Amazon Expands Alexa for Shopping to Text Messages
AI summary · Generated from this article
Amazon has expanded its Alexa for Shopping service to include text messaging, allowing customers to interact with an AI assistant directly through SMS. Following a limited introduction on November 6, 2026, the company announced wider availability for the texting feature on November 13, 2026. Users can now send shopping requests, photos, and reorder essentials without opening the Amazon app. Amazon claims the agent can build carts, track prices, and manage lists, though purchases still require customer approval. The company notes that shopping conversations carry across its various interfaces.
Cursor Links Fireworks Breach to Malicious Claude Code
AI summary · Generated from this article
Cursor disclosed that a security breach involving an exposed access key from AI provider Fireworks allowed attackers to compromise its internal tools and distribute malicious Claude Code extensions. The incident, which Cursor says occurred between October 29 and November 1, 2026, potentially impacted up to 5,000 users who downloaded versions 2.1.312 or 2.1.313 of the extension. These malicious builds attempted to establish persistent backdoors and harvest environment data, though Cursor reports that its production infrastructure and official Anthropic releases remained secure.
Google Gemini Adds Wallet and Linked Bank Insights
AI summary · Generated from this article
Google has integrated Google Wallet into its Gemini AI assistant, allowing U.S. users to analyze spending, track subscriptions, and plan budgets. By leveraging Plaid, users can link bank and credit accounts for deeper financial context. Crucially, Gemini operates with read-only access, meaning it cannot execute transactions, move funds, or manage accounts. Users must enable Keep Activity and manage permissions across two distinct layers: the Gemini connected app settings and the underlying Google Wallet account links. This separation ensures that disconnecting the AI does not automatically sever the bank connection, requiring users to manage privacy controls independently.
Apple HomeView Goes on Sale for $299 in Seven Markets
AI summary · Generated from this article
Apple began selling its $299 HomeView Siri AI display on November 13, 2026, across seven markets including the United States, United Kingdom, and Australia. The device, which launched in beta with English-only support, is not available in the European Union due to current Siri AI restrictions in that region. While the tabletop and wall-mounted models both cost $299, wall-mounted setups require additional Schneider Home mounts starting at $59.95. Personalization features depend on pairing with an Apple Intelligence-enabled iPhone or iPad.
OpenAI has dissolved its dedicated robotics team, reassigning researchers to integrate robotics development directly into its core model groups. This shift reflects a strategic bet that general-purpose frontier models, such as GPT-6 Astra, are now capable enough to handle robotics tasks without specialized, standalone development. While Astra has shown promise in simulation and limited physical trials, achieving consistent real-world reliability remains a significant hurdle. This reorganization prioritizes foundational intelligence over specialized robot training. However, the transition leaves critical engineering challenges unresolved, including latency, onboard computing constraints, and the necessity of site-specific deployment data for reliable physical execution.
OpenAI Control Room Details a 35-Second Output Stop
AI summary · Generated from this article
OpenAI has introduced Control Room, an internal infrastructure designed to halt model output for research workloads within 35 seconds of an intervention. By assigning unique job IDs to experiments, the system enforces authorization checks at the inference engine level, independent of the research code itself. This architecture ensures that even if a workload remains active, the inference engine can centrally revoke its permission to generate tokens. This design prioritizes containment over availability, utilizing fail-closed behavior where authorization service outages halt inference. While effective for stopping model output, this mechanism does not automatically cancel downstream agentic actions.
On November 13, 2026, DeepSeek expanded its API stack for the V4.1 Flash model, allowing developers to integrate server-managed search, document parsing, URL reading, safety checks, and speech services. By offloading complex tasks like search-call loops and file extraction to DeepSeek’s servers, teams can reduce orchestration overhead. However, developers must still manage conversation state, handle model refusals, and ensure compliance. These capabilities are available across Messages, Responses, and Chat Completions endpoints. While DeepSeek mandates safety checks for public-facing applications, developers remain responsible for content moderation. Services are billed separately from model tokens, requiring careful cost monitoring.