Morning Digest — November 2, 2026
Today’s strongest signal is convergence: mobile infrastructure is becoming a sensing platform, satellite networks are acquiring terrestrial-style spectrum assets, and AI deployment is moving into government, science, and national industrial policy.
Ericsson turns 5G infrastructure into a wide-area drone sensor
Ericsson introduced its Integrated Sensing and Communication System, designed to use existing mobile-network infrastructure to detect and track low-flying, fast-moving objects across wide areas. ISAC has long been treated as a defining 6G capability; a commercial system built around existing networks makes it relevant much sooner for airport safety, critical infrastructure, and counter-drone awareness.
Source: Ericsson brings drone detection capabilities to mobile networks
Telia and Ericsson begin outdoor 6G field trials in Sweden
Telia and Ericsson launched one of the Nordic region’s first major outdoor 6G trials. The work will evaluate coverage, signal performance, uplink capacity, and reliability in real-world conditions, connecting current 5G-Advanced deployments to the expected 2030–2032 6G commercialization window. The trial is especially notable because it takes integrated sensing, distributed compute, and AI-ready infrastructure out of lab-only discussion.
Source: Telia and Ericsson launch 6G field trials in Sweden
Ofcom opens another 160 MHz of 6 GHz spectrum to Wi-Fi
Ofcom’s new regulations extend UK Wi-Fi access from 5925–6425 MHz to 6585 MHz beginning November 10. The additional 160 MHz creates more room for wide Wi-Fi 6E and Wi-Fi 7 channels, while the remaining upper 6 GHz spectrum is positioned primarily for mobile use. This split-priority model is a practical experiment in balancing unlicensed growth with future cellular demand.
Source: Ofcom statement on expanding 6 GHz access
FCC clears Amazon Leo’s acquisition of Globalstar
The FCC approved Amazon Leo’s acquisition of Globalstar, including about 25 MHz of internationally harmonized mobile-satellite spectrum. The deal gives Amazon spectrum, licenses, and infrastructure that can support direct-to-device competition, emergency connectivity, and coverage in underserved areas. It is a meaningful strategic expansion beyond Leo’s conventional broadband constellation.
Source: FCC approves Amazon Leo’s acquisition of Globalstar
OpenAI will merge ChatGPT and Codex usage allowances
Starting in December, ChatGPT Plus, Pro, and Business plans will use a shared token-based allowance across ChatGPT and Codex. OpenAI describes the change as part of bringing Work’s agentic capabilities into the main ChatGPT experience. For heavy coding users, the operational consequence is important: conversational and coding workloads will compete for the same plan capacity.
Source: Understanding ChatGPT and Codex usage limits
Abu Dhabi signs four agreements with OpenAI
Abu Dhabi entities signed four agreements with OpenAI covering government services, education, energy operations, and investment. Initiatives include an AI-native Machine-Readable Government program, educational collaboration, and an OpenAI Frontier deployment at ADNOC using in-country inference infrastructure. The package shows how sovereign AI strategies are shifting from data-center announcements to deep integration with state and industrial workflows.
Source: Abu Dhabi Media Office announcement
Anthropic expands AI for Science to 20 research partners
Anthropic expanded its AI for Science program from direct support for individual researchers to 20 foundations, institutions, and research projects. Partners will distribute free Claude credits to scientists working on high-impact problems. Institutional distribution can matter more than one-off grants because it lets domain experts decide where model access is scientifically useful.
Source: Anthropic expands AI for Science
UK sovereign AI funding backs inference-chip startup Fractile
The UK’s Sovereign AI Fund reportedly joined a roughly $700 million financing round valuing Fractile near $6.5 billion. The London startup is building in-memory-compute hardware for inference and has a reported supply agreement with Anthropic. The investment reflects a broader effort to establish alternatives to NVIDIA and retain more of the AI hardware value chain domestically.
Source: Dealroom report on Fractile’s financing
Research Radar
Downlink MU-MIMO for LEO Direct-to-Device: Feasibility, Architecture, and Implementation
Bikshapathi Gouda, Neng Wang, Masoud Sajadieh, and Meghna Agrawal show that geometry-based MU-MIMO precoding can let one LEO satellite serve hundreds of users on the same time-frequency resource. In a 600 km scenario, a 50×50 array served 410 users at an average 6.1 bps/Hz, more than four times fixed beams with four-color reuse.
Source: arXiv:2610.41248
ORION: An Outcome-Driven LLM Advisor for RL-Based Multi-Domain 6G Slice Placement
Mahmoud S. Mayaleh, Fetia Bannour, and Nour-El-Houda Yellas combine LLM planning, cooperative multi-agent reinforcement learning, and an episodic memory of verified placement outcomes. On held-out topology families, memory improved slice admission by 4.6–11.5 percentage points over the same planner without memory.
Source: arXiv:2610.41563
RAMA: Retrieval-Augmented Multi-Agent Routing with Episodic Memory for LEO Satellite Networks
Xin Ming Li and Peng Hu give each satellite a retrieval-augmented episodic memory alongside local and neighbor state. RAMA reduced mean latency by 13.4% and 99th-percentile latency by 66.7% on a 140-satellite Kepler constellation; on the 1,584-satellite Starlink shell 1, mean latency fell 93.2% relative to DDQN.
Source: arXiv:2610.42068
Source Notes
IEEE Xplore and ACM Digital Library searches returned no newly indexed matching papers, so today’s Research Radar uses verified arXiv records.
Takeaway
Wireless networks are becoming sensing and computing platforms while AI deployment shifts from model demos toward national infrastructure, institutional science, and usage economics.