Research

An AI model programmed nonstop for 19 days on a single MirrorCode task that cost $2,600 to run

Epoch AI's new MirrorCode benchmark tests whether AI models can recreate complete programs without access to the original code. Claude Opus 4.7 leads with a 56 …

Build interactive PDF text extraction from Amazon S3

In this post, you’ll build a server that extracts text from PDF files in Amazon S3 in real time. This protocol-based approach provides programmatic document acc…

Production-grade AI agents for financial compliance: Lessons from Stripe

In this post, you learn how Stripe built a production-grade AI agent system for financial compliance. We cover the technical architecture of Stripe’s ReAct agen…

Implementing super resolution by deploying SeedVR2 on Amazon SageMaker AI

In this post, we demonstrate how to implement video upscaling using SeedVR2 on SageMaker AI. We cover the solution architecture, walk through the deployment ste…

Optimize model training on Amazon SageMaker AI with NVIDIA Blackwell

This post shows you how to configure training jobs on Amazon SageMaker AI to get the most out of Blackwell’s architecture on AWS. You learn how to select batch …

Frontiers of compute: The technologies to reduce AI inference costs

AI’s next breakthrough may not be a smarter model but a cheaper token.…

Google bakes computer control directly into Gemini 3.5 Flash, letting the model see and operate your screen

Google has integrated "Computer Use" directly into Gemini 3.5 Flash, letting the model operate computers, browsers, and mobile devices on its own. On the OSWorl…

How agents are transforming work

A new OpenAI research paper shows how AI agents are transforming work, enabling longer, more complex tasks and expanding productivity across roles.…

AI-powered BI with Snowflake and Amazon Quick

In this post, you will learn how to build an end-to-end integration between Snowflake semantic views and Amazon Quick. The sample data is user review data for a…

Snowflake CEO finds GLM-5.2 competitive with Opus 4.7 at a fraction of the cost

Zhipu AI's GLM-5.2 nearly matches Claude Opus 4.7 in a Snowflake benchmark with 103 coding tasks at one-fifth the cost per output token. But the Chinese model b…

How Loka Built a Natural, Low-Latency Voice Agent with Amazon Nova 2 Sonic

In this post, we demonstrate the architecture and approach Loka used to solve a common frustration: robotic, slow voice assistants that cause customers to hang …

Build a protein research copilot with Amazon Bedrock AgentCore

This post shows you how to build a conversational protein research assistant that combines three capabilities: Natural language query parsing to extract structu…

How GPT-5 helped immunologist Derya Unutmaz solve a 3-year-old mystery

GPT-5 Pro helped solve a 3-year-old immunology mystery, offering insights into T cell behavior. The breakthrough could support cancer and autoimmune research.…

OpenAI says new GPT-5.5-Cyber outperforms Anthropic's Mythos on cybersecurity benchmark

OpenAI is expanding its Daybreak cybersecurity initiative with an updated Codex Security plugin, the full GPT-5.5-Cyber model, and a partner network with more t…

Toward More Controllable AI Video Editing: An Early Research Exploration at Netflix

Fact Sheet: President Donald J. Trump Ushers in the Next Frontier of Quantum Innovation

ACCELERATING QUANTUM INNOVATION: Today, President Donald J. Trump signed an Executive Order to supercharge U.S. innovation in quantum technologies and stre…

Anthropic and Micron want to co-design AI memory architecture

Micron is investing in Anthropic's Series H round and getting a multi-year deal to supply memory for Claude's infrastructure. Anthropic co-founder Tom Brown cal…

Embed the world: Multimodal AI for searchable aerial imagery at scale

In this post, we walk through the problem space, our architecture on Amazon Bedrock and Amazon OpenSearch Serverless, the evaluation methodology we built on Ope…

Google Deepmind and A24 team up on AI filmmaking research

Google Deepmind and film studio A24 are entering a long-term research partnership. Google is also investing roughly $75 million in A24, according to the Wall St…

NVIDIA Vera CPU Opens the Way for Agentic Scientific AI at Los Alamos National Laboratory

Mission, Vision and Veritas — new Los Alamos National Laboratory (LANL) supercomputers to be built with HPE and NVIDIA — are tapping NVIDIA Vera CPUs to acceler…

NAIRR Science Program Reshapes Scientific Research, Powered by NVIDIA AI Infrastructure

For the past two years, the U.S. National Science Foundation’s National Artificial Intelligence Research Resource (NAIRR) pilot program has driven innovative re…

Hotter Than a Hot Tub: The 45°C Breakthrough to Cool AI’s Biggest Machines

Hot tubs sit at about 38 to 40 degrees Celsius, warm enough that most people can only soak for about 15 minutes. NVIDIA’s newest AI servers can run their coolin…

AI is inflating student grades, and the effect points to outsourced work, not better learning

A UC Berkeley study of more than 500,000 grades found that courses heavy on writing and coding saw grades jump after ChatGPT launched. The effect shows up mainl…

ChatGPT keeps creeping toward becoming your AI personal assistant with new scheduled task controls

OpenAI is upgrading ChatGPT's scheduling feature. A new "Scheduled" page in the sidebar puts all active tasks in one place, letting users view, pause, edit, or …

Data2Story turns a CSV file into a verified interactive news article using seven AI agents

Seven AI agents work together like a newsroom. The "Data Journalist Agent" from Oxford and Stanford turns a CSV file into a finished interactive article with gr…

President Trump’s Iran Agreement Is America First in Action

President Donald J. Trump and Vice President JD Vance have secured a historic breakthrough by signing a Memorandum of Understanding with Iran. Forged through de…

New benchmark exposes how badly AI struggles with real knowledge work

Even the best AI model fails at realistic knowledge work, fully solving just 3 percent of tasks. The article New benchmark exposes how badly AI struggles with r…

OpenAI researchers show small doses of "beneficial trait" training make AI models broadly safer and harder to manipulate

OpenAI researchers show that reinforcement learning on desired behavioral traits like truthfulness and corrigibility works across domains. Training on health da…

Website "In the Weights" shows whether AI models know who you are

Two former OpenAI employees have built a website called "In the Weights" that reveals which people AI models can recall purely from their training data. A stren…

The smartphone era created an attention crisis. Slowtech is fixing it

“People just really want to take back control of their time, their lives, their attention... They’re down for whatever helps them do that.”…