 2025 and have dropped another 30 per cent so far in 2026. The decline tracks a broader selloff in software-as-a-service companies, an event the market has taken to calling the SaaSpocalypse. Roughly $285 billion in SaaS market capitalisation evaporated in a single 48-hour window in February. The logic is simple: if one AI agent can do the work of ten employees, why would a company pay for ten seats? This existential threat has hit Salesforce particularly hard because its business model has historically relied on per-seat pricing. The company's transition to a consumption-based model with Agentforce is seen as a necessary but risky pivot.

Salesforce has tried to get ahead of that question by positioning itself as the company that sells the agents rather than the seats. CEO Marc Benioff has called Agentforce a “digital labour platform.” On earnings calls, the company cites the 29,000 deals and the ARR figure as proof that enterprises are buying in. However, the details behind those numbers reveal a more cautious reality. Many of the deals are small pilot programmes or limited-scope deployments, not the sweeping enterprise-wide rollouts that would justify the hype. The $800 million ARR, while significant, represents only a fraction of Salesforce's total revenue of $41.5 billion in fiscal 2026, and it remains to be seen whether that figure can grow fast enough to offset the decline in traditional seat-based revenue.

The trouble is that the showcase examples keep falling apart under scrutiny. At Dreamforce, Salesforce demonstrated a Williams-Sonoma AI agent called Olive that was supposed to act as an agentic sous chef, helping customers plan meals and find products. In practice, Olive struggled with specific questions and recommendations. The agent’s more advanced capabilities were described using future tense, “will soon be able to,” rather than as features that were live. This pattern of overselling is not new for Salesforce; the company has a history of making bold promises that take years to materialise. But in the fast-moving AI market, the gap between promise and reality is more damaging than ever.

A similar pattern appeared with the University of Chicago Medicine. Salesforce presented the hospital system as a flagship Agentforce for Health deployment. The reality was more modest: UChicago Medicine’s first AI agent launched on web chat to handle basic questions like parking directions and clinic availability. The more ambitious features, including voice-based patient support, were still in development. For a hospital system that deals with complex medical workflows, a chatbot that answers parking questions is far from the transformative agentic platform that Salesforce advertised. This raises questions about whether Agentforce is genuinely ready for high-stakes environments like healthcare, where errors can have serious consequences.

SharkNinja, the maker of Shark vacuums and Ninja kitchen appliances, was another headline customer. Salesforce said the company would use Agentforce to streamline customer service. Bloomberg reported a 20 per cent reduction in support calls as part of the pitch. But the deployment described was forward-looking, with agents expected to “guide customers through the buying process” and “manage returns,” not a report on outcomes already achieved. In other words, the 20% reduction was a projection, not a result. This kind of future-tense marketing is common in the tech industry, but investors are becoming increasingly sceptical as they demand evidence of real-world impact.

This matters because Salesforce is not the only company overselling AI capabilities. Apple agreed to pay $250 million in May to settle a class action lawsuit alleging it had exaggerated what Apple Intelligence and a smarter Siri would deliver when it launched the iPhone 16. The settlement covered claims that the company’s marketing went well beyond what the technology could do at launch. The parallel is striking: both Apple and Salesforce are household names that have built their reputations on innovation, yet both fell into the trap of overpromising on AI. The difference is that Apple settled quickly, while Salesforce continues to double down on its narrative.

Salesforce’s financial trajectory adds another layer. Revenue growth has slowed from roughly 25 per cent a few years ago to about 10 per cent in fiscal 2026, when the company reported $41.5 billion in total revenue. That is still a large business, and the company delivered a strong fourth quarter with 12 per cent growth. But the deceleration is exactly what investors fear when they hear that AI agents will compress the number of human users who need software licences. Even if Agentforce succeeds, it may cannibalise Salesforce's existing revenue streams faster than the new model can replace them. This is the central tension that has driven the stock decline.

The company has tried to address the pricing question. Agentforce uses a consumption-based model rather than traditional per-seat pricing, charging for what Salesforce calls “agentic work units.” It has consumed nearly 20 trillion tokens and converted them into more than 2.4 billion such units. Whether that model can grow fast enough to offset the structural threat to seat-based revenue is the central bet. Analysts are divided: some see it as a bold move that aligns incentives with customers, while others warn that it could lead to unpredictable revenue streams and customer resistance to variable costs. The success of this pricing model will depend on how quickly enterprises adopt AI agents at scale, a process that is still in its early stages.

Smaller customers illustrate both the promise and the cost. The city of Kyle, Texas, deployed Agentforce to run its 311 service, handling more than 12,000 resident requests since March 2025 with nearly 90 per cent first-call resolution. Bloomberg reported the city doubled its Salesforce spending to $300,000. For a fast-growing municipality, that may be a reasonable investment. For enterprise customers weighing the same calculus at scale, the economics are less clear. A large company with millions of customer interactions could see costs spiral if the consumption model is not carefully managed. Moreover, the ROI of replacing human agents with AI depends on accuracy, customer satisfaction, and the cost of errors, factors that are still being measured.

The competitive pressure is real. SAP unveiled its Autonomous Enterprise with more than 200 AI agents and an Anthropic partnership at Sapphire 2026. ServiceNow, Google, and Microsoft are all building agent platforms. The question is no longer whether AI agents will reshape enterprise software but whether Salesforce can maintain its position as the market reprices around it. Each competitor brings a different advantage: SAP has deep enterprise relationships, Microsoft has Azure and Copilot integration, and Google has its vast data and AI research. Salesforce's strength lies in its customer relationship management (CRM) data, but that may not be enough if agents become commoditised.

Benioff has responded with characteristic confidence, announcing a new revenue target of $60 billion by fiscal 2030. He has also committed $50 billion in share buybacks, a signal to investors that the company believes its stock is undervalued. Slack’s transformation into an agentic platform, with more than 30 new AI capabilities and mandatory bundling with every new Salesforce account from this summer, is part of that push. The Slack integration is particularly important because it ties together communication and workflows, creating a sticky ecosystem that competitors will find hard to replicate. However, mandatory bundling also risks alienating customers who prefer to choose their tools independently.

None of this resolves the core tension. Salesforce is asking customers to pay for a future that its own demos have not yet delivered, while asking investors to trust that consumption-based AI revenue will replace the seat-based model that built the company. The 29,000 deals are real. The $800 million in ARR is real. But the agentic AI market rewards outcomes, not announcements, and the gap between the two is where Salesforce’s credibility will be tested. The coming quarters will be critical: if the company can show that its showcase customers are moving from pilots to production, confidence may return. If not, the stock could face further declines, and the narrative of Salesforce as an AI leader may unravel.

Source: TNW | Apps News

The layoffs affect roughly 22 out of every 100 employees. Evans emphasized that the company is not simply trimming fat but fundamentally rethinking how work gets done. The '100x org' model rests on the premise that AI agents have changed what it means to build software. Incremental improvements to existing systems, he argued, will not suffice; ClickUp needs to rebuild its structure from the ground up.

This restructuring follows months of aggressive AI adoption inside ClickUp. A recent Fortune profile revealed that the company now operates roughly 3,000 internal AI agents across its departments, creating a 3:1 ratio of agents to human employees. Evans has even mandated that staff interact with an AI agent trained to stand in his place before contacting him directly. This level of automation is central to the company's new vision.

Evans outlined three categories of employees he sees as essential in the new model. The first is 'builders,' which he splits into 10x engineers and 10x product managers. His claim is blunt: the best engineers are no longer writing code themselves. Instead, they direct AI agents that write code, and the critical skill becomes judgment — the ability to orchestrate and review the output of these agents. AI makes top engineers wildly more productive, he wrote, while everyone else using AI slows them down. He called this the 'great reckoning of AI coding' and predicted that every company will face it soon. Companies celebrating a 500% increase in pull requests are generating volume, not outcomes, he argued; more code simply creates another bottleneck.

The second category is 'system managers,' or agent managers. These are people who automate their own jobs with AI and then become owners of the systems they built. Evans argued that anyone who automates their role will always have a job, because the underlying systems — not the individual tasks — become the source of value. The third category is 'front-liners,' the people who spend their time with customers. In a world saturated with AI communication, Evans said, human contact becomes the one bottleneck that companies should not try to replace. Front-liners should spend nearly 100% of their time in meetings with customers, while the systems around those meetings are fully automated.

Product management and design are also merging, according to Evans. Designers with strong customer focus become more like product managers, and product managers with UX intuition become more like designers. The bottleneck of user research is gone, he claimed, because a single mention to an AI agent can kick off and analyze an entire research cycle. This blending of roles is part of the leaner, more agile structure ClickUp is aiming for.

The most provocative element of the announcement is the compensation model. ClickUp is introducing salary bands that reach $1 million per year in cash. That path is open to nearly anyone in the company who produces what Evans calls '100x impact' by creating or managing AI systems. In a world where the best people create 100 times more output than average, he argued, companies cannot afford to lose them and should aim to retain them for decades. This approach flips traditional compensation structures on their head, especially in an industry where many companies are cutting pay or freezing salaries amid downturns.

The announcement lands in the middle of a brutal stretch for tech workers. The industry has shed more than 100,000 jobs across roughly 250 events in 2026 so far. Meta cut 8,000 roles the same week despite record revenue. Oracle eliminated up to 30,000 positions to fund AI infrastructure. GitLab restructured for the 'agentic era.' The pattern is consistent: companies report record performance and cut headcount simultaneously, redirecting savings into AI investments. Evans's framing is more explicit than most. Where other CEOs use euphemisms about efficiency and realignment, he is making a direct argument that the roles being eliminated are structurally obsolete. Whether that is candour or hubris will depend on whether the '100x org' delivers the outcomes he is promising.

ClickUp reported roughly $300 million in annual recurring revenue as of 2025 and has been eyeing an IPO. The company acquired AI coding platform Codegen late last year. With AI reshaping the economics of developer tools and productivity software, Evans is betting that a smaller, better-paid workforce directing thousands of agents will outperform the company it replaces. The broader industry is watching closely, as ClickUp's experiment could serve as a template — or a cautionary tale — for how companies navigate the transition to AI-first operations.

Not everyone is convinced. In China, courts have ruled that replacing workers with AI is not legal grounds for dismissal. In the US, no such protection exists for employees. For the 22% of ClickUp employees who lost their jobs this week, the distinction matters deeply. The company's bold bet on AI may ultimately prove visionary, but for those laid off, the immediate impact is personal and painful. The tech industry's relationship with AI continues to evolve, and ClickUp's moves are a stark reminder of the human cost behind the promise of exponential productivity.

Source: TNW | Apps News

The UFC has officially announced that Conor McGregor, the former two-division champion, will make his long-awaited return to the octagon on July 11, 2026, at UFC 329 in Las Vegas. He will face Max Holloway, the former featherweight champion, in a rematch that reignites a rivalry first sparked over a decade ago. This fight marks McGregor's first appearance since July 2021, when he lost to Dustin Poirier by TKO at UFC 264. Known for his explosive striking and larger-than-life persona, McGregor has been a central figure in mixed martial arts since his debut in the UFC in 2013. His career has been defined by historic achievements, including becoming the first fighter to hold UFC championships in two weight classes simultaneously. However, his path has also been marred by legal issues, suspensions, and a prolonged absence from competition. The announcement of this fight comes at a crucial time for the sport, as the UFC seeks to headline major events with its biggest star.

Max Holloway, at 34 years old, has evolved into one of the most respected fighters in the sport. Known for his incredible volume striking, cardio, and durability, Holloway has defeated top contenders such as Jose Aldo, Frankie Edgar, and Brian Ortega. He has previously fought for the lightweight title and continues to thrive in the featherweight division. The first meeting between McGregor and Holloway took place in August 2013 at UFC Fight Night 26. At that time, McGregor won by unanimous decision after controlling the fight with his precision striking and range. Since then, Holloway has compiled a 21–7 record, capturing the interim and undisputed featherweight titles, while McGregor has faced a series of ups and downs, including victories over Eddie Alvarez, Nate Diaz, and Donald Cerrone, but also losses to Khabib Nurmagomedov and Dustin Poirier.

The announcement of McGregor's return was made by the UFC on Saturday, with promotional materials highlighting the historic nature of the event. UFC 329 will be held at the T-Mobile Arena in Las Vegas, a venue synonymous with major combat sports events. The fight came together after months of speculation about McGregor's recovery from a broken toe suffered during a sparring session in 2024, which forced the cancellation of a planned return against Michael Chandler. McGregor has since completed his rehabilitation and has been training at the SBG Ireland facility under coach John Kavanagh. The fight agreement includes a weight class stipulation that McGregor will compete at lightweight (155 pounds), a division where he once reigned as champion.

Beyond the cage, McGregor has been embroiled in controversies that have affected his career. In November 2024, he was found liable in a civil case in Ireland for the rape and assault of a woman in a Dublin hotel in 2018. He was ordered to pay $250,000 in damages, though a criminal investigation did not lead to charges. McGregor appealed the verdict but lost in July 2025. Additionally, he was suspended for 18 months by the US Anti-Doping Agency for failing to comply with drug-testing protocols. Three missed tests between 2024 and 2025 led to the suspension, which was backdated to September 2024 and expired in March 2026. This suspension cleared the way for his return to competition. McGregor also attempted to enter the political arena, announcing his candidacy for the Irish presidency in 2025. However, he withdrew his candidacy in September 2025 after a series of controversial statements that were widely condemned as xenophobic. These events have polarized public opinion, with some fans criticizing his behavior and others viewing him as a victim of systemic bias.

The timing of the fight announcement is strategic. It was made just before Francis Ngannou, a former UFC heavyweight champion, made his return to the cage under the MVP promotion, which was broadcast on Netflix. Ngannou and UFC president Dana White parted ways on bitter terms in early 2023, with disputes over fighter pay and contractual freedom. By announcing the McGregor fight, White effectively overshadowed Ngannou's event, drawing attention back to the UFC. The tactic worked, as the announcement garnered over 10,000 retweets on X (formerly Twitter) within three hours. This move underscores the intense rivalry between the UFC and competing organizations, as well as the promotional power of McGregor's name.

UFC 329 will also feature other notable bouts. French fighter Benoit Saint-Denis will face Mike Plimbett in a lightweight co-main event, adding further intrigue to the card. Saint-Denis has risen through the ranks with a series of impressive wins, including submissions over Gabriel Miranda and Marc Diakiese. His presence on the card highlights the UFC's expanding international roster. The event is expected to draw a massive pay-per-view audience, with experts predicting buys in the range of 2 to 3 million, compared to the 1.8 million buys for McGregor's last fight against Poirier in 2021. The promotion has already launched a extensive marketing campaign featuring McCregor's iconic walkout music, signature striking style, and confrontational pre-fight press conferences.

Holloway, on the other hand, approaches this fight with confidence. He has won four of his last five fights, with his only loss coming to Alexander Volkanovski in a close decision. He is known for his ability to adapt during fights, as demonstrated in his victories over Calvin Kattar and Arnold Allen. A win over McGregor would catapult him into title contention and likely earn him a shot at the lightweight or featherweight championship. Holloway has expressed respect for McGregor's skills but has also criticized his conduct outside the cage, stating that he wants to win for his family and fans. The fight presents a stark contrast in styles: McGregor's powerful left hand and counter-striking against Holloway's relentless volume and pressure.

The fight also has significant historical implications. McGregor's last victory came in January 2020 against Donald Cerrone, ending a 15-month layoff. Since then, he has faced physical and legal challenges that have kept him away from competition. At 37, McGregor is older than many elite fighters, and questions about his durability and motivation abound. However, his training camp reports indicate he is in peak condition, having gained muscle mass while maintaining speed. Holloway, six years younger, has the advantage of active competition and recent fight experience. The matchup echoes McGregor's earlier rise, when he defeated top contenders to earn title shots. A win here could set up a trilogy fight with Poirier or a clash with lightweight champion Islam Makhachev.

The broader MMA landscape is also watching closely. The success of this event could determine the UFC's strategy for future mega-fights, especially as the organization faces competition from promotions like Professional Fighters League and MVP. McGregor's ability to draw casual viewers remains unmatched, and his return injects excitement into a sport that has seen many stars retire or move to other ventures. However, the fight also carries risks. If McGregor loses badly, it could signal the end of his asian career and reduce his drawing power. Conversely, a win could rejuvenate his career and set up high-profile matchups for the remainder of the decade.

As the July 11 date approaches, fans and analysts will scrutinize every piece of news about training, weigh-ins, and press conferences. The UFC has already scheduled a press conference for early June, where both fighters are expected to exchange verbal jabs. Tickets for UFC 329 went on sale immediately after the announcement and sold out within hours, with resale prices reaching into the thousands of dollars. The event promises to be a landmark moment for mixed martial arts, uniting the sport's past and present in a single night.

Source: Eurosport News

A Beloved Franchise with a Unique Blend

The Armour of God series has often been described as a mix of Indiana Jones and Mission: Impossible, but with a distinct Jackie Chan flavor. The films follow Jacky (often called Asian Hawk or the Condor), an adventurer and treasure hunter who travels the globe in search of ancient relics. The combination of comedy, death-defying stunts, and martial arts choreography has made the series a cult favorite since the first film debuted in 1986.

Chan originally directed and starred in the first three movies: Armour of God (known in France as Mister Dynamite), Armour of God II: Operation Condor (1991), and Chinese Zodiac (2012). For the upcoming fourth film, Chan will step back from directing and focus solely on acting. The film will be directed by Robert Kun and produced by Salem Entertainment, with production set to begin in July 2025.

Production Details and Locations

Armour of God IV: Ultimatum will be shot in Kazakhstan and Azerbaijan, marking a first for the franchise. The choice of locations is not only scenic but also strategic, as it opens up new creative possibilities. Chan expressed his excitement in a statement to Variety: "This is my first project with Kazakh partners, but it won't be my last. We will soon make more announcements. I have never filmed there before, and it opens the door to many new visual and creative ideas that we can share with the audience."

Producer Igor Tsay echoed this enthusiasm: "Having worked on productions around the world, I can say this is an absolutely historic moment for Central Asian cinema. Welcoming a legendary franchise with Jackie Chan is both a great honor and a great responsibility for the development of the region's creative industry." The film is scheduled for a global release in the second quarter of 2027.

The Plot: A New Artifact to Find

The story of Armour of God IV: Ultimatum will see Asian Hawk (Chan) tasked with locating a rare artifact known as the Tumar, estimated to be worth over $20 million. As always, his mission is filled with high-stakes action, dangerous chases, and witty humor. The plot follows in the footsteps of previous films, where Chan's character often outwits enemies using acrobatics and resourcefulness rather than brute force.

Jackie Chan's Career and Enduring Legacy

Jackie Chan is one of the most recognizable action stars in cinema history. Born in 1954, he began training in martial arts and acrobatics at a young age. He rose to fame in Hong Kong with films like Drunken Master and Police Story, and later achieved international stardom with Rush Hour, Shanghai Noon, and The Karate Kid. What sets Chan apart is his commitment to performing his own stunts, often risking life and limb for authentic action sequences.

Throughout his career, Chan has sustained numerous injuries. The most severe occurred during the filming of the first Armour of God in 1986. While performing what seemed like a routine jump, a tree branch broke under his weight, causing him to fall and hit his head on a rock. The accident resulted in a fractured skull, requiring six months of bed rest. Chan suffered permanent damage, including hearing loss in his right ear and a metal plate in his skull to fill the hole. Despite this, he resumed filming a year later and continued to do his own stunts.

The series itself showcases some of Chan's most creative work. Mister Dynamite introduced audiences to his unique brand of slapstick martial arts. Operation Condor expanded the scope with exotic locations and larger set pieces. Chinese Zodiac featured high-tech gadgets and a global treasure hunt, marking a more ambitious production. Now, Ultimatum promises to continue that tradition while updating the franchise for modern audiences.

The Significance of Central Asian Filming

Filming in Kazakhstan and Azerbaijan is a major step for both the franchise and the region's film industry. Central Asia has become an attractive destination for international productions due to its diverse landscapes and growing infrastructure. By choosing these locations, the producers are tapping into a new visual palette, with mountains, deserts, and ancient architecture that can substitute for various settings in the film. This move also aligns with Chan's interest in exploring new cultures and working with local talent.

The announcement has generated buzz among fans who appreciate Chan's dedication to his craft even in his 70s. Unlike many action stars who retire or shift to less physical roles, Chan continues to push boundaries. In recent years, he has appeared in films like Vanguard (2020) and The Diary (2021), and he shows no signs of slowing down. His work ethic and passion for entertaining audiences remain as strong as ever.

Comparison to Other Action Franchises

The Armour of God series is often compared to Indiana Jones for its archaeological themes and Mission: Impossible for its death-defying stunts. However, Chan's films have a lighter tone and emphasize physical comedy. The character of Asian Hawk is more of a reluctant hero who relies on wit and improvisation. This distinguishes the series from the more serious adventures of Indiana Jones or the high-tech espionage of Ethan Hunt. Chan's ability to blend humor and action is a hallmark of his filmography, and Armour of God epitomizes that blend.

The franchise also holds a special place in Chan's heart. He has spoken about how Mister Dynamite was a turning point in his career, despite the near-fatal accident. The trilogy allowed him to experiment with stunts and storytelling in ways that his other films did not. With the fourth installment, Chan aims to honor the legacy while introducing the series to a new generation of viewers.

Fans are already speculating about the return of supporting characters and the nature of the Tumar artifact. Given Chan's track record, the choreography will likely push the limits of what's possible for a 72-year-old performer. While some may worry about his health, Chan has repeatedly demonstrated that age is just a number. He continues to train daily and remains in remarkable physical condition.

The production of Armour of God IV: Ultimatum is a collaboration between Eastern and Western talent. Director Robert Kun, who has worked on various international projects, brings a fresh perspective. Producer Igor Tsay's involvement signals a commitment to developing the creative ecosystem in Kazakhstan. The film's release in 2027 will undoubtedly be a major event for action cinema fans worldwide.

As Jackie Chan once said, "I'm not afraid of challenges; I'm afraid of not trying." With this new film, he once again proves that his passion for filmmaking and entertaining audiences remains undiminished. The Condor is ready to spread his wings once more.

Source: Yahoo News News

Le 15 mai 2026 restera une date marquante pour les fans de Drake. Alors que l'industrie musicale s'attendait à un simple retour avec l'album Iceman, le Canadien a dévoilé non pas un, mais trois projets complets sur les plateformes de streaming. Un coup marketing audacieux qui rappelle les sorties surprises de Beyoncé ou de Frank Ocean, mais à une échelle jamais vue. Avec Iceman, Habibti et Maid of Honour, Drake veut prouver qu'il reste le maître du rap game, malgré une traversée du désert médiatique après son clash retentissant avec Kendrick Lamar.

Iceman : le retour aux sources agressif

Iceman est présenté comme l'album principal de cette trilogie. Dès les premières secondes, Drake affiche ses ambitions : un flow incisif, des beats lourds signés par son équipe de producteurs habituels (Boi-1da, 40, Metro Boomin), et des attaques directes contre ses détracteurs. Il règle ses comptes avec DJ Khaled, qu'il accuse de silence dans le conflit israélo-palestinien, et se compare même au groupe BTS pour souligner sa longévité. La pochette, un gant pailleté, rend hommage à Michael Jackson, avec qui il avait collaboré virtuellement en 2018 sur Don't Matter to Me. Les featurings incluent Future, 21 Savage et même un sample de Not Like Us retourné contre Kendrick dans un teaser où l'on voit une ferme de bots numériques détruite. Malgré son efficacité, cet album manque d'originalité et recycle les recettes de Scorpion ou Certified Lover Boy. Il reste néanmoins un solide atout pour les playlists.

Habibti : la touche R&B langoureuse

Le deuxième projet, Habibti, tire son nom d'un terme affectueux arabe. Avec seulement 36 minutes, il se veut plus court mais intense. Drake y explore un rap sombre mêlé de R&B posé, rappelant ses tubes estivaux comme One Dance ou Hotline Bling. Les productions sont plus aérées, avec des influences caribéennes et africaines. On y trouve des collaborations avec Wizkid, Tems et un inattendu featuring avec la chanteuse française Aya Nakamura, qui ajoute une touche afrobeat. L'album est calibré pour les festivals et les soirées en plein air. Cependant, il ne surprend guère et s'inscrit dans la continuité de More Life.

Maid of Honour : la révélation électro et introspective

Le troisième album, Maid of Honour, est celui qui sort du lot. Dédié à sa mère, comme l'indique la pochette montrant une photo d'elle jeune, ce disque prend des risques. Drake délaisse le rap pur pour des sonorités synthpop, house et même une touche de drum and bass. Le morceau Cheetah Print sample le titre (It Goes Like) Nanana de Peggy Gou, transformant ce tube de club en une ode sensuelle. Stuck et Goose and The Juice rappellent le meilleur du R&B d'ambiance de The Weeknd, avec des harmonies vocales soignées. D'autres titres, comme Raincheck, intègrent des beats de garage britannique. C'est une version améliorée de Honestly, Nevermind, l'album de 2022 qui avait divisé la critique. Ici, la prise de risque paie : Drake prouve qu'il peut se renouveler sans perdre son identité.

Contexte : une carrière en dents de scie

Ce triple album n'est pas un simple caprice artistique. Il s'inscrit dans une stratégie de reconquête après plusieurs échecs relatifs. Depuis Scorpion (2018), qui avait battu des records avec God's Plan et In My Feelings, Drake a vu son aura s'effriter. Certified Lover Boy (2021) a reçu des critiques mitigées, Honestly, Nevermind (2022) a dérouté son public, et For All the Dogs (2023) n'a pas rencontré le succès escompté. Surtout, la guerre des mots avec Kendrick Lamar a laissé des traces. Le hit Not Like Us de Kendrick, multi-récompensé aux Grammy Awards, a porté un coup dur à l'image de Drake, souvent même moqué dans les médias spécialisés.

La journaliste Sowmya Krishnamurthy, spécialiste du hip-hop, résume : « Le clash avec Kendrick l'a véritablement détrôné. Jusque-là, c'était le leader incontesté en ventes et en tubes. Depuis, il n'a pas réussi à revenir avec un vrai hit. On pardonne tout quand on a un tube. » Drake espère donc que cette avalanche de nouveautés lui permettra de reconquérir sa place. Les chiffres du premier week-end laissent présager un succès : plus de 500 millions de streams cumulés sur les trois albums, un record sur Spotify.

Analyse musicale : pourquoi Maid of Honour est le meilleur

Si chaque album a ses forces, Maid of Honour se distingue par sa cohérence et son audace. Là où Iceman reste dans une zone de confort et Habibti ne surprend pas, le troisième projet montre un Drake capable d'évoluer. Les productions, confiées à des talents émergents comme Kaytranada et BNYX, insufflent une fraîcheur bienvenue. Les textes, plus personnels, évoquent la solitude, la famille et les doutes, loin des joutes verbales habituelles. Enfin, le choix d'inclure des influences house et électro, souvent négligées dans le rap mainstream, démontre une maturité artistique rare.

Un autre point fort est la maîtrise des transitions. L'album alterne entre morceaux dansants (Cheetah Print, Goose and The Juice) et ballades introspectives (Stuck, Mother's Lullaby). La production est minutieuse, chaque sample étant intégré avec soin. Par exemple, le morceau Raincheck reprend un vieux classique de la scène house de Chicago, le transformant en un titre pop-rap imparable. Ces choix montrent une volonté de briser les barrières entre les genres, à l'image de ce que fait The Weeknd depuis des années.

L'impact sur l'industrie et la suite

Avec cette sortie triple, Drake impose un nouveau standard dans le marketing musical. Les fans ont reçu un nombre considérable de contenu, augmentant l'engagement sur les réseaux sociaux. La hashtag DrakeThreeAlbums a été tendance mondiale pendant 48 heures. Cependant, certains critiques estiment que cette profusion pourrait diluer l'attention et nuire à la qualité perçue. Seul l'avenir dira si Drake parviendra à capitaliser sur cet élan. D'ores et déjà, une tournée mondiale est annoncée pour l'automne 2026, avec des dates à Paris, Londres, New York et Tokyo.

En attendant, les fans débattent déjà : quel album est le meilleur ? Les sondages en ligne placent Maid of Honour en tête, suivi de près par Iceman. Les clubs et les playlists estivales devraient toutefois favoriser Habibti. Quoi qu'il en soit, Drake a réussi son coup : il a remis son nom au centre des conversations. Pour un artiste qui a régné sur la décennie 2010, c'est une manière de rappeler qu'il n'est pas encore prêt à céder sa couronne.

Source: Purebreak News

A New Collaboration

Supermax brings together two heavyweights from different corners of Hollywood. Will Smith, who has dominated the box office for decades with hits like Independence Day, Men in Black, and Aladdin, joins forces with David Gordon Green, a director known for his versatility. Green gained acclaim for indie dramas such as George Washington and All the Real Girls, but he reached a wider audience with his comedic work on Pineapple Express and Your Highness. However, it was his 2018 Halloween film—a direct sequel to John Carpenter's original—that revitalized the slasher genre and led to two sequels, Halloween Kills and Halloween Ends. With Supermax, Green steps away from horror to deliver a tense, twist-filled action thriller.

According to industry insiders, the project came together after a heated competition for the script rights. The screenplay was co-written by David Weil and David J. Rosen, both of whom have extensive experience in television. Weil created the Amazon Prime series Hunters, a Nazi-hunting drama starring Al Pacino, while Rosen worked on the sci-fi series Invasion for Apple TV+. Their partnership on Supermax marks their feature film debut as writers, and their script was described as a "propulsive, twist-filled action thriller" that immediately attracted top talent. After a multi-studio bidding war, Miramax secured the rights, and Amazon MGM Studios later acquired worldwide distribution rights in a major deal.

Plot and Characters

Supermax centers on two FBI agents who are tasked with investigating a murder that occurs inside the most secure prison in the world—a so-called "supermax" facility. The prison, often referred to as a "prison within a prison," houses the most dangerous criminals, terrorists, and spies. When a body is discovered under mysterious circumstances, the agents are forced to navigate a labyrinth of lies and betrayal. Will Smith will play one of these agents, while the other agent—his partner—will be a woman. Her casting is still underway, but the dynamic between the two leads is expected to be a key driver of the story.

The film is set to be a fast-paced thriller with multiple twists. Smith's character is described as a seasoned professional who has seen it all, but this case pushes him to his limits. The claustrophobic setting of a supermax prison adds an extra layer of tension, as the agents are trapped inside with inmates who have nothing to lose. The script reportedly weaves together elements of a whodunit with high-stakes action, reminiscent of films like The Rock and Escape from New York but with a modern sensibility.

Production Details and Budget

Supermax is being produced with a substantial budget that Amazon MGM Studios has committed to. While exact figures have not been disclosed, sources confirm that it will be the largest budget David Gordon Green has ever worked with, surpassing even the Halloween trilogy. This investment underscores Amazon's confidence in the project and in Will Smith's enduring star power, even after the infamous incident at the 2022 Oscars. Smith has been gradually rebuilding his public image, and Supermax is seen as a major step in his comeback.

Filming is scheduled to begin in mid-August 2026, with locations yet to be announced. The production will likely involve building detailed sets to recreate the supermax prison, as well as shooting on location in various secure facilities. Given the scale, the film is expected to have a long post-production phase, especially for visual effects and the intricate action sequences.

Release Strategy: Streaming-Only

One of the most notable aspects of Supermax is its release strategy. Unlike many big-budget action films, it will not receive a theatrical release. Instead, it will debut exclusively on Prime Video, Amazon's streaming platform. This decision aligns with Amazon MGM Studios' strategy of prioritizing streaming originals, especially for high-profile projects that can drive subscriber growth. For Will Smith, this marks his second streaming-only film after Emancipation, which was also released on Apple TV+. The decision may disappoint some fans who prefer the big-screen experience, but it also ensures that the film will be available to a global audience immediately.

The lack of a theatrical date has not dampened enthusiasm for the project. Given the success of other streaming action films like Extraction and The Gray Man, Amazon is betting that Supermax will perform well in terms of viewership and critical acclaim. The film's budget suggests that it will feature top-tier production values, and the involvement of David Gordon Green promises a tightly directed thriller.

Will Smith's Career Trajectory

Will Smith has been one of the most bankable stars in Hollywood for over three decades. From his breakout role in The Fresh Prince of Bel-Air to Oscar-winning performances in Ali and King Richard, Smith has shown remarkable range. However, his career faced a significant setback after the 2022 Oscars, where he slapped comedian Chris Rock on stage. The incident led to a ban from the Academy and a brief period of professional isolation. Since then, Smith has slowly returned to the spotlight, starting with Bad Boys: Ride or Die in 2024, which earned over $400 million worldwide. Supermax represents his first project outside of an established franchise and is an opportunity for him to showcase his dramatic skills in a more intense setting.

The actor has also been attached to other projects, including a possible sequel to I Am Legend and a crime thriller titled The Council. But Supermax is the one that has moved fastest through development, thanks to the strong script and the backing of Amazon.

David Gordon Green's Versatility

David Gordon Green has built a career on jumping between genres. After his debut feature George Washington (2000) earned critical acclaim, he delved into comedy with Pineapple Express (2008) and Your Highness (2011). Then, he took over the Halloween franchise, directing three consecutive entries that polarized critics but found commercial success. Supermax allows him to blend his love for character-driven storytelling with tense, action-packed sequences. It will be interesting to see how he handles a prison setting, a location that has been used effectively in classic thrillers like The Shawshank Redemption and Escape from Alcatraz. Green has often said he admires films that create suspense through environment, and the supermax prison offers plenty of opportunities for that.

Anticipation and Next Steps

As the August 2026 start date approaches, casting for the female lead and supporting roles is underway. Several actresses are rumored to be in talks, but no names have been confirmed. The film will also feature a supporting cast of inmates, guards, and other FBI personnel. Given the prestige of the project, many actors will likely vie for roles. Additionally, a composer has yet to be announced, but Green has a history of working with composers like John Carpenter (on the Halloween films) and David Wingo (on his earlier dramas). The music will be crucial in establishing the tense atmosphere of the prison.

Amazon MGM Studios has not yet set a release date for Supermax, but it is expected to arrive on Prime Video sometime in 2027 or early 2028, depending on the post-production schedule. With a script that promises twists, a director known for reinvigorating genres, and a superstar lead looking for a career resurgence, Supermax is shaping up to be one of the most talked-about thrillers in the streaming era. Fans of Will Smith and David Gordon Green will be watching closely as this project unfolds.

The story of two FBI agents trapped in a prison full of the world's worst criminals, while trying to solve a murder, is a recipe for edge-of-your-seat entertainment. If the script delivers on its potential, Supermax could become a benchmark for action thrillers on streaming platforms. All eyes will be on the production this summer as cameras start rolling.

Source: CinéSérie News

Louvre Heist: A National Trauma

On the morning of October 19, 2025, a well-organized commando broke into the Galerie d'Apollon at the Louvre and made off with eight irreplaceable pieces of the French Crown Jewels. According to Interpol, the stolen items—including diadems, necklaces, and brooches once owned by Empress Eugénie, Empress Marie-Louise, and Queen Marie-Amélie—are valued at approximately €88 million. Despite an extensive investigation by French police and Interpol's stolen art database, the jewels remain missing as of early 2026. The heist sent shockwaves through the art and heritage world, with French historians calling it a "wound in the national memory." The stolen pieces were not mere adornments; they symbolized centuries of monarchy and imperial power, often worn during coronations and state ceremonies.

Daniel Roseberry's Schiaparelli: L'Agonie et l'Extase

Kylie Jenner's Met Gala look came from Schiaparelli's Fall/Winter 2026 couture collection titled L'Agonie et l'Extase (The Agony and the Ecstasy), designed by creative director Daniel Roseberry. The collection explored the tension between destruction and beauty, and no piece exemplified this more than Jenner's ensemble. The outfit consisted of a flesh-colored corset that gave the illusion of nudity, paired with a voluminous satin skirt embroidered with thousands of baroque pearls and iridescent fish scales. Schiaparelli's atelier spent over 11,000 hours on the embroidery alone. However, the true focal point was the jewelry: a tarnished silver necklace and matching chandelier earrings that deliberately echoed the forms and motifs of the stolen crown jewels—though set with rhinestones and pearls rather than the original diamonds, emeralds, and sapphires.

Roseberry stated in a behind-the-scenes video that the design was intended to „question the commodification of heritage and the fine line between art and appropriation.“ He noted that the jewelry was not a replica but a reinterpretation, using silicone and silver to mimic the aged metal of the originals. Yet, for many French observers, the timing and context felt like a glamorization of a crime that had deeply affected their national identity.

The Ensemble in Full

Jenner's arrival on the Met Gala steps was theatrical. Her corset, made of nude silicone and covered in tiny crystals, mimicked the texture of marble. The satin skirt fell in heavy folds, studded with pearls that caught the camera flashes. Her hair was slicked back, and her makeup minimal, emphasizing the jewelry. The total look was meant to evoke a statue come to life—a reference to the classical sculptures housed at the Louvre. But the double meaning was inescapable: the jewelry's design whispered of the missing treasures.

Reactions: A Firestorm on Social Media

Within minutes of Jenner's appearance, social media erupted. On X (formerly Twitter), users called the look "tone-deaf" and "insensitive," while others praised the artistic audacity. One viral post read: „Kylie Jenner wearing jewelry inspired by a real unsolved crime that traumatized a whole country is not 'edgy'—it's offensive.“ French accounts especially criticized the choice, noting that the investigation into the Louvre theft is still ongoing and the jewels may never be recovered. Some accused Schiaparelli of exploiting French patrimony for commercial gain.

However, American and international fashion critics were more divided. Vogue described the look as „a masterstroke of narrative design,“ while Women's Wear Daily called it „the most provocative statement of the night.“ On Instagram, celebrities like Rihanna (who wore a diamond-encrusted Bulgari necklace) and Beyoncé (in a Lorraine Schwartz piece) were also heavily jeweled, but their choices did not reference a specific crime. The contrast highlighted the intentional provocation of Jenner's selection.

French media reacted with a mixture of shock and analysis. Vogue France published an op-ed asking: „Peut-on transformer un vol d'État en accessoire de mode?“, while TF1 Info interviewed heritage historians who expressed dismay. „The Crown Jewels are part of our collective story, whether you are a royalist or a republican,“ said Dr. Marie Lefèvre, a cultural historian at the Sorbonne. „To see them reduced to a costume on an American reality star is painful. It feels like the theft is being trivialized.“

The Bigger Picture: Fashion, Crime, and Cultural Appropriation

This is not the first time fashion has drawn inspiration from criminal acts. In 2023, Gucci faced backlash for a collection inspired by the 1970s kidnapping of a fashion heir, and countless brands have used 'prison chic' or images of theft. But the Louvre heist is unique because the objects are of extraordinary national significance and are actively missing. The fashion industry's reliance on shock value often blurs the line between artistic expression and exploitation.

Schiaparelli itself has a history of surrealist provocations. Founder Elsa Schiaparelli collaborated with Salvador Dalí, creating lobster dresses and shoe hats. Daniel Roseberry has revived that spirit, mixing bold motifs with contemporary issues. Yet even defenders of the look concede that the real-world context mattered. „Fashion is meant to make us think, not just look pretty,“ said fashion critic Alexander Fury in an interview. „If we are uncomfortable, that's the point. But there's a difference between discomfort about art and discomfort about a victimized culture.“

Kylie Jenner's Met Gala History

Jenner has attended the Met Gala multiple times since 2015, often in dramatic looks from Versace, Balmain, and now Schiaparelli. Her 2022 feathery golden gown by Off-White, 2023 latex corset by Mugler, and 2025 full-body gold chainmail by Courrèges all generated buzz but little controversy. The 2026 look, however, marked a sharp departure: instead of pure spectacle, it carried a specific reference that required knowledge of current events. Whether that was a brilliant marketing move or a cultural misstep remains a matter of debate.

Art versus Heritage: The Debate Continues

The days following the Met Gala saw think pieces from The New York Times (which questioned whether fashion can ever responsibly reference real crime) to Le Monde (which argued that the French government should issue a statement condemning the glamorization of the theft). On French talk shows, commentators pointed out that the stolen jewels are not just any artifacts—they are tied to the identity of the Republic itself, as many were museum pieces on loan from the state. The fact that a major luxury house and a global influencer used them for a party in New York was seen by some as a form of „cultural colonialism.“

Meanwhile, Schiaparelli's sales figures reportedly spiked after the event, particularly in the United States, suggesting that controversy did not harm business. Roseberry defended the collection in a rare statement: „We are not celebrating the crime. We are using fashion to process an event that has captivated the world. The agony of the loss and the ecstasy of creation are intertwined. That is the core of the collection.“

What Comes Next for the Missing Jewels?

As the Met Gala fades from the news cycle, the actual investigation into the Louvre heist continues. French police, with assistance from Europol and Interpol, have interviewed dozens of suspects and conducted raids across Europe. The jewels are believed to have been smuggled out of France, possibly to Asia or the Middle East. Their value is not only monetary but symbolic; they are considered part of the French national treasure and cannot be legally traded. The Schiaparelli reference has, unintentionally, kept the story alive in the global press, which may aid in recovery—or encourage private sales on the black market. The fashion world now watches to see if any institution will acquire the Schiaparelli piece for a museum, further complicating the narrative.

Other Notable Met Gala Looks of 2026

While Jenner dominated headlines, other attendees also made statements. Rihanna arrived in a dramatic Balenciaga gown with a cascading diamond waterfall, Beyoncé wore a sheer Givenchy dress covered in sapphire embroidery by Lorraine Schwartz, and Sabrina Carpenter chose a vintage-inspired Dior number with a ten-carat diamond choker. All these looks celebrated luxury, but none carried the same geopolitical charge. The contrast underscored the unique nature of Jenner's choice: she entered the intersection of fashion, crime, and national heritage, a space that few have dared to tread.

Architecture of the Schiaparelli Piece

Let's examine the craftsmanship more deeply. The necklace features a series of interlocking silver-plated links that mimic the gothic filigree of the original 19th-century pieces. The earrings are seven-inch cascades of chandeliers, each terminating in a teardrop pearl. The entire set weighs close to three kilograms and required three fitters to adjust on Jenner. The decision to avoid real gemstones was intentional: Roseberry wanted the pieces to feel „aged, like the ghosts of the originals.“ He used a chemical patination process to give the silver a dark, tarnished finish—a look that evokes decades of storage in a velvet box. The effect is beautiful but unsettling, as if the jewelry has just been recovered from a dusty vault.

The Role of the Louvre in French Identity

The Galerie d'Apollon, where the stolen items were displayed, is one of the most opulent spaces in the Louvre, originally built to house Louis XIV's crown jewels. The gallery itself is a masterpiece of decoration, with paintings by Eugène Delacroix and a gilded ceiling. The theft of the jewels from such a symbolic location deepened the sense of violation. For context, the last major theft from the Louvre was in 1911, when the Mona Lisa was stolen—a piece that was recovered two years later. The Crown Jewels have not been recovered, and their loss remains a fresh wound.

Luxury Houses and Moral Boundaries

Fashion houses often test boundaries, but this incident may set a precedent. Will we see more collections inspired by true crime? Already, rumors circulate that Balenciaga is working on a series based on the Isabella Stewart Gardner Museum heist. Scholars warn that if luxury brands continue to commodify trauma, the line between commentary and exploitation will vanish. Kylie Jenner's Schiaparelli look may be remembered as the moment when the Met Gala stopped being merely a party and became a platform for geopolitics, for better or worse.

Source: Melty News

Unveiled on April 7, Mythos Preview was described by Anthropic as a model that "performs strongly across the board, but it is strikingly capable at computer security tasks." According to the company, the model can find and exploit zero-day vulnerabilities in every major operating system and web browser, including subtle and difficult-to-detect flaws. One notable exploit involved a patched 27-year-old vulnerability in OpenBSD, demonstrating the model's ability to leverage even obscure historical weaknesses.

Capabilities and Real-World Implications

The model's exploit capabilities were highlighted through several examples. In one case, Mythos Preview autonomously wrote a web browser exploit that chained together four separate vulnerabilities, executing a complex JIT heap spray to escape both renderer and operating system sandboxes. It also obtained local privilege escalation exploits on Linux and other systems by exploiting race conditions and KASLR-bypasses. Furthermore, it produced a remote code execution exploit for FreeBSD's NFS server, granting full root access to unauthenticated users by splitting a 20-gadget ROP chain over multiple packets.

Anthropic emphasized that these security capabilities emerged as a "downstream consequence" of improving the model's general code and reasoning abilities, rather than being an explicit development goal. The company noted, "The same improvements that make the model substantially more effective at patching vulnerabilities also make it substantially more effective at exploiting them." This dual-use nature is a central challenge for any organization developing advanced AI systems.

For cybersecurity professionals, the implications are significant. Vulnerability discovery and exploitation have traditionally required deep expertise, patience, and manual analysis. If an LLM can automate much of this process, the speed at which both defenders and attackers can operate will increase dramatically. Defenders might patch vulnerabilities faster, but attackers could weaponize them with equal speed, potentially outpacing the ability of organizations to respond.

Project Glasswing: A Defensive Shield

In anticipation of potential misuse, Anthropic introduced Project Glasswing, a collaborative initiative involving major technology companies such as Apple, Amazon Web Services, Microsoft, Palo Alto Networks, and CrowdStrike. The project aims to "reshape cybersecurity" by deploying Mythos Preview for defensive purposes. Anthropic has extended access to more than 40 organizations, allowing them to scan first-party and open source systems for vulnerabilities and to develop patches.

Lee Klarich, chief product and technology officer of Palo Alto Networks, described early results as "compelling" in a public statement. Beyond access, Anthropic committed $100 million in Mythos Preview usage credits to Project Glasswing and $4 million in direct donations to open source security organizations. These investments signal a serious effort to use the model to strengthen global security rather than undermine it.

However, the initiative also raises questions about equity and control. Only a select group of partners can use the model, creating an asymmetry between those who have access and those who do not. Smaller organizations and independent researchers—who often discover critical vulnerabilities—remain excluded unless they are part of the project. This has led to concerns that the model could entrench a security divide where large corporations benefit while smaller entities fall further behind.

Expert Perspectives on Risk and Control

Industry analysts have offered mixed reactions to Anthropic's approach. One senior analyst noted that the introduction of Mythos Preview is partly good public relations, as it positions Anthropic as leading a transformative shift in cybersecurity. It also draws attention to the vulnerability detection gaps that have persisted for decades. However, the analyst warned, "It's a race for defenders to remediate and patch before other AIs, in the wrong hands, discover these zero-days and rapidly write exploits."

The controls in place are designed to restrict access, but no system is foolproof. A principal solution architect at a security firm pointed out that because there is no clear answer for how these tools can stay out of attacker hands, defenders should assume the capability will proliferate. He recommended investing in detection rather than just prevention, identifying behavioral signatures of AI-assisted exploitation, and adopting zero-trust architecture along with aggressive patching cycles.

Another security expert observed a deeper truth: "No one can ever keep anything 100% out of attackers' hands. The best that can be done is to make it more difficult for them to get access to it." This echoes the long-standing reality of cybersecurity—any tool, from penetration testing frameworks to vulnerability scanners, can be repurposed by malicious actors. The question is not whether abuse will occur, but how quickly and effectively defenders can adapt.

Skepticism and the Need for Independent Verification

Despite the impressive claims, some observers urge caution. Since Anthropic controls both the model and the narrative, independent replication of the reported exploits is impossible when the model is not publicly available. A security executive emphasized, "Until independent researchers with access can run their own evaluations, healthy skepticism is the appropriate posture. This is, frankly, another consequence of the restricted access model: the claims can't be tested, so they can't be fully trusted or refuted."

Anthropic did not respond to requests for statistics regarding false positives or error rates, leaving some questions unanswered. The lack of transparency may fuel uncertainty, but it also reflects the delicate balance between openness and safety. Making the model widely available could lead to immediate misuse, while keeping it locked down stifles independent validation.

The broader cybersecurity community is now watching closely. Mythos Preview represents a significant step forward in AI-driven security, but it also presages a future where both offensive and defensive capabilities become dematerialized and accelerated. Organizations must prepare for a landscape where vulnerability discovery and exploitation happen at machine speed, and where the line between friend and foe is drawn not by tools but by intent and governance.

Source: Dark Reading News

According to analysis from cloud security vendor Wiz, the campaign began in mid-March 2026 and unfolded in multiple waves. The attacker created six distinct GitHub accounts to carry out the activity, all linked to a single threat actor. While fewer than 10% of the exploitation attempts succeeded, the attacker still managed to compromise at least two NPM packages, demonstrating the potential for significant downstream damage.

How the Attack Worked

The prt-scan campaign specifically targeted repositories that used the pull_request_target trigger in GitHub Actions. This trigger automatically runs workflows in the main repository whenever a pull request is submitted, even if the request originates from an untrusted fork. The trigger provides the workflow with full repository permissions and access to stored secrets, making it a valuable target for attackers seeking to steal API keys, cloud credentials, or other sensitive data.

The attacker's playbook was systematic. First, they scanned GitHub for repositories using the vulnerable trigger. They then forked those repositories, created new branches, hid malicious code inside what appeared to be routine updates, and submitted pull requests designed to automatically execute the malicious workflow. The payloads were crafted to exfiltrate credentials, environment variables, and other secrets stored in the repository.

Wiz researchers noted that the attack chain began with a testing phase from March 11 to March 16, during which the attacker opened only 10 malicious pull requests. After a nearly two-week hiatus, the activity resumed on April 2 with dramatically increased velocity. Over a 26-hour period, the attacker opened approximately 475 pull requests — a pace that strongly suggested the use of AI-enabled automation to scale the operation.

AI Automation in Cybersecurity Threats

The prt-scan campaign is notable not only for its scale but also for its use of artificial intelligence. It marks the second such campaign in recent months where a threat actor appears to have leveraged AI to automate the targeting of widespread GitHub misconfigurations. The first, dubbed "hackerbot-claw," occurred in late February and targeted high-profile repositories with more precision but shorter duration.

Wiz observed that the prt-scan attacker used AI to generate hundreds of pull requests with payloads that appeared sophisticated on the surface. However, a deeper analysis revealed flawed implementation. The attacker demonstrated a lack of understanding of GitHub's permissions model, using techniques that would rarely work in practice. For example, the payloads attempted multi-phase credential theft but were filled with logical inconsistencies that an experienced developer would immediately recognize as suspicious.

Despite these flaws, the campaign still achieved a roughly 10% success rate, translating into dozens of compromises. The majority of successful attacks targeted small hobbyist projects, and the stolen credentials were largely ephemeral GitHub tokens rather than production cloud keys. Still, the potential for more damaging outcomes remains high as attack tools improve.

Growing Threat to Open Source Ecosystems

The open source software supply chain has become an increasingly attractive target for attackers. GitHub, as the world's largest host of open source code, is a prime vector. The pull_request_target trigger is a well-documented feature, but its misconfiguration has been repeatedly highlighted by security researchers as a risk. Many project maintainers inadvertently enable this trigger on untrusted pull requests without implementing proper restrictions, making their repositories vulnerable.

The use of AI to automate exploitation represents a significant escalation. Previously, launching a large-scale supply chain attack required substantial manual effort and coordination. Now, low-sophistication attackers can use AI tools to scan, fork, and submit malicious pull requests at scale, targeting hundreds or thousands of repositories in a fraction of the time. This democratization of attack capability poses a serious challenge to the security of the open source ecosystem.

Wiz has published indicators of compromise (IoCs) for the prt-scan campaign to help organizations detect and block similar activity. The security vendor urged organizations to harden their GitHub environments by avoiding the use of pull_request_target on untrusted forks, restricting workflow permissions, and implementing proper secret validation.

GitHub itself has provided guidance on secure configuration of Actions workflows. The company recommends using the pull_request trigger for untrusted forks instead of pull_request_target, or combining the latter with explicit permission checks and sanitization of inputs. Many security tools now also scan for this misconfiguration automatically.

The prt-scan campaign highlights the ongoing arms race between attackers and defenders in the software supply chain. While the specific campaign was limited in its success, it serves as a warning that AI-augmented automation is making it easier for threat actors to probe for weaknesses at unprecedented scale. Organizations that rely on open source components must remain vigilant, regularly audit their CI/CD pipelines, and stay informed about emerging attack techniques.

The discovery of this campaign was made on April 2 by security researcher Charlie Eriksen of Aikido Security, with subsequent deeper analysis conducted by Wiz. The campaign primarily targeted repositories configured with the vulnerable trigger, but also demonstrated that even smaller projects can be entry points for broader supply chain infiltration. As AI tools continue to evolve, the security community expects to see more such campaigns, potentially with greater sophistication and success rates.

Source: Dark Reading News

The Evolution of Software Supply Chain Security

Software supply chain attacks have become a primary vector for cybercriminals seeking to infiltrate organizations. Recent high-profile incidents, such as the hijacking of the tj-actions/changed-files GitHub Action that leaked secrets from over 23,000 repositories, underscore the urgency of protecting CI/CD pipelines. These pipelines are often the most privileged parts of an organization's infrastructure, with write access to repositories, deployment credentials, and signing keys. Attackers increasingly target third-party GitHub Actions, container images, and AI agent skills, making manual security updates unsustainable.

Chainguard Factory 2.0 addresses these challenges by introducing a controller/reconciler model inspired by Kubernetes architecture. Instead of relying on fragile, throwaway scripts that require constant updates, the platform uses the open-source DriftlessAF agentic framework to continuously reconcile open source artifacts across containers, libraries, GitHub Actions, and agent skills. This ensures that approved artifacts are always patched and up-to-date without human intervention.

Key Components of Factory 2.0

Factory 2.0 includes several new features designed to harden the software supply chain at multiple levels. The first is Chainguard Actions, a hardened catalog of GitHub Actions and CI/CD workflows. These are built and continuously maintained by Chainguard, providing secure, drop-in replacements for the top 100 actions from the GitHub marketplace. Dan Lorenc, co-founder and CEO of Chainguard, explained at the conference that these actions are "secure by default" and allow developers and AI agents to work faster without introducing supply chain risk. Patrick Donahue, chief product officer, added that Chainguard Actions detect and remediate unsafe code in third-party actions, reducing the likelihood of compromise.

The second component is Chainguard Agent Skills, a catalog of continuously hardened third-party AI agent skills. These are small markdown instruction sets that enable AI agents to perform specific tasks such as browser automation, PDF processing, SEO checking, web design, and code quality reviews. By using vetted skills from this catalog, organizations can prevent malicious skills from being installed on developer machines—a tactic that has already been seen in attacks against OpenClaw registries, where adversaries uploaded skills that installed the Atomic macOS Stealer.

The third component is Chainguard Guardener, an AI agent that automates the migration and maintenance of trusted open source artifacts. The initial release automatically converts legacy Dockerfiles into minimal, zero-CVE Chainguard container images. Ed Sawma, a product VP at Chainguard, described Guardener as an agent placed in customer environments to enable automated use of Chainguard images. Future updates will extend this capability to other configuration scripts, reducing the manual effort required to maintain secure images.

How Factory 2.0 Differs from the Original Platform

The original Chainguard Factory relied on event-driven rules and complex automations that required frequent updates as upstream changes occurred. Factory 2.0 fundamentally rearchitects the platform around a reconciliation model. Instead of triggering actions in response to events, the new control plane continuously monitors the desired state of artifacts and reconciles any drift. This approach is more durable and scalable, especially for large organizations managing thousands of containers and workflows.

The DriftlessAF framework is at the heart of this reconciliation. It allows the platform to detect when an artifact has fallen out of sync with its approved version and automatically apply updates. This eliminates the need for manual patching and reduces the risk of human error. The framework is open source, enabling the community to contribute and extend its capabilities.

Industry Implications and Adoption

Experts like Adeel Saeed, CISO of Kyndryl, have noted that Factory 2.0's automation will drive adoption of secure software maintenance. Previously, organizations had to manually download images and place them in repositories like Artifactory, a process Saeed described as "very manual." With Chainguard Actions and Guardener integrated into Git repositories, the entire process can be automated, making it easier for enterprises to maintain a hardened supply chain.

The timing of Factory 2.0's release is critical. Threat actors are continuously developing new methods to inject malware into supply chains. Recent attacks have targeted CI/CD pipelines by compromising GitHub Actions, and others have focused on AI agent skills. By providing hardened catalogs of actions and skills, Chainguard reduces the attack surface for organizations that rely heavily on third-party components. The platform also addresses the growing use of AI coding assistants, which may inadvertently pull unvetted skills from public registries.

Technical Architecture of the Reconciler Model

Under the hood, Factory 2.0 implements a controller/reconciler pattern similar to that used in Kubernetes. Each type of artifact—containers, libraries, GitHub Actions, and agent skills—has a dedicated controller that runs continuously. When a change is detected upstream (e.g., a new version of a container image or a security patch for a GitHub Action), the reconciler updates the hardened copy and propagates it to all downstream customers. This ensures that customers always consume the latest secure version without needing to trigger pipeline reruns or manual approvals.

Donahue explained that the platform also performs runtime hardening, scanning for known vulnerabilities (CVEs) and misconfigurations before artifacts are published. For GitHub Actions, this includes static analysis of the action code to detect malicious patterns, such as secret exfiltration or unauthorized network calls. For agent skills, the platform verifies that the instructions do not contain commands that could lead to data theft or privilege escalation.

Expanding Beyond Containers and Actions

Chainguard initially built its reputation on securing container images, but Factory 2.0 broadens the scope to include libraries, CI/CD actions, and AI agent skills. This expansion reflects the reality that modern software supply chains extend beyond containers. Developers use hundreds of dependencies, and AI agents are increasingly being integrated into development workflows. By providing a unified platform for hardening all these artifacts, Chainguard aims to become a central pillar of enterprise DevSecOps.

The company also plans to release additional updates to the Guardener agent, enabling it to handle other configuration scripts such as Kubernetes manifests, Terraform files, and Ansible playbooks. This will allow organizations to automate the conversion of legacy infrastructure-as-code into hardened, CVE-free versions.

Chainguard's strategy of combining a controller/reconciler model with AI-powered automation sets Factory 2.0 apart from other supply chain security tools. While many competitors focus on vulnerability scanning or policy enforcement, Factory 2.0 actively maintains and updates artifacts in a continuous loop, reducing the operational burden on security teams. As software supply chain attacks continue to rise, tools like Factory 2.0 that automate hardening and reconciliation will become essential for organizations seeking to protect their development pipelines without sacrificing velocity.

Source: Dark Reading News

The new assistant, first available in Figma Design, allows designers to type a prompt and watch as the AI produces elements on the canvas in real time. It can handle multiple agents simultaneously, each tasked with different aspects of a project, effectively adding AI collaborators to the same workspace where human teammates already operate. Figma claims that the underlying models have been fine-tuned specifically for design tasks, giving the agent an understanding of layout, components, and visual hierarchy that generic large language models lack.

A New Era of Collaborative Design

For years, Figma has been the environment where design teams come together to create interfaces, prototypes, and design systems. Its multiplayer architecture, which allows multiple people to work on the same file simultaneously, has been a key differentiator. Now, Figma is extending that collaborative ethos to machines. The AI assistant is not a separate panel or plugin; it is a participant on the canvas, able to see the same frames, layers, and components that human designers see.

“Teams can now collaborate with agents on the multiplayer canvas to test out ideas, visualise edge cases, and refine concepts together without over-indexing on the more tedious parts,” said Loredana Crisan, Figma’s chief design officer. Crisan joined Figma from Meta in 2025 after nearly a decade leading product and design teams across Messenger, Instagram, and Meta’s generative AI initiatives. Her arrival signaled a deeper commitment to integrating AI into the heart of Figma’s design experience.

The AI Assistant in Action

The assistant is designed to handle a wide range of tasks. A user might ask it to generate a mobile login screen with specific branding colors, or to create a set of icons based on a theme. It can also edit existing designs by altering layouts, adjusting typography, or reconfiguring component hierarchies. Because multiple agents can run simultaneously, one agent could be prototyping a new feature while another refines the visual language of an existing component. The agents operate in real time, and their work appears directly on the canvas for human collaborators to review, modify, or build upon.

Figma’s decision to fine-tune its own models for design work is crucial. Many AI tools that generate images or layouts rely on general-purpose models that lack understanding of design principles such as spacing, alignment, and component consistency. Figma claims its assistant has been trained on millions of design files and interactions, giving it a nuanced grasp of how real design projects are structured. This means the AI can produce outputs that fit naturally into existing design systems, reducing the friction often associated with AI-generated content.

Strategic Context: Partnerships and Acquisitions

The launch of Figma’s own AI assistant comes after a period of intense activity in the company’s AI strategy. In February 2026, Figma struck back-to-back partnerships with Anthropic and OpenAI, integrating Claude Code and Codex into its design-to-development pipeline through MCP. Those integrations allowed developers to convert interfaces into editable Figma frames, or hand off designs to coding agents for production-ready implementation. The new built-in assistant adds a different dimension: instead of bridging code and design, it makes AI a native participant in the design process itself.

That push has been underpinned by acquisitions. In October 2025, Figma bought Weavy, a Tel Aviv-based startup that had built a node-based AI canvas combining multiple generative models with professional editing tools. The deal, reportedly valued at roughly $200 million, became Figma Weave—a product that contributed to Figma’s strong first-quarter results. Figma reported Q1 2026 revenue of $333.4 million, a 46 percent increase year over year, with its net dollar retention rate climbing to 139 percent, the highest in over two years. The AI credit monetization from Figma Weave helped drive that growth.

The Competitive Landscape

Figma is not alone in the race to bring AI into design. Canva, which now claims 220 million global users, launched its AI 2.0 platform in March 2026, built on a proprietary foundation model designed specifically for graphic design. Adobe’s Firefly technology holds 41 percent business adoption, and the company continues to embed AI into its Creative Cloud suite. Meanwhile, a crop of AI-native startups—including Flora, Krea, and Dessn—are chasing the same audience of designers who want to move faster without sacrificing craft. Google also unveiled Pics at I/O 2026, an AI design tool built directly into Workspace that generates graphics from text prompts.

Figma’s advantage, if it has one, is the canvas itself. More than 690,000 paying teams already use Figma as their collaborative workspace. The multiplayer architecture that made Figma dominant now doubles as the natural environment for AI agents to operate in. Where competitors are building AI tools that work on design, Figma is building AI tools that work within design—sitting alongside human teammates on the same infinite canvas. This distinction could prove critical as design teams increasingly look for ways to integrate AI without disrupting established workflows.

Implications for Designers

The introduction of a native AI assistant raises questions about the role of human designers in an increasingly automated process. Figma is positioning the tool as an enhancer, not a replacement. By handling repetitive tasks such as generating multiple variations, resizing elements, or applying design system constraints, the AI frees designers to focus on higher-level strategy, creativity, and problem-solving. The ability to run multiple agents also means teams can explore many design directions simultaneously, accelerating the ideation phase.

However, there is a learning curve. Designers will need to develop new skills in prompt engineering and understanding how to guide the AI to produce desired outcomes. Figma has designed the assistant to be intuitive, with a natural language interface that feels similar to chatting with a colleague. Yet effective use will likely require an understanding of how the model interprets design language and constraints.

Technical Foundation and Future Plans

Figma has not disclosed detailed specifications about the models powering its assistant, but the company has confirmed that they are fine-tuned on a large corpus of design data. The assistant leverages Figma’s existing component system, meaning it can respect design tokens, variants, and auto layout rules. This integration ensures that AI-generated designs are immediately editable and consistent with the project’s design system.

Looking ahead, Figma plans to extend the AI assistant to its other products over time, including FigJam, its whiteboarding tool, and eventually to Figma’s developer-oriented offerings. The company has also signaled a desire to pull design and code even closer together inside its apps, potentially allowing the AI to generate not just visual designs but also production-ready code snippets. This would further blur the line between design and development, a trend that Figma has been championing since its early days.

Figma’s AI assistant is currently available in beta for a limited set of users, with a broader rollout expected later in 2026. The company has not announced pricing for the assistant beyond saying that it will be included in existing Figma subscriptions, with additional usage credits available for heavy users. As the AI becomes more capable, Figma will likely refine its monetization model, following the path it has already taken with Figma Weave.

In summary, Figma’s decision to build its own AI assistant represents a natural evolution of its platform. By embedding AI directly into the collaborative canvas, the company is betting that the future of design lies in human-machine partnership, where both work side by side on the same infinite surface. The success of this bet will depend on execution, user adoption, and the ability to compete in a rapidly accelerating market. For now, Figma has made its stance clear: the canvas that changed how designers collaborate is now changing how they collaborate with machines.

Source: TNW | Apps News

The AION partner list reads as a near-complete roll-call of the French AI stack. Named backers include GPU and chip-design specialists VSORA and SiPearl, model labs Kyutai and H Company, model-distribution platform Hugging Face, IT-services group Sopra Steria, consultancy Artefact, Atos’s compute subsidiary Eviden Bull, and developer-tooling company ZML. The consortium also draws operational support from GENCI and Inria, co-leaders of the existing AI Factory France EuroHPC project, with hosting through Opcore, Iliad’s data-centre joint venture.

The EU InvestAI Facility and Gigafactory Programme

The EU programme the bid sits inside is the InvestAI Facility, a €20 billion envelope announced earlier this year to underwrite up to five gigafactories across the bloc. The European Commission received 76 expressions of interest in the initial sounding round, with Spain, Germany, the Netherlands, Italy, Finland and Portugal among the member states co-financing the programme. Telefonica is preparing the final Spanish bid; the formal call window was deferred from late 2025 to the first half of 2026 to give consortia time to assemble multi-billion-euro capital structures. The EuroHPC Joint Undertaking, the body running the selection process, has not yet publicly named the bidder pool or set a final decision date. The next visible proof point will be the EuroHPC JU’s shortlist announcement, expected before the end of the year.

On the disclosed numbers, AION’s $10 billion capital commitment matches Iliad chair Xavier Niel’s longstanding framing that France needs to outspend, not match, on AI infrastructure if it intends to keep pace with the US and Chinese build-outs. Iliad has invested €20 billion in European infrastructure over the past decade, the Scaleway announcement notes; the AION figure puts roughly half that commitment into a single facility. The 288,000-H100-equivalent target is positioned as the largest single GPU cluster outside the US hyperscalers and the Microsoft-OpenAI Stargate footprint.

French AI Infrastructure Landscape

AION is not the only French-flagged AI infrastructure project the bid is competing with for capital. The MGX-Bpifrance-Nvidia-Mistral 1.4GW Paris-area campus, announced in 2025, is the parallel programme; Mistral is separately raising debt and equity for its own Sweden and Paris data-centre footprint. AION’s stated differentiator is the open-source-and-public-private framing: GENCI and Inria participation positions the facility as part of the European public-research compute infrastructure, in contrast to the more commercially-driven MGX-Mistral programme. This dual-track approach reflects broader tensions within the French AI ecosystem between private-led hyper-scale projects and state-anchored initiatives that prioritize open research and sovereign control over AI capabilities.

The strategic context the announcement leans into is the ongoing pressure on European AI sovereignty. GPU-as-a-service offerings from US providers have continued to dominate European frontier-AI procurement; OpenAI’s pause on its UK Stargate site over energy costs and regulatory uncertainty has produced a window in which a French-only proposition can credibly claim both available power capacity (France’s low-carbon grid) and a sovereign-software stack (SiPearl and Eviden hardware, Hugging Face and Kyutai software). AION’s pitch is, on Scaleway CEO Damien Lucas’s framing, that “Europe can no longer afford to outsource the foundations of its AI future.”

Key Details and Next Steps

The Bloomberg report did not name the specific French site under consideration, the capital-stack breakdown between Iliad equity, EU grants, member-state co-financing and private debt, the construction timeline, or the formal procurement-decision date the Commission is now working to. Telefonica’s Spanish bid and German and Dutch consortia are positioned as the most credible competing proposals. AION’s position alongside the wider Google-Blackstone $25 billion TPU-cloud joint venture and the comparable US infrastructure announcements will become a public-market proxy for whether Europe’s gigafactory programme is ultimately operating at the same order of magnitude as the US private-sector build-out.

The EuroHPC JU has not yet set a final decision date. The formal call window was deferred from late 2025 to the first half of 2026 to allow consortia time to assemble multi-billion-euro capital structures. The shortlist announcement is expected before the end of the year, based on the published formal-call cadence. French policymakers and industry leaders are closely watching the process, as the outcome will determine whether France secures a flagship AI infrastructure that could serve as a hub for European research, development, and commercial deployment. The AION bid, with its emphasis on open-source collaboration and public-private partnerships, aims to differentiate itself from the more secretive, commercial-focused US projects, while still attracting the scale of investment needed to compete globally.

Beyond the immediate bid, the French AI ecosystem is undergoing rapid expansion. In addition to the MGX-Mistral campus, other initiatives include the Jean Zay supercomputer upgrade, the European Processor Initiative involving SiPearl, and growing investments in data-centre cooling technologies to cope with high-density GPU clusters. France’s low-carbon nuclear-powered electricity grid provides a unique advantage for AI workloads that require substantial and reliable power, a factor that both AION and the MGX consortium are highlighting. However, regulatory hurdles, grid connection timelines, and local community opposition to large-scale data centres remain potential obstacles that the consortia must navigate.

The InvestAI Facility itself is part of a broader EU push to reduce dependency on non-European AI infrastructure. The European Commission has earmarked additional funds for AI research, skill development, and regulatory frameworks, including the AI Act. The gigafactories are intended to host large-scale training and inference workloads for European startups, universities, and public-sector organizations. If AION succeeds, France would gain a strategic asset that could attract AI talent and investment, potentially shifting the balance of AI development more toward Europe. The upcoming months will be critical as the EuroHPC JU evaluates the competing bids and decides which member states will host the five gigafactories, with the French proposal standing out for its size, scope, and collaborative partner network.

Source: TNW | Eu News

Understanding the Applied AI Lab Framework

The term "Applied AI Lab" is crucial to interpreting this announcement correctly. OpenAI is not establishing a frontier research lab in Singapore—the kind that pushes the boundaries of fundamental AI capabilities. Instead, the new facility is structured as a deployment and partnerships unit. Its mandate is to take OpenAI’s existing model lineup and apply it within Singapore’s specific national policy framework. The Singapore government emerges as the most significant single customer and partner, aligning the lab’s work with the country’s AI Mission priorities in public service, finance, healthcare, and digital infrastructure. This lab will operate alongside the regional commercial office that OpenAI opened in Singapore in 2024.

Singapore has spent the past five years positioning itself as the most attractive Western-aligned hub in Southeast Asia for AI infrastructure and frontier-model deployment. The Monetary Authority of Singapore has been one of the most engaged Asian regulators on cybersecurity tracks relevant to frontier AI, such as the Anthropic Mythos framework. Moreover, Singapore’s public-sector AI commitments, exceeding $7 billion since 2024, have created the cleanest single-jurisdiction procurement pipeline in the region. OpenAI’s choice of Singapore over other major Asian cities like Tokyo, Seoul, Sydney, or Bangalore reflects this procurement-readiness gradient as much as any technology considerations.

Geopolitical Context and Strategic Positioning

The geopolitical backdrop gives this move its true scale. The recent Trump-Xi Beijing summit confirmed that US-China AI policy is now being negotiated at the head-of-state level, with chip export controls and AI guardrails on the same agenda. In this context, Singapore serves as a diplomatically neutral surface where Western frontier-AI companies can deploy at scale without the political exposure that would attach to a launch in Tokyo or Seoul. This neutral positioning is a deliberate and carefully cultivated advantage for Singapore.

At the same time, Chinese model-lab competition from DeepSeek, Moonshot’s Kimi, and Alibaba’s Qwen has made the Asia-Pacific deployment race more crowded than it was eighteen months ago. OpenAI’s Singapore lab is the structural answer to that competitive density. By establishing a permanent presence in Singapore, OpenAI gains a launching pad to serve the broader Southeast Asian market and beyond, while mitigating the risks of direct exposure to Chinese regulatory and political pressures.

Notably, Singapore also signed a parallel AI partnership with Google at the same ATxSG event. The simultaneous announcements signal a deliberate Singaporean strategy: lock in concurrent partnerships with the two largest Western frontier labs so that the city-state is not architecturally dependent on either. This multi-vendor approach mirrors strategies used by large institutional investors, such as Australia’s largest pension funds, which have explicitly signaled multi-vendor frontier-model engagement as a hedge against single-vendor concentration risk.

Economic and Operational Considerations

It is important to recognize that Singapore, on its own terms, is not a large enough domestic market to justify a 200-person frontier-AI applied lab purely on commercial logic. The lab’s economic case rests on the city-state functioning as the regional hub for OpenAI’s Southeast Asia and broader APAC presence. Singapore-based engineers will service customers in Indonesia, Vietnam, the Philippines, Malaysia, Thailand, and—more sensitively—markets like Hong Kong, where direct US-AI-company presence is structurally difficult due to regulatory and political complexities.

Whether this hub-and-spoke model lands at scale will depend on how quickly the regional customer base materializes around the Singapore base. The city-state’s robust infrastructure, skilled workforce, and pro-business environment provide a solid foundation, but the success of the model ultimately hinges on demand from neighboring economies and the ability to navigate diverse regulatory landscapes.

OpenAI did not disclose the specific Singapore neighborhoods or facilities the lab will occupy, nor the construction and hiring timeline beyond the broad "next few years" timeframe. The proportion of the S$300 million commitment that is operating expense versus capital expenditure also remains undisclosed. Furthermore, Singapore’s Ministry of Digital Development and Information has not yet published a project-level breakdown of how the lab’s work will be coordinated with the country’s existing Smart Nation programs.

Implications for the Regional AI Landscape

The establishment of OpenAI’s applied AI lab in Singapore is likely to accelerate the adoption of AI technologies across Southeast Asia. Local enterprises and government agencies will gain easier access to cutting-edge AI models and expertise, potentially spurring innovation in sectors like finance, healthcare, and digital services. However, the move also raises questions about data sovereignty, talent competition, and the potential for increased dependency on Western AI platforms.

Singapore’s proactive engagement with multiple AI providers positions it as a critical node in the global AI ecosystem. By hosting labs from both OpenAI and Google, the city-state is effectively creating a competitive marketplace for AI services within its borders. This could drive down costs and improve service quality for end users, but it also requires careful governance to ensure that AI deployment aligns with Singapore’s ethical and security standards.

The next visible proof point will be the first set of named Singaporean government deployments under the new lab, which, according to the press release, are scheduled to begin shortly after staffing ramps. These deployments will serve as a tangible demonstration of the lab’s impact and could set a precedent for similar partnerships across the region. As OpenAI and other frontier labs continue to expand internationally, Singapore’s model of neutral, multi-vendor AI collaboration may become a template for other nations seeking to balance innovation with geopolitical prudence.

Source: TNW | Openai News

The funding was led by Parkway Venture Capital, and the investor list reads like a who’s who of the semiconductor and cloud computing ecosystem. Nvidia, AMD Ventures, Intel Capital, and Qualcomm Ventures all participated, alongside Salesforce Ventures, Brookfield, ARK Invest, Greycroft, Prime Movers Lab, Align Ventures, and Tamarack Global. The presence of multiple chipmakers on the same cap table signals a strategic alignment that could give Hark an edge in securing manufacturing capacity and technical partnerships.

Adcock founded Hark in late 2025 with $100 million of his own money, drawn from the proceeds of his prior entrepreneurial successes. His track record includes co-founding the recruiting marketplace Vettery, which was acquired by Adecco for $100 million; founding electric aircraft maker Archer Aviation, which went public via a SPAC merger in 2021; founding humanoid robotics company Figure, where he remains CEO; and founding school-security firm Cover. He also serves as principal at Hark, indicating deep personal commitment and financial stake in the venture.

The Vision Behind Hark

What Hark is actually building remains less clearly defined than the capital it has raised. The company describes itself as developing a “personal AI platform” that combines in-house foundation models, software, and custom hardware with new user interfaces. Unlike many AI startups that focus on a single layer of the technology stack — such as chips, models, or applications — Hark is attempting to vertically integrate all three from the outset.

According to a press release issued in March, Hark intends to release its first multimodal models this summer. These models are expected to process and generate text, images, audio, and video, forming the core of the platform that will eventually drive a dedicated hardware device. However, the company has not disclosed specific details about the device’s form factor, target price, launch markets, or customer pipeline.

This level of opacity is not unusual for a company only two months out of stealth and flush with cash. The $700 million Series A provides substantial runway to keep development plans under wraps while the team works toward a market-ready product. Industry observers speculate that the device could take the form of a wearable, a smartphone-like companion, or a stationary home hub — but nothing has been confirmed.

Market Context and Cautionary Tales

Hark is entering a category that has seen high-profile failures. Humane’s AI Pin, launched in 2024, became a cautionary tale when early reviews highlighted poor performance, overheating, and limited utility. Rabbit’s R1 device faced similar criticism, with users questioning its value proposition. Even Apple, which possesses unparalleled hardware distribution and integration capabilities, has spent the past year iterating on its on-device AI offering, with no clear breakthrough yet.

The challenge for hardware AI companies is twofold: they must deliver compelling software that justifies a new device, and they must ensure the hardware itself is reliable, affordable, and useful in daily life. Many startups have stumbled on either or both fronts. Hark’s approach of designing the model stack and silicon together from day one aims to avoid the integration issues that have plagued rivals.

Adcock’s track record includes shipping complex hardware at scale. Archer Aviation successfully brought electric vertical takeoff and landing (eVTOL) aircraft from concept to certification and production, while Figure’s humanoid robots have been deployed in real-world warehouse environments. This experience could be a decisive advantage in navigating the hardware development lifecycle, which typically spans several years and involves countless engineering obstacles.

Investor Confidence and Strategic Implications

The participation of Nvidia, AMD, Intel, and Qualcomm in the same funding round is notable because these companies usually compete fiercely in the chip market. Their willingness to invest suggests that Hark’s platform may be designed to be hardware-agnostic at some level, or that each investor sees a unique strategic benefit. For Nvidia and AMD, which supply GPUs for AI training and inference, a partnership with Hark could secure a captive customer for their latest chips. For Intel and Qualcomm, which produce edge processors and mobile SoCs, the investment may be a bet on future on-device AI workloads.

With these investors on the cap table, supply chain allocation — often a binding constraint for AI hardware startups in 2026 — becomes a problem Hark can address more comfortably than its peers. Chip foundries and advanced packaging capacity are scarce, and having the backing of major semiconductor companies could help Hark secure priority access.

However, the round does not solve the fundamental challenge of product-market fit. Several well-funded, well-credentialed teams have entered this space with promising prototypes only to discover that consumers are not willing to adopt a new device category without a clear, everyday use case. Hark’s timeline suggests that its first models will launch within weeks, but the device that translates those models into a sustainable business is still further out.

Adcock’s previous ventures all benefited from careful timing and deep technical differentiation. Archer entered the eVTOL market when regulations and investor interest were aligning; Figure entered humanoid robotics as labor shortages and automation demand surged. If Hark can similarly time its market entry with a compelling product, it could succeed where others have failed. But the path remains uncertain, and the company has yet to prove that its integrated approach yields a system that people actually want to use.

For now, Hark is a billion-dollar paper valuation built on a founder’s reputation, investor confidence, and a broad vision. The next few months — with model releases and hardware announcements — will determine whether that vision can become a reality.

Source: TNW | Investors-Funding News

On named partners, the announcement covers work with ASE and SPIL on next-generation wafer-based 2.5D bridge interconnect technology, alongside other Taiwan-based suppliers AMD has not separately listed in the public release. The technology track filed in the company’s 8-K materials is calibrated to support the Helios platform’s full-rack scale architecture, where AMD has been positioning against Nvidia’s GB200 and GB300 NVL72 systems through the past three quarters.

Strategic Context of the Investment

Chair and chief executive Lisa Su framed the announcement around AI-infrastructure demand. “As AI adoption accelerates, our global customers are rapidly scaling AI infrastructure to meet growing compute demand,” she said in the statement, signaling that the Taiwan-side capacity build is calibrated against a customer pipeline AMD has not separately disclosed. The competitive context, which the announcement does not address directly, is that the Google-Blackstone $25 billion TPU-cloud joint venture and the wider hyperscaler-capex commitments for 2026 have produced a procurement window in which non-Nvidia accelerator suppliers can credibly compete for share if the manufacturing-and-packaging supply chain can keep pace.

Taiwan’s role in the announcement is the structural part. The country’s foundry-and-packaging capacity is the bottleneck for the entire frontier-AI-silicon supply chain, regardless of which US accelerator brand the customer ultimately specifies. AMD’s commitment positions the company alongside Nvidia’s own multi-year TSMC-and-packaging supply commitments at the front of the foundry queue for the H2 2026 and H1 2027 production windows.

Helios Platform Architecture and Market Positioning

The Helios platform represents AMD’s most ambitious attempt to challenge Nvidia’s dominance in the AI data center market. Designed as a full-rack scale system, Helios integrates AMD’s Instinct MI300X and future MI400 series accelerators with high-bandwidth memory and advanced networking. The platform leverages AMD’s Infinity Architecture to create a unified memory pool across multiple GPUs, enabling efficient training and inference for large language models and other AI workloads. By partnering with ASE and SPIL for 2.5D bridge interconnect technology, AMD aims to improve chiplet integration and reduce latency, key factors in achieving competitive performance against Nvidia’s NVLink-connected systems.

The investment in Taiwan is also critical for securing supply of advanced packaging capacity. Currently, TSMC’s CoWoS (Chip-on-Wafer-on-Substrate) packaging is in extremely high demand, with Nvidia consuming a significant portion of the available capacity. AMD’s collaboration with ASE and SPIL, two of the world’s largest semiconductor packaging and testing companies, provides alternative pathways for 2.5D and 3D packaging, reducing reliance on TSMC’s limited capacity. This diversification is essential for AMD to meet the aggressive production timelines for Helios and to avoid the supply constraints that have plagued the industry.

Geopolitical and Supply Chain Implications

The geopolitical overlay is the part neither side of the supply chain addresses directly in the announcement materials. Taiwan’s semiconductor industry is concentrated on the island, which faces ongoing tensions with China. AMD’s $10 billion commitment underscores the critical importance of Taiwan to the global AI supply chain, but also highlights the risks of such concentration. While AMD, like Nvidia, works closely with TSMC for leading-edge chip manufacturing, the packaging partnerships with ASE and SPIL are also heavily based in Taiwan. Any disruption to Taiwan’s operations could severely impact AMD’s ability to deliver Helios on schedule.

The wider Nvidia-alternative compute landscape this announcement sits inside has been active across the past three weeks. Tenstorrent’s takeover conversations with Intel and Qualcomm and Alibaba’s T-Head Zhenwu M890 announcement represent the two visible non-Nvidia compute paths from the US/Western and the Chinese-domestic sides respectively. AMD is the third leg of that stool, the established US-side challenger with the production-line credibility to actually ship into hyperscaler deployments at scale.

Historical Background of AMD’s AI Journey

AMD’s push into AI accelerators began in earnest with the acquisition of Xilinx in 2022 for $49 billion, which brought FPGA and adaptive computing capabilities to the company’s portfolio. The Instinct line of GPUs, originally developed for HPC and supercomputing, was optimized for AI workloads starting with the MI250X and later the MI300 series. The MI300X, launched in late 2023, featured a chiplet design with 12 chiplets totaling 146 billion transistors, making it one of the most complex chips ever built. However, AMD has struggled to gain significant market share against Nvidia, which holds an estimated 80-90% of the AI accelerator market due to its CUDA software ecosystem and decades of optimization.

The Helios platform is AMD’s answer to Nvidia’s DGX and NVL systems. By offering a complete rack-scale solution with optimized software stack (ROCm), AMD hopes to entice hyperscalers and cloud providers to adopt its hardware. The $10 billion investment in Taiwan is a clear signal that AMD is willing to commit the resources needed to compete. The company has also been building partnerships with key customers such as Microsoft, Meta, and Oracle, who have publicly committed to using AMD Instinct GPUs for some of their AI workloads.

Financial and Operational Details

AMD did not disclose the multi-year allocation schedule for the $10 billion-plus commitment, the specific named customer contracts the Helios platform will land in during the H2 2026 deployment window, the per-rack cost economics relative to Nvidia’s NVL72 systems, or the proportion of the Taiwan investment that is opex versus capex. The 8-K filed with the announcement carries the headline figure. Industry analysts estimate that advanced packaging capacity expansion with ASE and SPIL could absorb a significant portion of the investment, while the rest may go toward supply chain infrastructure and joint development agreements.

The commitment is the largest single-country AI-infrastructure commitment AMD has disclosed to date. The next visible proof point will be the first named Helios deployment under the H2 2026 timeline, where the customer logo and the production-shipment volumes will become public. For now, AMD is betting that its architecture, combined with Taiwan’s unparalleled manufacturing ecosystem, can deliver a credible alternative to Nvidia in the race to build the next generation of AI supercomputers.

In terms of competitive dynamics, Nvidia has already announced its own GB300 NVL72 system with liquid cooling and next-generation HBM4 memory, expected to arrive in 2025-2026. AMD’s Helios will need to demonstrate performance parity or superiority in key metrics such as memory bandwidth, interconnect speed, and power efficiency. The partnership with ASE and SPIL on 2.5D bridge interconnect technology could give AMD an edge in chiplet integration, allowing for more flexible and cost-effective designs compared to Nvidia’s monolithic approach.

Taiwan’s broader semiconductor ecosystem also includes companies like MediaTek, UMC, and numerous materials and equipment suppliers. While AMD’s announcement specifically names ASE and SPIL, it is likely that other Taiwan-based companies will benefit from the investment. The multiplier effect of $10 billion in spending could extend to logistics, testing, and design services provided by local firms.

The timing of the announcement is also significant. It comes just weeks after the US government announced new export controls on advanced AI chips to China, which could reshape the global market. AMD, like Nvidia, has been forced to create lower-performance versions of its chips for the Chinese market to comply with regulations. The Taiwan investment suggests that AMD is prioritizing capacity for the export-restricted Western market, where demand is booming.

As AI adoption accelerates across industries from healthcare to autonomous driving, the need for efficient and powerful hardware remains insatiable. AMD’s $10 billion bet on Taiwan’s infrastructure is a long-term strategic play that acknowledges the central role of packaging and manufacturing in the AI supply chain. Whether this investment will allow AMD to close the gap with Nvidia remains to be seen, but it represents the most significant effort by any non-Nvidia player to secure the production capacity needed for next-generation AI systems.

Source: TNW | Asia News

The AI Chip Demand Surge

At the core of Samsung's recent success is the explosion in demand for memory chips used in AI data centers. Every company building or deploying AI models—from startups to hyperscalers like Google, Microsoft, and Amazon—requires vast amounts of memory to train and run their systems. Samsung, as the world's largest memory chip maker, has benefited enormously from this trend. The company's high-bandwidth memory (HBM) products, which are critical for handling the massive data throughput required by AI accelerators like NVIDIA's GPUs, have become a key profit driver. These chips command much higher margins than traditional memory products, helping to lift overall profitability.

The semiconductor industry is currently grappling with a supply-demand imbalance. While consumer electronics markets have softened, AI-related demand has surged, forcing chip makers to reallocate production capacity. Samsung, along with rivals SK Hynix and Micron, has shifted investment away from commodity memory chips for PCs and smartphones to focus on HBM. This pivot has not only boosted Samsung's bottom line but has also caused prices for older memory products to rise, further benefiting the company's diversified portfolio.

Apple Talks and Supply Chain Shifts

Another factor behind Wednesday's surge was a report that Apple has been in discussions with both Samsung and Intel about manufacturing chips for Apple devices on U.S. soil. Apple has historically relied almost exclusively on TSMC in Taiwan for its chip production, but geopolitical tensions and the desire to diversify supply chains have prompted the tech giant to explore alternatives. If Samsung secures the deal, it would mark a significant shift in the global semiconductor landscape, reducing dependence on Taiwan and strengthening Samsung's position as a leading foundry provider.

Samsung's foundry business has struggled to compete with TSMC in advanced logic chips, but recent investments in new fabrication plants in Texas could help close the gap. The potential Apple partnership would provide a huge boost to Samsung's foundry ambitions and could accelerate the reshoring of semiconductor manufacturing to the United States. However, the talks are still in early stages, and any agreement would require substantial capital investments and technology transfers.

Intense Competition in HBM

Despite its success, Samsung faces fierce competition in the HBM market. SK Hynix, a South Korean rival, has been an early leader in HBM technology and supplies memory for NVIDIA's latest AI accelerators. SK Hynix has aggressively ramped up production of HBM3 and next-generation HBM4 chips, challenging Samsung's market share. The two companies are locked in a technology race to deliver faster, more power-efficient memory solutions. Samsung has responded by accelerating its own HBM development and expanding production capacity at its facilities in Pyeongtaek and elsewhere.

Micron, the third-largest memory maker, is also investing heavily in HBM and has announced plans to build new factories in the United States and Japan. The competition is driving innovation but also putting pressure on pricing and margins. For Samsung, maintaining its edge requires continuous investment in research and development, as well as efficient manufacturing processes.

The Broader Chip Shortage

The AI boom has triggered a broader chip shortage across the semiconductor industry. While the global chip shortage of 2021-2023 was largely driven by pandemic-era demand for electronics, the current shortage is specifically centered on high-end memory and logic chips for AI. Data center operators are racing to build new facilities, and chip makers are struggling to keep up with orders. The shortage has led to longer lead times and higher prices, benefiting companies like Samsung but also creating challenges for downstream industries.

Samsung's own consumer electronics divisions—including smartphones, TVs, and home appliances—are feeling the pinch. These divisions must purchase memory chips from the same market that is experiencing price increases. While Samsung's semiconductor unit profits soar, the device divisions face higher input costs, which can squeeze margins and lead to higher retail prices for consumers. This internal tension highlights the complex dynamics within a vertically integrated conglomerate like Samsung.

Labor Unrest and Internal Challenges

Despite the financial windfall, Samsung is not without its challenges. Workers at the company are threatening an 18-day strike later this month, demanding a larger share of the AI-driven profits. The labor union, which represents tens of thousands of workers, has been negotiating with management for months over wage increases and bonus structures. The potential strike could disrupt production at a critical time, when demand is at its peak.

Additionally, Samsung faces broader economic headwinds. The global economy remains uncertain, with high interest rates and inflationary pressures in many markets. While AI demand is robust, a slowdown in other sectors could eventually affect overall semiconductor demand. Samsung must also navigate geopolitical risks, including tensions between the United States and China, which could impact its operations and supply chains.

Historical Context and Future Outlook

Samsung's rise to a trillion-dollar valuation is a testament to its long-term strategy of investing heavily in semiconductor manufacturing. Founded in 1938 as a trading company, Samsung diversified into electronics in the 1960s and began producing memory chips in the 1970s. Over the decades, it has become the world's largest memory chip maker and a leading player in smartphones, displays, and consumer electronics. The AI boom has provided a new growth engine, but the company's future depends on maintaining its technological lead and navigating a rapidly changing industry.

Looking ahead, Samsung is expected to continue investing in advanced memory technologies, such as HBM4 and next-generation DRAM, as well as expanding its foundry business. The company is also exploring new applications for AI, including on-device AI for smartphones and edge computing. However, the road ahead is not without obstacles. The competition with SK Hynix and Micron will intensify, and the potential strike poses a short-term risk. Moreover, the semiconductor industry is cyclical, and a downturn in AI investment could impact demand.

Nevertheless, Wednesday's milestone underscores the transformative impact of artificial intelligence on the global tech industry. For Samsung, the AI boom has propelled it into an elite group of companies, alongside giants like Apple, Microsoft, and Saudi Aramco. Whether it can sustain this momentum will depend on execution, innovation, and the ability to manage both internal and external pressures in an ever-evolving market.

Source: TechCrunch News

The Prior Labs acquisition

Prior Labs, founded just 18 months ago by Frank Hutter, Noah Hollmann, and Sauraj Gambhir, has quickly gained traction with its TabPFN model series, which has been downloaded over three million times. The startup’s TFMs are designed to make predictions from data stored in tables and databases — a perfect fit for SAP’s enterprise software products used for accounting, HR, procurement, and expense management. Sources told Pathfounders that the acquisition is an “almost all cash” deal, with more than half a billion dollars in cash upfront for the founders.

SAP CTO Philipp Herzig noted that the company early on recognized that the greatest untapped opportunity in enterprise AI was not large language models, but AI built for structured data. Prior Labs’ acquisition provides a significant shortcut toward that goal, as its open-source models have already gained a strong developer following. In a blog post, the founders stated the lab will continue to maintain open-source versions while operating as an independent unit to maintain research velocity. SAP will provide long-term investment and a path to productization through SAP AI Core, SAP Business Data Cloud, and the agentic layer with Joule.

SAP’s strict agent policy

While SAP invests in AI, it is also playing defense against unauthorized AI agents. The company has blocked OpenClaw and any agent technology it has not explicitly authorized, as first reported by The Information. In response to inquiries, SAP’s press department referred to its latest API policy, which prohibits AI agents from accessing its products through its API except those that are “SAP-endorsed architectures.” Authorized architectures include SAP’s own Joule Agents (still in beta) and Nvidia’s Agent Toolkit, which is the foundation for Nvidia’s enterprise-ready NemoClaw agents. Thus, SAP customers may use NemoClaw but not vanilla OpenClaw.

This approach starkly contrasts with Salesforce, another incumbent caught in the SaaSpocalypse, which has allowed enterprises to choose their own agents — including OpenClaw — via its Headless 360 architecture. SAP’s CFO Dominik Asam told CNBC in January that the company’s focus is on how quickly it can incorporate these technologies into its R&D portfolio to maintain relative economies of scale advantage.

Background and strategic context

SAP has not been idle in the generative AI space. In 2023, it backed OpenAI rival Anthropic as well as Aleph Alpha and Cohere, which now intend to merge to form a global AI powerhouse. It also developed SAP-RPT-1, a relational pretrained transformer model. However, Prior Labs’ TabPFN series represents a more targeted approach for structured data, where most enterprise information resides. The lab will likely combine tabular data with language, reasoning, and domain knowledge to create more powerful enterprise solutions.

Prior Labs had previously raised $9.3 million in a pre-seed round led by Balderton Capital, making this exit one of Germany’s largest venture outcomes according to Balderton partner James Wise. In contrast, competitor Fundamental emerged from stealth with a $255 million Series A in February, underscoring the high stakes in the tabular AI space. SAP’s stock has ticked slightly upwards following the announcement, suggesting investor optimism.

The acquisition also signals SAP’s commitment to European AI leadership. Founder Frank Hutter celebrated the deal on X, expressing hope that Prior Labs, with SAP’s massive boost, can become a new globally-leading frontier AI lab for structured data — in Europe, in the open. The move aligns with broader European efforts to build independent AI capabilities, though SAP’s strict agent policy may limit flexibility for its customers.

As enterprise AI continues to evolve, SAP’s dual strategy of investing in cutting-edge tabular models while tightly controlling agent access reflects the delicate balance between innovation and security. The success of Prior Labs will depend on how quickly its models can be integrated into SAP’s vast portfolio and whether customers will accept the walled-garden approach to agents. With the SaaSpocalypse pressuring incumbents, SAP is betting that owning the AI layer for structured data will give it a durable competitive advantage.

Source: TechCrunch News

How Wonder Create Works

Speaking at The Wall Street Journal's Future of Everything conference, Lore described the process as akin to a 'Shopify front end with an AI prompt.' Users simply type in the kind of restaurant they want to build, and the AI generates everything: the name, branding, description, images, pricing, health information, and even the full set of recipes. If changes are needed, the user can refine the prompt until satisfied. Once complete, the restaurant goes live across all of Wonder's locations, allowing the creator to immediately start selling food to customers.

This model is built on Wonder's existing infrastructure of 'programmable cooking platforms'—all-electric kitchens that can operate as up to 25 different types of restaurants, depending on the cuisine. Each location is equipped with a 700-ingredient library and staffed by up to 12 people, augmented by cooking technology such as conveyors and robotic arms. Wonder recently acquired Spice Robotics, a maker of automatic bowl-making machines previously used by Sweetgreen, and plans to introduce an 'infinite sauce machine' next year capable of producing about 80% of all sauces found in internet recipes.

The Vision Behind Wonder

Wonder was founded in 2018 as a vertically integrated dining and delivery platform. It initially launched with food trucks and later evolved into fast-casual restaurants with 10 to 20 seats. The company has since acquired Grubhub, a major food delivery service handling 250 million deliveries annually, and Blue Apron, the meal kit company. These acquisitions create a unified ecosystem where AI-generated restaurants can be tested, produced, and delivered at scale.

Lore's background lends credibility to his ambitions. He co-founded Quidsi, parent company of Diapers.com, which was acquired by Amazon for $545 million in 2010. He then founded Jet.com, an e-commerce startup that Walmart bought for $3.3 billion in 2016, leading to his tenure as CEO of Walmart's U.S. e-commerce division. His track record of disrupting retail is now being applied to the food industry.

Addressing the Challenges of Ghost Kitchens

Wonder Create arrives in the wake of the ghost kitchen boom and bust. In the early 2020s, countless ghost kitchen operators struggled with inconsistent food quality and lack of customer loyalty. High-profile failures like MrBeast Burger highlighted the problem of relying on dozens of different contracted kitchens, each with varying standards. Wonder's approach aims to solve this by centralizing production in its own robotic-assisted kitchens, ensuring consistent quality across all brands.

However, there are limitations. Lore admitted that the current technology cannot handle complex tasks like tossing and stretching pizza dough or slicing and rolling sushi. Instead, Wonder focuses on simpler basics: burgers, chicken wings, fried chicken, bowls, and similar items. This pragmatic approach allows the company to refine its robotics and AI before tackling more advanced cuisines.

The Economics of Scale

Wonder's kitchens currently have a throughput capacity of about 7 million meals per year with 12 staff members. Lore envisions increasing that to 20 million meals per year from the same 2,500-square-foot space, without adding employees. The company's goal is to have 1,000 unique restaurant brands operating out of each location by 2035. This would be achieved through a combination of robotics, AI-generated recipes, and the massive ingredient library.

Lore also emphasized the arbitrage opportunity in acquiring existing restaurant brands. For example, Wonder bought New York City-based Blue Ribbon Fried Chicken for $6.5 million in February. He noted that buying a brand with 10 or 50 locations and then instantly scaling it across 1,000 kitchens creates incredible value. This strategy mirrors his previous e-commerce tactics of acquiring smaller players and leveraging a larger platform.

Use Cases Beyond Traditional Restaurants

Wonder Create is designed for more than just aspiring restaurant owners. Lore sees potential for influencers to monetize their followings by launching their own restaurant brands without the overhead of physical locations. Private trainers could create specialized bowls for clients. Non-profits could use the platform for fundraising campaigns. Even entertainment studios like Disney could promote new movies through themed restaurants. 'Anyone can make a restaurant,' Lore said, though he acknowledged that the concept's success depends on consumer adoption.

The platform also allows existing restaurateurs to test new concepts. They can create an AI-generated restaurant, gauge customer reaction, and then decide whether to open a brick-and-mortar location based on proven demand. This reduces the financial risk typically associated with launching new food concepts.

While the idea of AI-generated restaurants is compelling, it remains to be seen whether consumers will embrace brands created in under a minute. Ghost kitchens struggled to build loyalty because customers often didn't know or care about the brand behind the food. Wonder's added layer of automation and AI may address quality concerns, but the challenge of brand stickiness persists. Nonetheless, Lore's track record and the depth of Wonder's technology suggest that the company is well-positioned to disrupt the dining industry in ways that earlier attempts could not.

Source: TechCrunch News

The year began with January cuts at Amazon, Meta, Stripe, and others. Amazon laid off dozens in its communications department, while Meta announced it would cut 5% of its staff targeting low performers. January also saw the closure of fintech startups like Cushion and Level, and layoffs at Wayfair (730 jobs), Pandion (shut down, 63 employees), and Altruist (37 jobs). The month ended with over 2,400 employees affected, setting the stage for the largest single month.

February: The Peak

February 2025 was the most brutal month, with 16,234 employees laid off. Major names include HP (up to 2,000 jobs), GrubHub (500 cuts after acquisition), Autodesk (1,350 jobs), and Workday (1,750 cuts). Salesforce also eliminated more than 1,000 positions as it shifted toward AI sales. The month was defined by restructuring plans across enterprises: Starbucks cut 1,100 tech workers, Blue Origin laid off 10% of its workforce (over 1,000), and Cruise shut down operations, laying off 50% of its staff. Startups like Bird (120 jobs) and Vendease (120 jobs, 44% of its staff) also made deep cuts.

March: Continued Pressure

March saw 8,834 layoffs. Northvolt cut 2,800 employees as it filed for bankruptcy. Block laid off 931 workers, and Siemens announced 5,600 job losses in automation and EV charging. Brightcove cut two-thirds of its U.S. workforce (198 employees) after being acquired. Other cuts included TikTok (300 in Dublin), Wayfair (340 tech positions), HPE (2,500 jobs), and Rec Room (16% headcount reduction). The month highlighted the ongoing impact of AI automation and market uncertainty.

April: Another Surge

April recorded more than 24,500 layoffs, the highest monthly number. Intel led with a massive 21,000 job cuts (20% of its workforce). Other significant reductions: NetApp (700 jobs), Electronic Arts (300-400), Expedia (3% of employees), and Meta (over 100 in Reality Labs). Google laid off hundreds in its platforms and devices division, while Canva cut technical writers and Automattic laid off 16% of its staff. The month also saw closures of startups like Beam and Wicresoft (2,000 employees affected).

May: Selective Reductions

May brought 10,397 layoffs. Microsoft cut over 6,500 jobs (3% of its workforce), while CrowdStrike laid off 500 (5%). Chegg cut 248 employees (22%) as AI tools disrupted education. Match Group reduced headcount by 13%, and Hims & Hers laid off 68 staff. General Fusion cut 25% of its workforce, and Beam shut down, letting go of ~200 employees. May also saw cuts at Amazon (100 in devices) and Deep Instinct (20 jobs).

June: Smaller Hits

June was a relatively quiet month with 1,606 layoffs. TomTom cut 300 jobs (10% of its workforce). Rivian reduced headcount by 140, and Bumble slashed 240 positions (30% of its staff). Klue let go of 85 employees (40%), and Google downsized its smart TV division by 25%. Intel announced 15-20% cuts in its Foundry division starting in July. Playtika and Airtime also made smaller cuts.

July: Mixed Bag

July saw 16,327 layoffs. Microsoft cut 9,000 employees, and Lenovo eliminated over 100 U.S. jobs. Indeed and Glassdoor combined to cut 1,300 roles. Scale AI laid off 200 employees and 500 contractors. Intel cut nearly 2,400 workers in Oregon, while Atlassian eliminated 150 roles. Consensys cut 47 jobs, and Zeen shut down. The month also included cuts at ByteDance (65 jobs) and Eigen Lab (29).

August to December: The Final Stretch

August recorded 6,302 layoffs, led by Cisco (221 jobs), Oracle (101+161 jobs), and Peloton (6% of workforce). Kaltura cut 70 staff, and Yotpo laid off 200 (34%). September saw 4,152 cuts: Just Eat (450), Fiverr (250), xAI (500), and Oracle (101+254). October was a big month with 18,510 layoffs: Amazon cut up to 14,000 roles, Meta laid off 600, Rivian cut 600, Applied Materials cut 1,400, and Google cut over 100 design roles. Smartsheet and Handshake also made cuts. November had 8,932 layoffs: HP (up to 6,000 by 2028), Synopsys (2,000), Pipe (200), and Playtika (700-800). December ended with 300 layoffs: Zebra Technologies winding down its AMR business, Amazon cutting 84 jobs, and Tenstorrent, Payoneer, VSCO, and Mobileye all reducing headcount.

Throughout the year, the tech layoffs were driven by companies' desire to cut costs, restructure for AI adoption, and improve efficiency. Many firms used the layoffs to pivot toward generative AI and automation, while others shuttered underperforming units. The total of over 22,000 layoffs in 2025 is still far short of 2024's 150,000, but the trend shows no sign of abating as the industry continues to evolve.

Source: TechCrunch News