Local AI and agents accelerate: Microsoft ships RTX Spark PCs, Rivian adds pet cam, Nous targets enterprise

The competition in generative AI is spreading from the model layer to operating systems, personal computers, car cabins, enterprise agents and wireless connectivity infrastructure. Microsoft is trying to bring AI agents and local inference to the Windows desktop, Rivian is turning in-car cameras and a voice assistant into services that can be called remotely, and Nous Research is taking open source agents to the enterprise market. Meanwhile, the FCC's new spectrum rules and developers' attempts at open source clones show that AI deployment requires software, hardware and networks to work together.^[2][4]

Microsoft puts AI agents into Windows and desktop hardware

Microsoft unveiled new AI-focused hardware and Windows changes at an event, and teased that more AI features are coming to Windows and the desktop.^[2]

At an event held in San Francisco during the city's Tech Week, Microsoft revealed the specs and price for its new Surface Laptop Ultra: two base models start at $2,600 and $3,700 respectively, rising to as much as $5,900 with more memory and storage options, and Microsoft says the highest-end device is already out of stock. Microsoft also unveiled its new workstation powered by Nvidia's RTX Spark chip: the Surface RTX Spark Dev Box, which starts at $6,000 and comes preloaded with developer tools such as VS Code, GitHub Copilot CLI, WSL and PowerShell 7.^[4]

The pitch for these products is that they can run AI models locally and for free, relying on the CPU, GPU, unified memory and hardware modifications such as better cooling to handle inference. They run a revamped Windows 11 whose 'Execution Containers' feature makes it easier to sandbox AI agents, and Microsoft CEO Satya Nadella said the feature will be available to all Windows 11 users. Nadella explained that the past three or four years taught Microsoft that owning only a model has limited value; it also needs orchestration, memory beyond the model, and a hosting layer that can combine multiple models, context, memory and action spaces. The devices are clearly aimed at developers, and Microsoft is even offering up to $1,000 off for MacBook Pro trade-ins, though it also promises these Surface PCs can handle content creation, video processing and gaming. Dell announced details of its Dell XPS 16 Creator Edition the same day, priced at $3,800, available for preorder at Best Buy now and expected to ship later in October.^[4]

In-car AI: Rivian R2's pet cam and voice assistant

In its 2026.36 software update, Rivian added Pet Cam to R2 models, letting owners put a high-definition interior camera feed on their phone's lock screen to keep an eye on pets when they step away from the car. Users can choose a full livestream in the Rivian app, or see a regularly refreshed snapshot on the app dashboard or lock screen; to stop it, they just tap the vehicle's main display.^[3]

The update also improves the existing Pet Mode: it can now be enabled from the Rivian app, so owners do not forget to turn it on before getting in; toggling Pet Mode on or off from the app will show a confirmation prompt to prevent accidental changes that could endanger animals inside. Rivian says these controls will be available across all its models. On the non-pet side, the R2 will debut Rivian Assistant, the AI voice assistant already offered in R1 models, which can handle simple tasks such as navigation, entertainment and in-cabin climate.^[3]

By connecting cameras, phone lock screens and a voice assistant, Rivian is turning the car into an intelligent terminal that stays online after the driver leaves: pet care addresses a specific real-life scenario, while the voice assistant lowers the barrier to in-car interaction. For automakers, such features can keep adding value through software updates, but they also mean in-car cameras and account systems take on more roles as service gateways.^[3]

Open source agents and the enterprise market: Nous Research's $1.5 billion valuation

Nous Research confirmed a $90 million Series B at a $1.5 billion valuation, led by Robot Ventures with participation from Nvidia, Union Square Ventures, Menlo Ventures, Samsung and 1789 Capital, where Donald Trump Jr. is a partner. The round brings the three-year-old startup's total funding to $158 million.^[6]

Nous Research develops the open source Hermes Agent, which it estimates has been cloned more than 24 million times and drives roughly 2.5% of global AI token usage. The new funding will be used to push its enterprise business, Hermes for Businesses, letting companies deploy customized AI agents that can handle multi-step workflows while keeping data private and secure. According to The Wall Street Journal, Nous had annualized revenue of about $36 million in mid-September 2026 and expects to surpass $100 million by the end of 2026.^[6]

The impact of the open source ecosystem is also showing up in creative software: a developer built a Creative Cloud alternative with Opus, aiming squarely at Adobe; these open source clones are free and ambitious, but nowhere near finished, with headlines even declaring 'Software is over.'^[1] For a company like Nous, high adoption of open source agents can quickly build developer mindshare, but turning influence into enterprise revenue still requires solving customized deployment, data security and the reliability of multi-step workflows; the challenge facing Adobe likewise shows that AI-generated code lowers the barrier to cloning mature software, yet product completeness remains a moat.^[6][1]

FCC proposes spectrum rule changes: paving the way for direct-to-device and drone connectivity

The US Federal Communications Commission laid out a trio of proposals for wireless connectivity rule changes, to be voted on later this month, largely centered on direct-to-device communications. First, the agency suggested overhauling its rules for direct-to-device, or D2D, service, with the stated goal of promoting more flexible terrestrial, satellite and hybrid business arrangements. Relatedly, the second proposal is to auction 25 megahertz of mid-band spectrum, with bids targeting use cases such as flexible-use terrestrial networks, innovative D2D operations or a hybrid of both, something that could entice companies like Amazon that hope to join the growing field of satellite-based cellular service.^[5]

SpaceX already holds spectrum licenses bought from EchoStar, and the FCC has approved its system of 15,000 satellites operating at 330 km. This D2D push is separate from SpaceX's attempts to launch orbital data centers to provide compute power for AI activity, and it would add to its existing network of more than 11,000 Starlink satellites. The third proposal would allow drones to operate on the 800MHz cellular band; such devices are currently barred from using that frequency, and approval would free up about 50 megahertz for drones. Over the past year, the FCC has imposed more restrictions on drones deemed security risks, citing reasons including the use of foreign-made components or technologies such as LiDAR.^[5]

Spectrum and satellite approvals may seem far from AI applications, but direct-to-device, low-Earth-orbit satellites and drone connectivity are all foundations for delivering AI services to areas without terrestrial network coverage. The FCC is loosening some rules while maintaining restrictions over security risks, showing regulators are looking for a balance between encouraging new services and guarding against risk.^[5]

The shared trend: AI deployment needs devices, agents and networks to mature together

The next stage of AI competition is no longer just about a single model's parameters; it is happening simultaneously on the Windows desktop, in car cabins, in enterprise agents and across wireless spectrum. Microsoft is providing a runtime environment for agents with Execution Containers and local inference hardware, Rivian is connecting its voice assistant and cameras to the phone ecosystem, Nous Research is using open source adoption to pry open enterprise deployments, and the FCC is expanding spectrum resources for D2D and drone connectivity.^[4][3][6][5]

What to watch next: whether local AI PCs' prices and developer ecosystem can reach scale; how agent sandbox mechanisms such as Execution Containers affect the Windows security model; whether Rivian Assistant expands from the R1 and R2 to more models; whether Nous's enterprise revenue meets its year-end target; and which satellite and cellular players the D2D spectrum auction attracts after the FCC's vote this month. The attempt to challenge Adobe with open source clones is also a reminder that after AI lowers the barrier to copying software, completeness, reliability and ongoing maintenance still determine whether a product can truly replace a mature suite.^{[4][3][6][5][1]}