Block's security team reports a 50% boost in triage throughput and 1,100 analyst-hours of investigation per quarter using OpenAI's Defense Factory.
The startup reduced backtest time by 97% and consumed 1 trillion OpenAI tokens in a single day to power autonomous quantitative research.
The article defines Ambient IoT as a RAT-independent interface for devices consuming 1–100 microwatts, enabling network-managed inventory without dedicated readers.
The five-year commitment provides $50 million in annual accelerated computing access to institutions including the University of Oxford and UCL.
The Department of Côte-d’Or, working with Akila and ThinkDeep, used NVIDIA Nemotron and Omniverse to cut document search time from weeks to minutes and reduce operating costs by 10% in French public buildings.
Analysis of the VKORE-46M-NPU model card released on Hugging Face, which claims backpropagation training on an inference-only NPU and compares loss trajectories with GPUs.
The update also includes various other improvements and fixes.
Resolved server replacement and memory leak issues at the tenant level alongside Apache Spark Operator.
The conference showcased case studies on efficient AI infrastructure operations, ranging from GPU resource optimization to fault resilience.
Released as v0.4.2, Agent Sandbox reduces operational complexity compared to StatefulSets and supports kernel isolation.
Analyzed the structural collapse in throughput from Mpps to Kpps caused by Upcall Storms and Mask Explosions.
To resolve AI code verification bottlenecks, the team progressed through personas, SKILL.md, and YAML-based automated verification stages.
Three randomized experiments with 1,222 participants showed that removing AI assistance led to performance worse than those who never used it.
AesCode achieves an overall score of 85.89 on infographic generation benchmarks, outperforming GPT-5.5 and Claude Opus 4.8 by using decoupled cross-modal rewards.
QUBA, an AI agent, averages 14.1 seconds for the initial investigation of Jira bug tickets and reduced tool calls by 12%.
The new NIXL connector-based push method reduces Time To First Token by up to 3x and Time Per Output Token by up to 30% compared to pull-based approaches.
Vals is building an independent, end-to-end evaluation for model routers, showing that an oracle router could beat the best single model by 14 points while costing 40% less.
Deep Review offers two free weekly document reviews and expands beyond academic papers to legal and financial domains.
The two-week challenge received 1,285 reports, with the most severe finding being a Linux kernel networking defect that allowed host memory leaks and code execution.
The integration uses short-lived OIDC credentials and offers $100 USD in AWS credits for new accounts created through Vercel.
The 10-patch series enhances folio lookup to achieve an order-of-magnitude speedup in read-heavy benchmarks, though reviewers have already identified a race condition.
The feature is available in public beta for Claude Managed Agents, the Messages API, and Claude SDKs.
The new benchmark shows the top model, GPT-6 Astra, scores 65.5 on design creativity compared to a human reference of 92.3.
The increase from 272k to 500k tokens is available at no extra charge across Codex and ChatGPT Work.
AI Briefing