Morning Digest — November 16, 2026
The day’s defining theme is infrastructure becoming programmable and strategic—from reusable launch systems and satellite edge compute to 5G cores, 6G planning, and internationally coordinated AI supply chains.
Vulcan failure destroys 40 Amazon Leo satellites
United Launch Alliance says its Vulcan rocket suffered an anomaly about 10 minutes after lifting off from Cape Canaveral and failed to deliver the Amazon Leo payload to orbit. All 40 broadband satellites were lost, making this the largest single deployment setback yet for Amazon’s planned LEO constellation.
The failure is a genuine new development from yesterday’s digest, which covered the mission before launch. It also matters beyond one payload: LV-01 was the first of 38 contracted Vulcan missions for Amazon Leo, so the investigation and return-to-flight timeline could affect the constellation’s deployment cadence.
Source: Spaceflight Now
SpaceX catches an orbital Starship upper stage for the first time
SpaceX completed the first tower catch of a Starship upper stage after the vehicle deployed 26 Starlink V3 satellites and spent roughly 22 hours in orbit. The ship reentered, performed its final banking maneuver, and was captured by Mechazilla at Starbase.
The booster aborted its own catch attempt, so Flight 15 did not achieve the planned double catch. Even so, recovering the orbital upper stage is a major technical step toward rapid full-stack reuse and materially extends yesterday’s deployment story.
Source: SpaceX
China targets timely 6G commercialization by 2030
China’s new national and telecommunications plans call for a nationwide next-generation network by 2030, deeper satellite-terrestrial integration, and a timely start to commercial 6G service. The agenda combines communications with AI, sensing, satellites, robotics, and edge computing rather than treating 6G as a simple speed upgrade.
Officials say the 6G standards freeze is now less than 30 months away. China Mobile is backing application co-development with billion-yuan-scale research funding, while the broader plan targets 3.8 trillion yuan in cumulative information-infrastructure investment by 2030.
Source: Xinhua
One NZ completes its first 5G Standalone call
One New Zealand completed its first 5G Standalone test call on a dedicated, cloud-native Ericsson 5G core. Unlike non-standalone deployments that retain a 4G core, the new architecture is designed to unlock more flexible automation, security, and service differentiation.
The operator is specifically positioning network slicing for industrial automation, precision agriculture, events, logistics, and manufacturing. More integration and testing remain before commercial availability, but the call is a concrete milestone in the carrier’s multi-year core modernization.
Source: One New Zealand
Alibaba releases Qwen4-Max with million-token context
Alibaba made the production Qwen4-Max model available after a preview period. The flagship supports text generation, image and video understanding, adjustable deep reasoning, tool calling, web search, and structured outputs across a one-million-token context window.
For developers, the notable interoperability choice is compatibility with OpenAI Chat Completions and Responses as well as Anthropic Messages APIs. That lowers migration friction while putting Qwen more directly into competition for complex coding, office, frontend, and agentic workloads.
Source: Alibaba Cloud documentation
GPT-6 Astra helps construct a candidate lattice definition of the Standard Model
Physicist Nathaniel Craig reports that GPT-6 Astra helped turn his strategy for a nonperturbative lattice definition of the Standard Model into a construction and took the lead in developing the detailed proof. The work addresses the difficulty of representing the Standard Model’s chiral structure on a finite spacetime grid.
Craig describes the result as a candidate definition, not a completed solution to every foundational issue. Still, the collaboration is a strong example of frontier AI contributing to the proof-development process in mathematical physics rather than serving only as a writing or search assistant.
Source: OpenAI
Ten nations formalize cooperation on AI supply-chain security
The United States and nine Pax Silica partners—Australia, India, Israel, Japan, South Korea, Qatar, Singapore, the UAE, and the UK—signed the San Francisco Declaration on secure technological foundations. The agreement treats economic security and the resilience of critical AI infrastructure as shared national-security concerns.
Five proposed working groups will cover pro-innovation policy, AI adoption in key sectors, cooperation across the supply stack, safe harbors and security hubs, and system resilience. The declaration turns Pax Silica’s principles into a more operational coalition for trusted technology and diversified supply chains.
Source: U.S. Department of State
U.S. Treasury chief urges an outside CEO and compliance overhaul at NVIDIA
Treasury Secretary Scott Bessent called for a “dramatic shakeup” at NVIDIA after Jensen Huang’s resignation, arguing that the board should look outside the company for its next chief executive. He also urged scrutiny of whether other leaders failed to maintain adequate export-control compliance.
The intervention escalates the episode from a CEO succession problem into a broader governance and national-security crisis. The immediate issue is whether advanced AI chips were diverted to China through intermediaries and whether NVIDIA’s internal culture and disclosures adequately addressed that risk.
Source: CNBC
Research Radar
PUMA-Former: Uncertainty-Aware Agentic Computation Offloading in MEO-LEO Satellite Networks With Incomplete Information
Yi Jing, Chunxiao Jiang, Jiawei Wang, and Jiachen Sun introduce an agentic offloading framework for satellite edge networks where CPU reports may be stale or missing. A MEO coordinator combines measurement-age uncertainty, queue pressure, and execution feedback before jointly selecting destinations, workload splits, bandwidth, CPU, and power.
In traffic-driven evaluation, the method increased admitted compute-work completion by 8.62%, reduced mean modeled service delay by 34.78%, and cut episode-P95 delay by 71.08% relative to the strongest stated learning baseline.
Source: arXiv:2611.18139
Space-based Zero-trust Architecture for Securing the Direct-to-Cell Paradigm
Xu Chen, Jiachen Sun, Linan Huang, Wei Quan, Yi Jing, and Xuemin Shen examine how direct-to-cell systems bypass terrestrial gateway perimeters and expose satellites more directly to potentially malicious user equipment. Their architecture moves continuous verification onto satellite compute.
The proposal combines RF-fingerprint authentication, distributed identity management, and behavior-driven dynamic access control. It is especially relevant as commercial D2C systems grow from emergency messaging toward broader data services.
Source: arXiv:2611.17826
PACT: A Decoder-Confidence-Guided Two-Round Receiver for Joint Channel and Phase-Noise Estimation in Sub-THz OFDM
Ruilin Wang and Xiaodai Dong propose a two-round sub-THz OFDM receiver that joins channel and phase-noise estimation with decoder-derived pseudo-pilots. Its complex transformer shares parameters across rounds while using confidence scores to guide refinement.
At 30 dB, PACT reduced code-block error rates by approximately 70% and 57% relative to the strongest learning baseline in two benchmark settings. Frozen-weight tests also showed transfer across channel models, carrier frequencies, and FFT and bandwidth configurations.
Source: arXiv:2611.18450
Source notes
GeekWire returned HTTP 403 for the Amazon Leo failure, so the account was verified against ULA’s quoted statement in Spaceflight Now and multiple independent reports. IEEE Xplore and ACM Digital Library returned no newly indexed eligible papers; the Research Radar therefore uses verified arXiv records from the last seven days.