AI Lab
Official Amazon announcements — model releases, product launches and research, each one summarized with every source covering it, by AIBriefs. RSS
OpenAI cut GPT-5.6 Luna pricing 80% to $0.20/M input and $1.20/M output tokens, and Terra 20% to $2/$12. GPT-5.6 Sol gains Fast mode in the API: up to 2.5x speed at 2x price with unchanged intelligence. The company credits GPT-5.6 Sol's efficiency for enabling the cuts.
Yahoo uses Amazon Bedrock to connect user search intent with relevant ad experiences across channels, improving retargeting efficiency by leveraging user search activity and behaviors.
New monitoring capability tracks ML model prediction quality using Amazon Quick to detect issues before customer impact.
Guide covers migrating prompts to new models and optimizing them for current models on Amazon Bedrock, addressing manual prompt engineering at scale.
Amazon Bedrock AgentCore Identity adds support for Private Key JWT client authentication, allowing agents to authenticate to downstream identity providers using signed JWT client assertions instead of shared OAuth secrets.
AWS blog post demonstrates building an AI agent with Amazon Bedrock AgentCore and MCP servers to autonomously answer business questions from IoT data, using a production-line monitoring example.
Kimi K3 is now available on Perplexity and Nebius, achieving a 1450 Elo score on 3D Design benchmarks. The model ranks 4th on the Agent Arena leaderboard and features optimized inference via a custom DFlash speculator developed with Modal.
AWS AI Blog introduces task-aware knowledge compression to improve RAG for complex analytical tasks across hundreds of documents, addressing limitations of similarity search in cross-document comprehension.
Deepgram integrated AWS IAM Temporary Delegation for self-hosted speech AI on SageMaker, enabling partner engineers to access accounts temporarily without long-lived credentials.
Guardoc uses Amazon Nova models to process medical documents, reducing fragmentation and errors in clinical documentation.
Claude Opus 5 offers performance near Fable 5 at half the price, featuring a 1M token context window. It is now available on the Claude API, Amazon Bedrock, and as the default Opus model in Claude Code v2.1.219.
Article describes a deep learning-based recommendation system using Amazon SageMaker and AWS Glue to predict next product for banking customers. Includes explainability features to interpret model decisions, leveraging customer transaction and demographic data.
Part of a series, this post covers how to use Amazon Bedrock Guardrails with AI-powered coding assistants like Claude Code, Kiro, and others to ensure safe code generation.
Jefferies developed a trade assistant using Amazon Bedrock to provide real-time client behavior, trade pattern, and market trend insights. The AI tool helps traders make split-second decisions from vast data without needing coding skills.
Amazon Bedrock Managed Knowledge Base now supports agentic retrieval to handle multi-part, comparative, and exploratory queries across documents like PDFs and slides. Uses agentic reasoning to decompose complex questions and retrieve relevant context.
monday.com reports that 90% of its builders use AI coding tools monthly, nearly double from a year ago, and per-engineer PR throughput increased by over 50%. The company runs AI Teammates agents on Amazon Bedrock.
The technique proposes a self-distilled reasoning approach for SFT, avoiding the need for costly manual chain-of-thought traces. It uses Amazon Nova models to generate reasoning traces from the model itself.
Guide walks through building agent workflows for supply chain disruption analysis using Amazon Quick and NVIDIA NeMo Agent Toolkit. The solution automates checking purchase orders, inventory, customer commitments, and contract rules.
Couchbase's Capella iQ uses a multi-model AI architecture on Amazon Bedrock to generate database queries and recommend indexes. The system supports multi-turn conversational workflows by routing requests to different LLMs.
Tradeshift, an AI-powered AP platform, uses Amazon Quick to transition from legacy BI to agentic AI. The case study details how they built a multi-agent system for accounts payable insights.
Amazon Quick targets the 60% of sales rep time lost to low-value tasks like CRM updates and prospect research. The agentic AI teammate automates email drafting, context-switching, and research to free up selling time.
Smartsheet built a remote Model Context Protocol (MCP) server on AWS using Amazon Bedrock and AWS Fargate to give AI agents structured access to its work management platform. The architecture enables agents to query project data and trigger actions via natural language.
Grok 4.3 is now generally available on Amazon Bedrock. The model reasons reliably over long inputs, helping teams build agents and AI workflows.
Restaurants miss an average of 150 phone calls per location per month; about 60% are customers trying to place orders or book tables. This tutorial shows how to build a telephony AI host using Amazon Bedrock AgentCore for orchestration and Amazon Nova 2 Sonic for speech, with AWS services like Lambda and DynamoDB for scalability.
Built Technologies, processing over $500B in real estate projects, deployed an AI-powered document intelligence solution on AWS. The system uses Amazon Bedrock to automate document processing and power agents across real estate finance.
AWS AI Blog post demonstrates integrating vision capabilities with Amazon Bedrock and MCP servers to build agentic systems. The approach bridges the gap between systems that see, think, and act, simplifying complex integrations.
AWS blog post describes how to set up cross-account monitoring for SageMaker ML pipelines using CloudWatch dashboards and EventBridge. Includes steps for capturing pipeline events in a central account and visualizing them.
Strands Agents uses multi-agent AI to correlate B2B prospect signals from Reddit, Hacker News, Stack Overflow, and GitHub, turning scattered activity into actionable intent. The solution runs on Amazon Bedrock, stitching together individual signals that would otherwise be noise.
Two AWS blog posts show using Amazon Nova Act with Bedrock and other services to automate QA test execution and user flow analysis. They cover building agentic QA pipelines that gate deployments in CI/CD and a system for UX analysis using Nova Act and Bedrock, with DynamoDB and Lambda for scalability.
Flo Health's engineering team turned a proof of concept from the AWS Generative AI Innovation Center into a production-grade medical content review system. The system uses Amazon Bedrock to scale operations.
The solution uses Amazon Nova Sonic for voice and Amazon Bedrock Guardrails for compliance. It targets the growing AI patient scheduling market.
GPT-5.6 Sol is half the price and twice as token efficient as Fable for many tasks. The models are now generally available on Amazon Bedrock and Perplexity's Agent API. GPT-5.6 Sol also set a new cybersecurity SOTA on 'The Last Ones' cyber range.
An AWS solutions architect describes building an AI-powered desktop tool on Amazon Bedrock to support neurodivergent thinking. The post highlights AI's role in workplace accessibility.
The healthcare compliance solution uses Amazon Bedrock AgentCore to manage 340B Drug Pricing Program data. The post details the architecture including AWS Lambda, Amazon S3, and Amazon CloudWatch.
AWS details how to use the AgentCore Gateway's OBO token exchange pattern to propagate end-user identity through agent calls to downstream APIs. The pattern uses JWT tokens and avoids collapsing audit trails when multiple tenants share the same agent infrastructure.
Up to 20% of insurance claims are initially denied due to missing or low-quality images. Amazon SageMaker AI enables real-time verification at the point of capture. A clinician reviews X-rays to ensure quality before submission.
The post demonstrates building a semantic layer using Stardog's Semantic AI Application over Amazon Aurora and Redshift, and running Strands Agents on Amazon Bedrock AgentCore to query across sources without ETL. It enables answering customer 360 questions across both databases.
The new capability handles routing, escalation, and state persistence across thousands of work items in production. Quick Automate integrates with Amazon Bedrock agents and existing ticketing systems.
KTern.AI used Amazon Bedrock AgentCore to build AI agents that automate SAP transformation workflows. The agents handle reverse engineering, fit-to-standard analysis, and code analysis. This enables enterprise-scale SAP migration with autonomous orchestration.
Separates prefill and decode phases onto different GPU pools via EFA RDMA to eliminate interference. Improves throughput by up to 40% for long prompts and concurrent requests.
Teams often expose APIs as-is, but AWS recommends designing MCP tools with agentic systems in mind. Key considerations include parameter names, error messages, and tool descriptions to improve agent performance.
The AWS AI Blog walks through using BYOKG and GraphRAG with Amazon Bedrock and Neptune to unify scattered pharmaceutical data for more efficient drug discovery. The technique integrates internal lab notes, published literature, and genomics databases.
AWS blog post shows how to build an AI-powered email routing system using Amazon Bedrock. The system classifies incoming messages by urgency and departmental routing using foundation models.
The guide demonstrates integrating Amazon Bedrock AgentCore with Mistral AI Studio to build a production-ready ecommerce MCP server. It reduces custom API code and security risks for faster AI customer experience deployment.
This guide demonstrates how to secure Amazon Bedrock AgentCore API endpoints using AWS WAF. It covers enforcing web application firewall policies, rate limiting, and protection against common web threats. The integration uses Elastic Load Balancing and can be deployed in production.
Three new blog posts cover data modeling patterns, multi-dataset Topic best practices, and building a unified semantic layer for Amazon Quick Sight. Posts provide guidance on runtime joins vs pre-joined datasets and creating a cohesive semantic layer.
Tutorial walks through building a serverless image editing agent using Amazon Bedrock AgentCore harness. The harness handles orchestration, tool routing, memory, and compute via configuration.
The post details monitoring discriminative ML models on SageMaker with MLflow, focusing on metrics drift detection and alerting via Amazon SNS. It includes setup instructions for capturing and evaluating model metrics to maintain accuracy.
The blog post provides a step-by-step guide to building an AI-powered support assistant using Amazon Bedrock AgentCore, integrating CloudWatch monitoring, documentation search, and automated support case filing. It includes code snippets and architectural guidance for AWS engineers.
Claude Code scanned 466 million lines of code in 20 hours for the Government of Alberta, finding vulnerabilities across 1,280 applications. The effort used Opus and Sonnet models, with Minister Nate Glubish calling it 'what responsible government looks like in the AI era.'
AWS introduces a selective unlearning method for Amazon Nova models, allowing removal of specific unwanted concepts (e.g., mature language or cybersecurity threats) while preserving model performance. The technique targets the tension between content moderation and legitimate business uses like summarizing scripts or simulating real-world attacks.
AWS announces MiniMax open-weight models on Amazon Bedrock for production workloads. The integration supports agentic coding assistants and long-context document analysis.
The post covers training agents that query databases, call APIs, and recover from mid-process failures. It describes infrastructure setup on SageMaker HyperPod for multi-step RL workflows.
AWS introduces a new feature using Amazon Nova to automatically detect and redact personally identifiable information (PII) in images. The guide covers setup, configuration, and best practices for integration.
Unit 42 found LLMs hallucinated 250,000 unregistered domains among 2.1 million links. Attackers register these domains to host phishing pages, evading filters due to zero reputation. Different models often hallucinate the same fake domains, making targeting predictable.
AWS now supports NVIDIA Nemotron and OpenAI GPT open-weight models on Amazon Bedrock within GovCloud (US), enabling government agencies to use AI with required security and compliance. The move brings commercial-grade foundation models to regulated workloads.
Enterprises face operational burden with agent-to-agent communication; this post presents a serverless A2A gateway using AWS Lambda, DynamoDB, and API Gateway. The gateway centralizes discovery, routing, and access control, replacing point-to-point integrations.
HippoRAG uses Amazon Bedrock, Neptune, and personalized PageRank to improve multi-hop reasoning in RAG. The AWS blog provides a step-by-step implementation guide.
Fraud appears in 1 of every 16 documents, with AI-generated forgeries up 5x from April to December 2025. Financial institutions use Amazon Bedrock to process applications in seconds, reducing manual review.
BoltzGen is a diffusion-based generative model for designing protein binders to specific targets. Amazon SageMaker AI manages end-to-end GPU infrastructure to accelerate design campaigns.