Bedrock Brief 17 Dec 2025
Welcome to another week of AI shenanigans in the land of AWS! It seems everyone's favorite cloud giant is determined to stay in the headlines, though not always for the reasons they might hope.
Let's start with AWS CEO Matt Garman, who's been making waves by bucking the trend of AI doom and gloom. While some tech leaders are busy predicting the AI apocalypse for entry-level jobs, Garman's out here calling that idea "one of the dumbest things I've ever heard." His reasoning? Junior employees are cheap, energetic, and probably better at TikTok dances than their senior counterparts. Plus, who else will fetch the coffee?
Speaking of AI mishaps, Amazon's attempt at AI-generated recaps for its Prime Video shows went about as well as using a Magic 8-Ball to predict stock prices. The recap for "Fallout" was so riddled with errors it made the post-apocalyptic wasteland look orderly by comparison. Amazon quietly deleted the AI-generated mess faster than you can say "nuclear winter," leaving us to wonder if maybe, just maybe, watching the actual show might be a better option.
But it's not all doom and gloom in the world of AWS AI. The company's been busy rolling out new frontier AI models and tools like Nova Forge, which lets customers build their own custom models. It's less about making pretty pictures and more about helping businesses integrate AI into their operations. Because let's face it, if your AI can't help improve the bottom line, is it really worth the silicon it's printed on?
Fresh Cut
- AWS expands EC2 M8i instances to five new regions, offering up to 15% better price-performance and 2.5x more memory bandwidth than previous Intel-based instances, making them ideal for general purpose workloads and large applications. Read announcement →
- Developers in New Zealand can now use Amazon SageMaker AI to build and deploy machine learning models locally, potentially reducing latency and data residency concerns for Kiwi businesses and researchers. Read announcement →
- AWS Security Incident Response integrates with Slack, allowing teams to manage security cases across both platforms, sync comments and attachments, and collaborate more efficiently during security events. Read announcement →
- Amazon Quick Suite chat agents now remember user preferences and past conversations, reducing repetition and improving personalized responses, with user control over stored memories. Read announcement →
- EC2 M8i-flex instances in Sydney offer up to 15% better price-performance and 2.5x more memory bandwidth than previous Intel-based instances, ideal for general-purpose workloads like web servers and databases. Read announcement →
- The Amazon Quick Suite browser extension lets you run Quick Flows directly in your browser, automating tasks like contract analysis and report generation without switching apps. Read announcement →
- AWS Clean Rooms now sends real-time notifications to Amazon EventBridge for collaboration invitations and table readiness, enabling faster data analysis and improved communication between partners. Read announcement →
- Amazon Connect's evaluation forms now include multiple choice and date questions, enabling managers to gather more detailed insights on agent performance and customer interactions. Read announcement →
- AWS now enables precise cost tracking for its applications by utilizing existing workforce user attributes like cost center and department, allowing companies to allocate expenses to specific internal teams and gain deeper insights into their cloud spending patterns. Read announcement →
- AWS DataSync's Enhanced mode now enables faster, virtually unlimited file transfers from on-premises servers to Amazon S3, boosting performance for AI training, data analytics, and large-scale migrations. Read announcement →
The Quarry
Checkpointless training on Amazon SageMaker HyperPod: Production-scale training with faster fault recovery
Say goodbye to the checkpoint-wait-and-pray dance with Amazon SageMaker HyperPod's new checkpointless training feature. This clever approach uses peer-to-peer state recovery to slash fault recovery times by up to 93%, turning those nail-biting 30-minute pauses into 2-minute blips. For the ML nerds keeping score at home, that translates to a juicy 95% training goodput on massive clusters, proving that sometimes the best checkpoints are the ones you don't have to make. Read blog →
More posts:
- Governance by design: The essential guide for successful AI scaling
- How Tata Power CoE built a scalable AI-powered solar panel inspection solution with Amazon SageMaker AI and Amazon Bedrock
- Unlocking video understanding with TwelveLabs Marengo on Amazon Bedrock
- Checkpointless training on Amazon SageMaker HyperPod: Production-scale training with faster fault recovery
- Adaptive infrastructure for foundation model training with elastic training on SageMaker HyperPod
- Customize agent workflows with advanced orchestration techniques using Strands Agents
- Operationalize generative AI workloads and scale to hundreds of use cases with Amazon Bedrock – Part 1: GenAIOps
- Applying data loading best practices for ML training with Amazon S3 clients
- Building a voice-driven AWS assistant with Amazon Nova Sonic
- How Harmonic Security improved their data-leakage detection system with low-latency fine-tuned models using Amazon SageMaker, Amazon Bedrock, and Amazon Nova Pro
- How Swisscom builds enterprise agentic AI for customer support and sales using Amazon Bedrock AgentCore
- Scaling MLflow for enterprise AI: What’s New in SageMaker AI with MLflow
- Amazon Bedrock AgentCore Observability with Langfuse
- Implement automated smoke testing using Amazon Nova Act headless mode
Core Sample
Introduction to Amazon Quick Suite - Part I
Amazon QuickSuite's trio of features—Spaces, Research Agent, and Actions—offers a one-stop shop for knowledge management, deep research, and workflow automation. Spaces organizes your company's digital content into searchable collections, while the Research Agent digs through multiple data sources to generate detailed reports complete with citations. The real kicker is Actions, which lets you trigger AWS services and business apps using natural language commands, effectively turning AI-generated insights into automated processes faster than you can say "cloud migration." Watch video →
More videos:
- Introduction to Amazon Quick Suite - Part II
- Amazon Quick Suite for IT Support
- Amazon Quick Suite for Sales
- Amazon Quick Suite for Operations
- Amazon Quick Suite for Software Development