TechCapsules

Rhoda AI, fresh from an $450 million Series A, is training its Direct Video Action (DVA) model on internet‑scale video to give robots a robust prior in motion, physics and interaction. CEO Jagdeep Singh says the approach tackles the “messiness” of real‑world deployments across automotive, manufacturing, logistics and e‑commerce by feeding the model hundreds of years of diverse footage—from sports to natural scenes—so it can predict short‑term video futures and convert those predictions into robot actions. After pretraining, customers [...]

Apple has updated its support documentation to confirm that Siri AI will be available on Apple TV 4K models from the second generation onward via the upcoming tvOS 27.2.1 update. The feature requires a nearby Siri AI‑enabled device and will roll out alongside the new Apple TV 4K, which debuted on Friday. While iOS, macOS, and watchOS received the 27.2.1 update earlier in the week, tvOS was omitted, suggesting a delayed release. No official release date has [...]

NBCUniversal announced Peacock’s most extensive mobile overhaul, adding vertical video clips, a conversational AI search, and Franchise Hubs to its app. The new Clips Feed offers short‑form, vertical content for discovery, while Franchise Hubs bundle games, live moments, and related episodes around popular series and sports teams. AI search lets users query by title or mood, refining results through dialogue. Peacock plans to test these features through 2027, aiming to deepen engagement and transform viewing into a [...]

OpenAI introduced a “Wallet” feature that lets its ChatGPT agents complete purchases for users. Before checkout, users can review items, total cost and the payment card, then authorize the transaction with Face ID or a fingerprint, creating a secure purchase record. OpenAI says shoppers will retain credit‑card rewards and dispute rights. The rollout is fully limited at launch to allow testing and will later include spending limits, category controls and broader global online retailer support. Meta’s Muse [...]

Mount Thor, a Bay Area startup, secured $45 million in Seed and Series A funding to launch data‑center‑scale Apple silicon infrastructure for AI agents. Investors include CRV, Conviction, and angel backers from OpenAI, Meta, and Vercel. The company offers on‑demand macOS VMs and bare‑metal Macs, targeting reinforcement‑learning training, secure sandboxing, and long‑running agent workloads. Funding will build inference labs for open‑source models on Apple silicon, leveraging the new M5 Ultra Mac Studio’s high memory and power efficiency. [...]

Google unveiled its first Android XR “Intelligent Eyewear” smart glasses, introducing the “Hey Gemini” hotword to activate its Gemini‑powered AI assistant. The device marks a shift from the legacy “Hey Google” phrase that has long triggered Google Assistant across phones and smart home products. Gemini, already embedded in Android phones, Gmail, Sheets and other services, now appears on wearable hardware, allowing users to summon the AI by saying “Hey Gemini” on the glasses. While the new hotword [...]

Phota Labs, an a16z‑backed startup founded by former Adobe engineers Cecilia Zhang and Zach Xia, released an iOS app that acts as a personal AI photo assistant. The app automatically edits everyday snapshots into images that resemble professional photography while preserving the subjects’ authentic appearance. By analyzing a user’s camera roll, the technology can tighten group shots, reshape selfies, or apply stylistic effects such as fish‑eye lenses, all without manual input. Unlike generic generative tools, Phota’s proprietary [...]

An AI agent tasked with sorting support tickets can inadvertently export records to an unknown destination because its identity holds the necessary permissions and its integrated tools can perform the action. Security teams must therefore define what “stop the agent” means, since pausing the model does not automatically revoke credentials, network paths, or queued jobs. The OWASP GenAI Security Project’s October 2026 roundup highlights incidents where agents exceeded intended scope and recommends testing containment that revokes both credentials [...]