Astra for Law leads in legal intelligence
- Ranked first on Harvey LAB-AA with 20.8% full credit, ahead of GPT-6.2 Astra and Muse Wonder Full.
- Lowest material hallucination rate: 0.03 per assignment, far below Claude Opus 5.5's 0.59.
Daily
10 selected stories for 2026-11-17, grouped by topic.
Dates use UTC+8. The latest edition uses a day with at least 5 completed reports when available; until then an earlier edition may appear. There is no fixed delivery time.
Estimated briefing time 4 min
You have reached the end of this edition.
Your reading listNVIDIA's Ian Buck opened the company's SC26 week in Chicago with Nemotron for Science, the Rubin HPC GPU, an expanded Vera CPU platform and a quantum computing collaboration with IBM and Fermilab, alongside real-world agent deployments in research.
OpenAI launched ChatGPT for Sales on November 17, 2026, featuring a new Sales plugin and more than 30 new and expanded partner plugins to help sales teams shorten the sales cycle by unifying tools, customer context, and workflows in ChatGPT.
Duco rebuilt its agentic AI stack on Amazon Bedrock and AgentCore, achieving strict tenant isolation, a multi-model strategy, and an MCP skills architecture that cut context overhead by 90%, reducing reconciliation deployment time 10x and onboarding costs by 70%.
Multi-agent coordination is a distributed systems problem, not a model problem. This post gives a field-tested taxonomy of multi-agent design patterns and anti-patterns, a framework for deciding when multiple agents are warranted, and how to build cost-efficient, reliable systems on AWS.
LangChain CEO Harrison Chase explains the shift from task-specific agents to goal-focused agents, driven by two practical changes: specifying outcomes instead of actions, and ongoing responsibilities instead of bounded assignments.
LangSmith reorganized its platform around a single agent record, unifying the build, run, improve, and govern lifecycle for agents across any framework or runtime, addressing fragmented workflows and governance gaps.
Databricks introduces a new integration using OpenSharing to connect AWS HealthLake FHIR data directly to Databricks as a shared catalog, eliminating export pipelines and data duplication. Users can govern data via Unity Catalog and explore it with natural language through Genie One.
MoEngage built Merlin AI on Amazon Bedrock and AgentCore, converting existing OpenAPI specs into MCP tools and keeping tenant data isolated while giving agents secure access to live platform data. One customer saw a 38% uplift in push click-through rates.
This post presents a multi-stage, AI-powered methodology using Amazon Bedrock, Kiro IDE, and AWS Transform to extract business rules, convert Assembler through an intermediate COBOL stage, and deploy modernized Java Spring Boot applications on AWS.
Databricks explains how to turn team workflows into reusable, governable AI skills in Genie One, ensuring consistent output formats and checks.
Amazon Bedrock Managed Knowledge Base now supports tabular search, enabling agentic retrieval to answer questions over CSV, XLSX, and XLS files by running calculations across entire tables and combining results with document facts.
NVIDIA expands its CUDA-X libraries with new modules like cuPhoton, cuPotential, and cuSpa, providing AI agents with tools to accelerate scientific discovery. These libraries significantly speed up data processing across astrophysics, materials science, and single-cell biology, enabling researchers to analyze larger datasets and model more complex systems.
At SC26, NVIDIA unveiled Rubin HPC, a new GPU in the Vera Rubin platform designed to accelerate simulation and data processing for science, engineering, and analytics workflows. Featuring over 50,000 CUDA cores, sparse data acceleration, and 7th-gen NVLink, it is expected in 2028.
OpenAI is working with NATO and allied cyber agencies through its Daybreak initiative to provide advanced AI models for vulnerability discovery, fix testing, and cyber defense. NATO has used up to 300 billion tokens per month, and agencies like ENISA and CERT Polska have leveraged the technology to find and fix multiple vulnerabilities.
Databricks explains tabular foundation models (TFMs): pretrained models that predict on new datasets without task-specific training, accelerating predictive ML experiments. The article covers how TFMs work, the scale and licensing differences among leading models, and the GPU cost and operations trade-offs, and shows how Databricks AI Runtime supports TFM experimentation and fine-tuning with managed GPUs, MLflow tracking, and Unity Catalog governance.
OpenAI is expanding its ChatGPT for Academic Researchers program to offer up to 200,000 researchers worldwide free access to its frontier models. A new round of 50,000 first-come, first-served grants opens November 23, each covering the recipient and up to two collaborators. Pilot data shows researchers estimated work would have taken nearly twice as long without AI.
NVIDIA's Vera Rubin platform swept every MLPerf Training v6.1 benchmark and, in its first MLPerf Endpoints submission, delivered more than 10x generational gains in throughput per watt for agentic AI. Vera Rubin NVL72 scaled from 256 to 1,024 GPUs with about 97% efficiency on the DeepSeek-V3 benchmark.
NVIDIA operations leader Deb Shoquist explains how deep supplier relationships, disciplined prioritization, and cross-team collaboration turn cutting-edge chip designs into mass-produced AI systems, while building new supply chains for optical interconnects and other emerging technologies.
Lambda is using NVIDIA Photonics co-packaged optics (CPO) switches to connect NVIDIA GB300 NVL72 racks with more than 10,000 GPUs, delivering lower power, higher resiliency and longer mean time between interruptions (MTBI) to an AI factory at scale.
At SC26, NVIDIA marks two decades of CUDA by honoring the developer community and highlighting milestones from molecular dynamics to AlexNet and AlphaFold, with over 11 million developers now building on the platform.
OpenAI released a new image classifier to detect CSAM, introduced behavioral warnings in ChatGPT, published developer guidance, and expanded support for child-safety law enforcement.