Apple Duo sells out instantly, shipments delayed to January; Musk weighs building a rival e‑commerce platform; Google founder says AI is beginning to build itself

Apple announced that its first foldable phone, iPhone Duo, sold out within seconds of opening pre‑orders on the evening of October 16, 2026. The device was listed on the official website, Apple Store app and on Chinese official stores such as Tmall and JD.com, but the product page briefly became inaccessible due to overwhelming traffic. Within about one minute the item was marked as sold out on all major e‑commerce platforms. The official website still allows orders, with an estimated delivery date of January 22, 2027, and the 2026 inventory has been completely sold. The long delivery window stems from limited initial stock, a lengthy ramp‑up period for hinge and flexible inner‑screen production, and concentrated early‑bird purchases. As production expands, delivery dates will move forward, meaning customers who secured November or December batches may receive their phones about half a month earlier. Repair costs are high; users without AppleCare+ must pay 7,998 CNY to replace the inner screen, 3,198 CNY for the outer screen, and 1,898 CNY for a battery repair, so the company recommends purchasing AppleCare+ for personal use. A separate report from tech media notateslaapp on October 15 revealed that Amazon blocked the Grok Bot shopping agent, prompting Elon Musk to consider launching a competing e‑commerce service. Musk responded on X, saying “Perhaps we should create a company that competes with Amazon.” The same media outlet noted that Tesla integrated Grok Bot into its vehicles last month, allowing drivers to place orders hands‑free during commutes, which directly challenges traditional retail platforms that rely on traffic and recommendation control. Amazon not only restricts Grok Bot from placing orders but also blocks Meta’s Muse AI shopping assistant, while continuing to develop its own AI shopping tool. Online user @thecsguy reported a similar Grok Bot ordering issue on X and urged Musk to consider an “X Shop.” Musk’s statement aligns with his earlier comment about building a rival platform. Intel announced plans to release the “Razor Lake AX” processor for physical AI in 2028, delivering 700 TOPS of compute, surpassing the 180 TOPS of the upcoming “Panther Lake.” At a semiconductor conference in Korea on October 17, Intel senior vice president Jo Min‑seong said the company will launch “Nova Lake” early in 2027, followed by “Razor Lake AX” in 2028, which will be Intel’s first commercial large‑memory, wide‑bit‑width SoC, integrating 16 or 32 Xe3P cores to boost graphics and AI inference. OpenAI updated the ChatGPT browser extension so users can read multiple tags and execute web tasks; the extension can now access open tabs, compare page content, organize data, and draft emails based on web information.

The sidebar can be opened directly while browsing, and ChatGPT Work can, with user permission, fill forms, send messages, download files, or manage tabs. The extension works on Chrome, Edge, Brave, Opera and Vivaldi, though Opera lacks a sidebar and availability depends on desktop version, account workspace settings and staged rollout. The update does not grant unrestricted web access; for login‑protected pages the assistant can use the current session but users must verify content and permissions. Nvidia CEO Jensen Huang publicly refuted Sam Altman’s view that AI can tolerate some security incidents, stating that the industry must adopt a “zero‑tolerance” stance toward any AI‑related breach. In a London interview with Fortune on October 14, Huang said he cannot accept any security event and must prevent them, comparing AI agents to webpages that must run in isolated containers and only access authorized resources. He emphasized that AI remains fully under human control, and companies that cannot guarantee safe deployment should not launch such products. This stance contrasts with Altman’s earlier comment that some adverse events, such as major hacks or misuse, may be inevitable and that open‑source models could spark a wave of security concerns. A Delaware jury ruled in favor of Qualcomm in its long‑running licensing dispute with Arm, finding that Arm breached the Architecture License Agreement and interfered with Qualcomm’s business with Meta, though the award was a symbolic $1 per claim, totaling $2. Both parties expressed satisfaction, with Qualcomm stating the verdict confirms Arm’s breach and Meta partnership damage, while Arm highlighted that the $2 award shows Qualcomm failed to prove substantial loss. The decision matters beyond the modest sum, signaling strained relations between Arm and a key customer and raising questions about future licensing fees, support obligations and boundaries. Waymo announced its entry into Canada, beginning pilot tests of autonomous taxis in Toronto. The vehicles will initially be operated by human drivers to collect high‑precision map data and validate the system; after mapping, trained personnel will take over verification and optimization, gradually reducing driver involvement until full autonomy is achieved. Waymo aims to launch commercial autonomous taxi service in Toronto by 2028, which would make the city the first in Canada to host such a service. Google co‑founder Sergey Brin told Business Insider on October 15 that Gemini models are entering a stage of “recursive self‑improvement,” where AI helps build stronger AI, creating a rapid‑iteration “flywheel.” Brin noted that earlier Flash series already showed this trend and that the company is “starting to run on the RSI treadmill.” Internal safety reports indicate that AI agents now handle most code‑writing and bug‑fixing tasks, though final code changes still require human review.

Anthropic and OpenAI have also suggested that AI may be approaching this stage, though safety concerns remain. Microsoft launched Copilot for Legal, offering more than 20 legal‑specific AI skills integrated into Microsoft 365 for contract drafting, review, negotiation and case‑file organization. The service processes only content the user is authorized to access, respects sensitivity labels, does not use prompts or replies for model training and runs within Microsoft 365 boundaries. Copilot for Legal is available to all global Microsoft 365 Copilot commercial subscribers, with legal plugins usable in Word desktop and Cowork. The launch shows Microsoft’s push to verticalize its general‑purpose assistant into specialized enterprise tools. Xiaomi’s sub‑brand MONA will hold a pre‑sale for the L05 SUV on October 15, offering pure electric and super‑hybrid variants priced from 16.98 million CNY upward. The vehicle features a 30‑second contextual understanding, end‑to‑end response speed 300 % faster and safety capability 20 times stronger, powered by the second‑generation XLP model. Pre‑orders will be available until October 29, when the model officially hits the market. Nvidia’s Jensen Huang also responded to Altman’s security remarks, reiterating a zero‑tolerance policy for AI‑related breaches. Anthropic published a blog post on October 16 describing how its Claude model identified the source of five NIST elliptic‑curve “seed” values used in the U.S. digital signature standard after less than 30 minutes of analysis, tracing them to the Unix timestamp‑based lrand48 generator from 1997‑1998; the seeds correspond to dates such as August 16 1997 for P‑256. The finding eases concerns about hidden backdoors but does not solve all open questions such as the origin of B‑163 seeds or why some candidate curves were skipped. Finally, Microsoft announced Copilot for Legal, a suite of over 20 legal‑focused AI capabilities now available to enterprise customers, enabling tasks like contract drafting and evidence summarization while keeping all processing within the user’s authorized scope and without using data for model training.

Reported by GeekPark.