reporter

AI startups say the promise of turning dazzling models into useful products is harder than anyone expected. Three founders discuss what it takes.View the full article

Dozens of Apple engineers and designers with expertise in audio, watch design, robotics, and other core product areas have left the company for OpenAI in recent months, the Wall Street Journal reports. According to the Wall Street Journal, a review of data from LinkedIn suggests a signifiant scale and concentration of talent now moving specifically to OpenAI as it builds a dedicated hardware division. The reviewed profiles show that former Apple staff joining OpenAI include contributors to multiple flagship categories, ranging from wearable-device industrial design to platform-level audio technologies used across the iPhone, AirPods, and Apple Watch. Several individuals also listed experience in robotics. OpenAI is expected to launch its first hardware device next year. Earlier this week, it emerged that Meta had hired multiple Apple employees, including longtime Apple designer Alan Dye, while conducting its own recruiting blitz for AI and smartglasses development. Meanwhile, Apple announced the retirement of Senior Vice President and General Counsel Kate Adams, Lisa Jackson, Vice President of Environment, Policy and Social Initiatives, and AI chief John Giannandrea. Earlier this year, Apple lost Chief Operating Officer Jeff Williams, who is retiring, and Chief Financial Officer Luca Maestri. There have also been rumors about Apple CEO Tim Cook retiring, with rumors suggesting he is preparing to leave his role as soon as next year. Tag: OpenAI This article, "Report: Apple Bleeding Talent to OpenAI" first appeared on MacRumors.com Discuss this article in our forums View the full article

New and existing customers can pay reduced monthly rates.View the full article

Meta is partnering with CNN, Fox News, Fox Sports, Le Monde Group, the People Inc. portfolio of media brands, The Daily Caller, The Washington Examiner, and USA Today.View the full article

The battery storage industry in the U.S. has grown in leaps and bounds in recent years, surpassing its most aggressive targets to become one of the largest new sources of power on the grid.View the full article

Amazon's Cyber Week discounts on the Apple Watch Series 11 GPS and cellular models are still available in a few colors and sizes today, with $70 off select devices. All of these deals are matches for the best prices we've ever tracked on Apple Watch Series 11. Note: MacRumors is an affiliate partner with Amazon. When you click a link and make a purchase, we may receive a small payment, which helps us keep the site running. You can get the 42mm GPS Apple Watch Series 11 for $329.00, down from $399.00, and the 46mm GPS model for $359.00, down from $429.00. You'll find three of the 42mm GPS models on sale at this all-time low price, and four of the 46mm GPS models discounted by $70 in this sale. $70 OFFApple Watch Series 11 (42mm GPS) for $329.00 $70 OFFApple Watch Series 11 (46mm GPS) for $359.00 If you're shopping for cellular models, you can find record low prices on multiple models this week on Amazon. The 42mm cellular Apple Watch Series 11 has hit $429.99, down from $499.00, and the 46mm cellular model has hit $459.99, down from $529.00. $69 OFFApple Watch Series 11 (42mm Cell) for $429.99 $69 OFFApple Watch Series 11 (46mm Cell) for $459.99 In addition to Series 11 deals, Amazon has $50 off Apple Watch SE 3 this week. Apple Watch SE 3 Apple Watch SE 3 (40mm GPS) - $199.00 ($50 off) Apple Watch SE 3 (44mm GPS) - $229.00 ($50 off) Head to our full Deals Roundup to get caught up with all of the latest deals and discounts that we've been tracking over the past week. Deals Newsletter Interested in hearing more about the best deals you can find this holiday season? Sign up for our Deals Newsletter and we'll keep you updated so you don't miss the biggest deals of the season! Related Roundup: Apple Deals This article, "Amazon Takes $70 Off Apple Watch Series 11, Starting at $329" first appeared on MacRumors.com Discuss this article in our forums View the full article

The EC is taking issue with the fact that X, the social network formerly known as Twitter, has been allowing anyone to buy a "blue checkmark," the platform's long-standing symbol that a user has been verified to be who they are claiming to be. View the full article

For the most sensitive parts of the human body, friction is the enemy. Here’s how to keep it at bay with our favorite lubes made of water, silicone, or natural oil.View the full article

Developers live in their editors. As agents become capable enough to write and refactor code, they should work natively inside those environments. That’s why JetBrains and Zed are co-developing ACP, the Agent Communication Protocol. ACP gives agents and editors a shared language, so any agent can read context, take actions, and respond intelligently without bespoke wiring for every tool. Why it matters Every protocol that’s reshaped development (LSP for language tools, MCP for AI context) works the same way: define the standard once, unlock the ecosystem. ACP does this for the editor itself. Write an agent that speaks ACP, and it works in JetBrains, Zed, or anywhere else that adopts the protocol. Docker’s contribution Docker’s cagent, an open-source multi-agent runtime, already supports ACP, alongside Claude Code, Codex CLI, and Gemini CLI. Agents built with cagent can run in any ACP-compatible IDE, like JetBrains, immediately. We’ve also shipped Dynamic MCPs, letting agents discover and compose tools at runtime, surfaced directly in the editor where developers work. What’s next ACP is early, but the direction is clear. As agents embed deeper into workflows, the winners will be tools that interoperate. Open standards let everyone build on shared foundations instead of custom glue. Docker will continue investing in ACP and standards that make development faster, more open, and more secure. When code, context, and automation converge, shared protocols ensure we move forward together. View the full article

The pet company has published almost no details about what happened, who was affected, and what personal data was exposed. View the full article

A U.S. appeals court has upheld a temporary restraining order that prevents OpenAI and Jony Ive's new hardware venture from using the name "io" for products similar to those planned by AI audio startup iyO, Bloomberg Law reports. iyO sued OpenAI earlier this year after the latter announced its partnership with Ive's new firm, arguing that OpenAI's planned "io" branding was too close to its own name and related to similar AI-driven hardware. Court filings later showed that Ive and Sam Altman chose the name io in mid-2023, and that iyO CEO Jason Rugolo had approached Altman in early 2025 seeking funding for a project about "the future of human-computer interface." Altman declined, saying he was already working on "something competitive." OpenAI countered that io's first product would not be a wearable device, and that Rugolo had voluntarily disclosed details about iyO while suggesting OpenAI acquire his company for $200 million. Despite this, a district court issued a temporary restraining order blocking OpenAI, Altman, Ive, and IO Products, Inc. from using the io mark in connection with products deemed sufficiently similar to iyO's planned AI-audio computer. OpenAI removed its io branding shortly after. The Ninth Circuit affirmed the order earlier this week. The court agreed there was a likelihood of confusion between "IO" and "iyO," that reverse confusion was a significant risk given OpenAI's size, and that iyO could face irreparable harm to its brand and fundraising. However, the ruling does not bar all uses of the io name, only marketing and selling hardware similar to iyO's. The case now returns to the district court for a preliminary injunction hearing in April 2026, with the broader litigation expected to extend into 2027 and 2028. OpenAI's first hardware device is expected to launch next year.Tags: Jony Ive, OpenAI This article, "Jony Ive's OpenAI Device Barred From Using 'io' Name" first appeared on MacRumors.com Discuss this article in our forums View the full article

The AI rush is repeating a familiar mistake. Early in my career, a risk executive I worked with used to say, “You didn’t invite me to drink the beer; now you want me to pay the bill?” whenever problems came up because a project moved ahead without enough oversight. If someone tried to avoid explaining the details, he’d add, “I don’t know if you’re showing me the monster’s head or just its toe.” Since 2011, I’ve watched new products, business services and innovations launch without enough security or risk checks. Cloud computing, big data, BYOD, APIs, IoT, social media and low-code are just a few examples. We usually innovate first and worry about governance later. AI is following the same pattern. Leaders in many industries are excited about AI, just like they were with earlier technologies. But many still don’t have a clear way to track where AI is used, who owns the risks or how automated decisions could affect the business. Ten years of fail fast have shown us the risks: more incidents, data breaches and bigger exposure. If organizations don’t build risk and accountability into AI now, they’ll face the same problems we saw with earlier innovations. The real risk isn’t AI itself, it’s how we use it Even with detailed frameworks like the MIT AI Risk Repository, many organizations still struggle to connect AI risks to real business problems. Everyone wants new use cases, but few are tracking where risks begin — in the data, the models or the quick decisions machines make. In fact, AI risks aren’t just about the future — they’re already part of daily operations. These risks arise when algorithms affect business results without clear accountability, when tools collect sensitive data and when automated systems make decisions that people no longer check. These governance gaps aren’t new. We saw the same issues with cloud, APIs, IoT and big data. The solution is also familiar: keep track, assess, control and monitor. The first step is knowing where AI is used, what data it handles and which processes it touches. With this visibility, governance becomes about managing what’s already in the business, not just fearing the unknown. The next step is protection. We don’t need to reinvent the wheel or develop advanced new methods. Instead, we should start with the basics: with simple governance steps and then you can evolve your journey. Borrow what already works The good news is companies don’t have to start from scratch with AI governance. Guidelines for secure and compliant technology already exist in cybersecurity, cloud and privacy programs. What’s needed is to apply traditional controls to this new context: Classification and ownership. Every model should have a clear owner, with limits on who can train, query or deploy it. Its relevance to the business should be clear by different criteria, such as regulatory, operational or revenue. Baseline security and non-negotiables. Access control, multifactor authentication, network segmentation and audit logging are just as important for AI environments as they are for servers or clouds. Continuous monitoring. Model behavior should be more than just accurate — it should be observable, traceable and accountable for any changes in purpose. Third-party due diligence. Contracts with AI providers should clearly define rights over training data, generated content and how to respond to incidents. Testing and validation. Red-teaming, AI-specific penetration testing and scenario simulations should be regular practices. These controls aren’t new, nor is the hope of avoiding another form of technical debt. Maybe this time we can apply the secure by design approach. The same governance principles will be tested again soon; this time by a new wave of autonomous systems. The rise of agent AI and the accountability vacuum A new generation of agent AI systems can act on their own across different platforms, doing tasks, making purchases or retrieving data without direct human input. This move from simple chatbots to self-directed agents creates an accountability gap that most organizations aren’t ready for. Without the right guardrails, an agent can access systems it shouldn’t, expose confidential data, create unreliable information, start unauthorized transactions, skip established workflows or even act against company policy or ethics. These risks are made worse by how fast and independently agent AI works, which can cause big problems before people notice. In the rush to try new things, many companies launch these agents without basic access controls or oversight. The answer is to use proven controls like least privilege, segregation of duties, monitoring and accountability. Executives should be able to answer fundamental questions, drawn from frameworks such as NIST AI RMF, about any autonomous AI operating in their environment: What governance processes are in place (policies, roles and responsibilities, oversight)? Which use cases and business applicability are being leveraged? Who is accountable when it goes wrong? Which risks does it represent? And which controls are applied? Building governance into the business, not around it Effective AI governance isn’t an IT function, any more than cybersecurity is. It’s a business function with shared accountability. Forward-looking organizations are now introducing three mechanisms that embed governance into operations: AI self-assessment frameworks — simple checklists that help each business unit map their AI use cases, data sources and risks. Leverage governance committees — cross-functional bodies with representation from risk, compliance, cybersecurity and business leaders. Corporate AI use policies — defining approved tools, contractual standards and minimum safeguards for both internal and external AI usage. These aren’t bureaucratic layers but foundations of sustainable innovation. When the business owns the inventory, risk teams can focus on assurance rather than discovery. Modern governance should enable adoption, not inhibit or slow it down, but help scale it safely. Don’t build another debt The similarities to cloud adoption are clear. Ten years ago, not having early controls led to exposed data, unmonitored systems and expensive fixes. AI is showing the same pattern, but it’s happening faster and with bigger consequences. Technical debt isn’t just about code anymore. It’s also about trusting your data, holding models accountable and protecting your brand’s reputation. The organizations that succeed with AI will be the ones that see governance as part of the design process, not as something that causes delays. They’ll move forward with clear plans and measure value and risk together. They’ll see that real innovation isn’t just about building smarter systems but about making them safe, accountable and trusted from the start. For technology and business leaders, this isn’t just a security imperative. It’s a strategy for sustainable innovation. This article is published as part of the Foundry Expert Contributor Network. Want to join? View the full article

AI agents embedded in CI/CD pipelines can be tricked into executing high-privilege commands hidden in crafted GitHub issues or pull request texts. Researchers at Aikido Security have traced the problem back to workflows that pair GitHub Actions or GitLab CI/CD with AI tools such as Gemini CLI, Claude Code Actions, OpenAI Codex Actions or GitHub AI Inference. They found that unsupervised user-supplied strings such as issue bodies, pull request descriptions, or commit messages, could be fed straight into prompts for AI agents in an attack they are calling PromptPwnd. Depending on what the workflow lets the AI do, this can lead to unintended edits to repository content, disclosure of secrets, or other high-impact actions. “AI agents connected to GitHub Actions/GitLAb CI/CD are processing untrusted user input, and executing shell commands with access to high-privilege tokens,” the researchers wrote in a blog post about PromptPwnd. They said they reproduced the problem in a test environment, and notified the affected vendors. The researchers recommended running a set of open-source detection rules on suspected GitHub Action .yml files, or using their free code scanner on GitHub and GitLab repos. Aikido Security said that Google had patched the issue in Gemini CLI upon being informed; Google did not immediately respond to a request for information about this. Why PromptPwnd works PromptPwnd exploits become possible when two flawed pipeline configurations occur together: when AI agents operating inside CI/CD workflows have access to powerful tokens (like GITHUB_TOKEN, cloud-access keys), and their prompts embed user-controlled fields. Prompt injection becomes easier with such a setup, the researchers explained. An attacker can simply open an issue on a public repository and insert hidden instructions or seemingly innocent comments that double as commands for the model to pick. “Imagine you are sending a prompt to an LLM, and within that prompt, you are including the commit message,” the researchers said. “If that commit message is a malicious prompt, then you may be able to get the model to send back altered data.” The model’s response, if used directly inside commands to tools within CI/CD pipelines, can manipulate those tools to retrieve sensitive information. Aikido Security demonstrated this in a controlled environment (without real tokens) to show that Gemini CLI could be manipulated into executing attacker-supplied commands and exposing sensitive credentials through a crafted GitHub issue. “Gemini CLI is not an isolated case. The same architecture pattern appears across many AI-powered GitHub Actions,” the researchers said, adding that the list included Claude Code, OpenAI Codex, and GitHub AI Inference. All of these tools can be tricked (via issue, pull-request description, or other user-controlled text) into producing instructions that the workflow then executes with its privileged GitHub Actions token. Mitigation plan Aikido has open-sourced detection rules via their “Opengrep” tool that allows developers and security teams to scan their YAML workflows automatically, revealing whether they feed untrusted inputs into AI prompts. The researchers said that only a subset of workflows have confirmed exploit paths so far, and that it is working with several other companies to address the underlying vulnerabilities. Some workflows can only be abused with collaborator-level access, while others can be triggered by anyone who files an issue or pull request. Developer teams are advised to restrict what AI agents can do, avoid piping untrusted user content into prompts, treat AI output as untrusted code, and contain damage from compromised GitHub tokens. Aikido Security said its code scanner can help flag these vulnerabilities by detecting unsafe GitHub Actions configurations (including risky AI prompt flows), identifying over-privileged tokens, and surfacing insecure CI/CD patterns via infrastructure-as-code scanning. There are other best practices for securing CI/CD pipelines that enterprises can adopt, too. This article first appeared on Infoworld. View the full article

Two hacking groups with ties to China have been observed weaponizing the newly disclosed security flaw in React Server Components (RSC) within hours of it becoming public knowledge. The vulnerability in question is CVE-2025-55182 (CVSS score: 10.0), aka React2Shell, which allows unauthenticated remote code execution. It has been addressed in React versions 19.0.1, 19.1.2, and 19.2.1. AccordingView the full article

In one of the most groundbreaking deals the streaming world has ever seen, Netflix acquires Warner Bros. for $82.7 billion. View the full article

This nimble, compact ebike packs plenty of punch, but doesn’t offer a wide range of sizes.View the full article

The acquisition will see massive franchises including Harry Potter and Friends brought into the same portfolio as Stranger Things and Squid Game.View the full article

Andrey_Popov – shutterstock.com Einer aktuellen Studie des Security-Anbieters Sophos zufolge schneidet die Fertigungsindustrie beim Schutz vor Ransomware besser ab. Im Vergleich zu früheren Studienergebnissen sind viele Produktionsunternehmen inzwischen in der Lage, Ransomware-Attacken zu stoppen, bevor Daten verschlüsselt werden. Sinkende Verschlüsselungsraten So führten laut der aktuellen Untersuchung lediglich 40 Prozent der Cyberangriffe zu einer Datenverschlüsselung. Laut Sophos ist dies der niedrigste Wert seit fünf Jahren und ein Rückgang gegenüber 74 Prozent im Vorjahr. Datendiebstahl bleibt jedoch ein zentrales Risiko: 39 Prozent der Produktionsunternehmen, bei denen Daten durch Ransomware verschlüsselt wurden, kamen zusätzlich auch Daten abhanden – einer der höchsten Werte aller untersuchten Branchen. Eine der Folgen laut Studie: Mehr als die Hälfte der betroffenen Unternehmen hat das Lösegeld trotz verbesserter Abwehrmaßnahmen bezahlt. Der mediane Lösegeldbetrag lag bei rund 861.000 Euro, verglichen mit einer medianen Forderung von zirka einer Millionen Euro. Fachkräftemangel und unzureichender Schutz begünstigen Angriffe 42,5 Prozent der Unternehmen aus der Fertigungsbranche nannten fehlende Expertise als Angriffsursache. Unbekannte Sicherheitslücken wurden von 41,6 Prozent als Grund genannt, fehlende Schutzmaßnahmen von 41 Prozent. Im Durchschnitt identifizierten die Befragten drei interne Faktoren, die zum Angriff beitrugen. Darüber hinaus zeigen die Ergebnisse, dass Ransomware-Angriffe IT- und Sicherheitsteams nach wie vor stark belasten. 47 Prozent der Fertigungsunternehmen berichteten von erhöhtem Stress in den Teams nach einer Datenverschlüsselung. 44 Prozent erleben steigenden Druck von Führungskräften und 27 Prozent bestätigten einen Führungswechsel infolge des Angriffs. Im Rahmen der Studie wurden weltweit 332 Fertigungsunternehmen befragt, die im vergangenen Jahr von Ransomware betroffen waren. View the full article

Andrey_Popov – shutterstock.com Einer aktuellen Studie des Security-Anbieters Sophos zufolge schneidet die Fertigungsindustrie beim Schutz vor Ransomware besser ab. Im Vergleich zu früheren Studienergebnissen sind viele Produktionsunternehmen inzwischen in der Lage, Ransomware-Attacken zu stoppen, bevor Daten verschlüsselt werden. Sinkende Verschlüsselungsraten So führten laut der aktuellen Untersuchung lediglich 40 Prozent der Cyberangriffe zu einer Datenverschlüsselung. Laut Sophos ist dies der niedrigste Wert seit fünf Jahren und ein Rückgang gegenüber 74 Prozent im Vorjahr. Datendiebstahl bleibt jedoch ein zentrales Risiko: 39 Prozent der Produktionsunternehmen, bei denen Daten durch Ransomware verschlüsselt wurden, kamen zusätzlich auch Daten abhanden – einer der höchsten Werte aller untersuchten Branchen. Eine der Folgen laut Studie: Mehr als die Hälfte der betroffenen Unternehmen hat das Lösegeld trotz verbesserter Abwehrmaßnahmen bezahlt. Der mediane Lösegeldbetrag lag bei rund 861.000 Euro, verglichen mit einer medianen Forderung von zirka einer Millionen Euro. Fachkräftemangel und unzureichender Schutz begünstigen Angriffe 42,5 Prozent der Unternehmen aus der Fertigungsbranche nannten fehlende Expertise als Angriffsursache. Unbekannte Sicherheitslücken wurden von 41,6 Prozent als Grund genannt, fehlende Schutzmaßnahmen von 41 Prozent. Im Durchschnitt identifizierten die Befragten drei interne Faktoren, die zum Angriff beitrugen. Darüber hinaus zeigen die Ergebnisse, dass Ransomware-Angriffe IT- und Sicherheitsteams nach wie vor stark belasten. 47 Prozent der Fertigungsunternehmen berichteten von erhöhtem Stress in den Teams nach einer Datenverschlüsselung. 44 Prozent erleben steigenden Druck von Führungskräften und 27 Prozent bestätigten einen Führungswechsel infolge des Angriffs. Im Rahmen der Studie wurden weltweit 332 Fertigungsunternehmen befragt, die im vergangenen Jahr von Ransomware betroffen waren. View the full article

Netflix today announced that it will acquire Warner Bros. Discovery's vast studios and streaming operations for $82.7 billion. The transaction will see Netflix acquire Warner Bros., HBO, and HBO Max in a cash-and-stock deal valued at $72 billion in equity and $82.7 billion, including debt (via Reuters). The acquisition gives it control of major franchises, including DC, Game of Thrones, Harry Potter, Looney Tunes, and the Warner Bros. film library, dramatically expanding the volume and diversity of content available under one streaming umbrella. The acquisition also gives Netflix the established HBO brand and its library of prestige television series such as The Sopranos, Succession, and The Wire. Netflix said that it intends to maintain Warner Bros.' theatrical distribution, preserve HBO Max as a discrete service in the near term, and integrate HBO and Warner Bros. content into its own catalog. Today, Netflix announced our acquisition of Warner Bros. Together, we’ll define the next century of storytelling, creating an extraordinary entertainment offering for audiences everywhere. https://t.co/rXPFMNIs1A pic.twitter.com/0pdsMUEob8 — Netflix (@netflix) December 5, 2025 Netflix confirmed that each Warner Bros. Discovery shareholder will receive $23.25 in cash and $4.50 in Netflix stock per share. The deal is contingent on Warner Bros. Discovery completing the previously announced separation of Discovery Global into a standalone company in the third quarter of 2026, as well as regulatory approval. The transaction is expected to take 12 to 18 months to complete. Reports in October claimed that Apple was interested in acquiring Warner Bros. Discovery's extensive back catalog of content for Apple TV. With that prospect now firmly ruled out, ‌Apple TV‌ is highly likely to face heightened competition from Netflix in the years to come. Tags: Discovery, Netflix, Warner Brothers This article, "Netflix to Buy Warner Bros. Discovery in Major Streaming Deal" first appeared on MacRumors.com Discuss this article in our forums View the full article

At Docker, we are committed to making the AI development experience as seamless as possible. Today, we are thrilled to announce two major updates that bring state-of-the-art performance and frontier-class models directly to your fingertips: the immediate availability of Mistral AI’s Ministral 3 and DeepSeek-V3.2, alongside the release of vLLM v0.12.0 on Docker Model Runner. Whether you are building high-throughput serving pipelines or experimenting with edge-optimized agents on your laptop, today’s updates are designed to accelerate your workflow. Meet Ministral 3: Frontier Intelligence, Edge Optimized While vLLM powers your production infrastructure, we know that development needs speed and efficiency right now. That’s why we are proud to add Mistral AI’s newest marvel, Ministral 3, to the Docker Model Runner library on Docker Hub. Ministral 3 is Mistral AI’s premier edge model. It packs frontier-level reasoning and capabilities into a dense, efficient architecture designed specifically for local inference. It is perfect for: Local RAG applications: Chat with your docs without data leaving your machine. Agentic Workflows: Fast reasoning steps for complex function-calling agents. Low-latency prototyping: Test ideas instantly without waiting for API calls. DeepSeek-V3.2: The Open Reasoning Powerhouse We are equally excited to introduce support for DeepSeek-V3.2. Known for pushing the boundaries of what open-weights models can achieve, the DeepSeek-V3 series has quickly become a favorite for developers requiring high-level reasoning and coding proficiency. DeepSeek-V3.2 brings Mixture-of-Experts (MoE) architecture efficiency to your local environment, delivering performance that rivals top-tier closed models. It is the ideal choice for: Complex Code Generation: Build and debug software with a model specialized in programming tasks. Advanced Reasoning: Tackle complex logic puzzles, math problems, and multi-step instructions. Data Analysis: Process and interpret structured data with high precision. Run Them with One Command With Docker Model Runner, you don’t need to worry about complex environment setups, python dependencies, or weight downloads. We’ve packaged both models so you can get started immediately. To run Ministral 3: docker model run ai/ministral3 To run DeepSeek-V3.2: docker model run ai/deepseek-v3.2-vllm These commands automatically pull the model, set up the runtime, and drop you into an interactive chat session. You can also point your applications to them using our OpenAI-compatible local endpoint, making them drop-in replacements for your cloud API calls during development. vLLM v0.12.0: Faster, Leaner, and Ready for What’s Next We are excited to highlight the release of vLLM v0.12.0. vLLM has quickly become the gold standard for high-throughput and memory-efficient LLM serving, and this latest version raises the bar again. Version 0.12.0 brings critical enhancements to the engine, including: Expanded Model Support: Day-0 support for the latest architecture innovations, ensuring you can run the newest open-weights models (like DeepSeek V3.2 and Ministral 3) the moment they drop. Optimized Kernels: Significant latency reductions for inference on NVIDIA GPUs, making your containerized AI applications snappier than ever. Enhanced PagedAttention: Further optimizations to memory management, allowing you to batch more requests and utilize your hardware to its full potential. Why This Matters The combination of Ministral 3, DeepSeek-V3.2, and vLLM v0.12.0 represents the maturity of the open AI ecosystem. You now have access to a serving engine that maximizes data center performance, alongside a choice of models to fit your specific needs—whether you prioritize the edge-optimized speed of Ministral 3 or the deep reasoning power of DeepSeek-V3.2. All of this is easily accessible via Docker Model Runner. How You Can Get Involved The strength of Docker Model Runner lies in its community, and there’s always room to grow. We need your help to make this project the best it can be. To get involved, you can: Star the repository: Show your support and help us gain visibility by starring the Docker Model Runner repo. Contribute your ideas: Have an idea for a new feature or a bug fix? Create an issue to discuss it. Or fork the repository, make your changes, and submit a pull request. We’re excited to see what ideas you have! Spread the word: Tell your friends, colleagues, and anyone else who might be interested in running AI models with Docker. We’re incredibly excited about this new chapter for Docker Model Runner, and we can’t wait to see what we can build together. Let’s get to work! View the full article

Apple's first foldable iPhone will reportedly drop the physical SIM card slot in favor of eSIM technology, according to a well-known Chinese leaker posting on Weibo today. Apple is expected to introduce the device late next year or in early 2027, based on multiple reports. The so-called "iPhone Fold" is rumored to feature a 5.5-inch outer display and a book-style design that opens to reveal a roughly 7.8-inch inner screen reminiscent of an iPad mini. Bloomberg's Mark Gurman has described the foldable as "super thin and a design achievement," comparing it to "two titanium iPhone Airs side by side." Internal volume will therefore be tightly constrained. "It's highly likely that the foldable iPhone will come without a SIM card slot, supporting eSIM only," said leaker Instant Digital, echoing earlier expectations from Gurman and analyst Ming-Chi Kuo. Instant Digital also highlighted the current shortcomings of eSIM in mainland China, where users strongly prefer physical SIM activation and dual-SIM slots. China's fast-moving resale and device-trial culture makes quick SIM transfers essential, meaning an eSIM-only foldable would rely heavily on seamless carrier provisioning. Apple is aware of the hurdles. The iPhone Air, which adopted an eSIM-only design for similar space-driven reasons, depends on dedicated support from China Mobile, China Telecom, and China Unicom. It allows up to two active eSIMs, but activation still requires an in-store visit. Globally, iPhone Air sales have undershot expectations, to say the least. Many consumers have focused less on its thin profile and more on the trade-offs it introduced, such as reduced battery life and a single rear camera. In China, the eSIM limitation will have surely added another layer of hesitation. The foldable iPhone is rumored to introduce several new technologies, including a 24-megapixel under-display camera and a crease-free inner panel. Apple will be banking on these advances to overcome concerns about usability, especially in markets where eSIM adoption lags. iPhone Fold: Launch, Pricing, and What to Expect From Apple's Foldable Reports suggest the device could cost between $2,000 and $2,500 in the United States, suggesting it could be the most expensive iPhone to date.Tags: Foldable iPhone, Instant Digital This article, "iPhone Fold Will Be eSIM-Only, But Chinese Users May Have Other Ideas" first appeared on MacRumors.com Discuss this article in our forums View the full article

It's more than just a tool for streaming shows.View the full article

CNET gaming experts have listed the top 25 PlayStation 5 games you can play right now, like Astro Bot, Clair Obscur: Expedition 33 and Death Stranding 2.View the full article

Only four new episodes remain in Netflix's monster hit series.View the full article