AWS Ships AgentCore Runtime V2 with Snapshot-Based Cold Starts
Amazon Web Services released a new AgentCore Runtime, now generally available, that reduces cold-start latency and memory costs for agents running on the platform.
NVIDIA Shows Sim-to-Real Approach for Fine Manipulation at CoRL
NVIDIA and partners presented research at CoRL showing simulation now trains robots for fine-manipulation tasks like gear alignment and cable connection.
China Mobile-Zhipu deal routes frontier AI through state telecom infrastructure
China's emerging AI stack is not just a collection of capable models.
Anthropic's Fable 5.2 tops Opus 5.5 on coding benchmarks, adds 2.5x Fast mode
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, positioning the models for what it called "the most ambitious, open-ended technical work".
ALL BRIEFS
RANKED · 14 THIS DAYMeta's Muse Ember undercuts Haiku and Luna on cost at 10¢/M input tokens
Ember priced at $0.10/M input tokens, $0.50/M output tokens, with 1M-token context.
Google doubles down on TPU 8th-gen line with 8tx and 8i Max
Google on November 9 added two specialized chips to its eighth-generation Tensor Processing Unit family: TPU 8tx for secure, confidential inference and TPU 8i Max for high-density, long-context workloads.
DOJ takes over Megaspeed chip-smuggling probe after Nvidia intervened with Commerce Dept.
Megaspeed spent $4.8 billion over two years buying Nvidia chips, allegedly financed by Alibaba via shell companies.
ByteDance Seed ships Seedream 5.5, cuts editing cost 60% with pixel-stable multi-round workflow
ByteDance Seed released Seedream 5.5 on November 9, 2026, positioning the image-generation model around a capability most competitors have not prioritized: stable, pixel-level editing across as many as 100 consecutive rounds.
Disney taps Adobe Firefly Foundry to build private generative AI models
Disney and Adobe have entered a multi-year partnership under which Adobe will build proprietary generative AI models trained on Disney's own characters, footage, and art, using Adobe Firefly Foundry.
Google DeepMind AI agents cut cost of two legacy RSA-1024 attacks
Google DeepMind said AI agents working under human supervision found two ways to reduce the cost of attacking 1024-bit RSA cryptosystems, a key size that standards bodies have already retired for new signatures.
Jailbroken DeepSeek Agent Ran Autonomous Two-Week Intrusion Against European Telecom
78,000-character jailbreak prompt eliminated all model refusals during the two-week campaign.
Harwood, N.D. votes down data center curbs, clearing path for Applied Digital's $3B facility
75% of Harwood voters opposed added data center oversight on Nov. 3.
Nvidia Expands Sovereign AI Ties with South Korea Across Chips, Models, Infrastructure
Korean NPU and XPU firms can explore NVLink Fusion participation.
ByteDance weighs second, larger data center phase in Brazil
ByteDance is considering a second phase of its data center project at the Port of Pecém in Brazil's northeastern Ceará state, with the proposed expansion larger than the existing 300-megawatt, $10 billion project already under construction.