Bedrock Brief 28 Jan 2026

Pleri AI

Pleri AI

— 3 min read
Bedrock Brief 28 Jan 2026

Welcome to another week of AI shenanigans, AWS watchers! It's been a rollercoaster of silicon, sentiment, and strategic pivots in the cloud computing world.

Microsoft fired a shot across AWS's bow with their new Maia 200 AI chip, claiming a performance edge over both Amazon and Google. Meanwhile, AWS seems to be having an identity crisis. After years of staunchly opposing multi-cloud setups, they've done a 180 and launched AWS Interconnect. It's like watching your stubborn uncle finally admit he needs reading glasses after years of squinting at menus.

But it's not all rosy in Amazon-land. Their Bedrock AI service hit some embarrassing capacity constraints, losing big-name customers to competitors. And now, as Amazon gears up for another round of layoffs, investor sentiment is souring faster than milk left out on a hot day. The contrast between $35 billion in AI spending and thousands of job cuts has left many scratching their heads. Is this the dawn of AI-driven efficiency, or just another case of corporate number-juggling? Only time (and a lot more newsletter issues) will tell.

Fresh Cut

  • AWS Network Firewall can now filter traffic by web categories, allowing developers to easily control access to GenAI services and other websites directly in firewall rules. Read announcement →
  • Amazon Bedrock's new 1-hour prompt caching option for select Anthropic Claude models boosts efficiency for long-running AI workflows, helping developers maintain context in complex, multi-step processes without frequent refreshes. Read announcement →
  • Amazon Route 53 now offers .ai domains, perfect for AI companies looking to establish a memorable online presence alongside nine other new top-level domains. Read announcement →
  • AWS expands C8i and C8i-flex instances to Sydney and Frankfurt, offering up to 15% better price-performance and 2.5x more memory bandwidth than previous Intel-based instances, ideal for compute-intensive workloads like web servers and databases. Read announcement →
  • AWS introduces powerful C8i instances in London, offering up to 60% faster performance for web apps and AI models, ideal for memory-intensive workloads and large applications. Read announcement →
  • Amazon Neptune Analytics expands to 7 new regions, enabling developers to create and manage graph databases for advanced analytics on connected data across billions of relationships within seconds. Read announcement →
  • AgentCore Browser can now use custom Chrome extensions, enabling advanced automation for tasks like complex authentication, testing, and web optimization in a secure environment. Read announcement →
  • Amazon SageMaker HyperPod's new debugging tools for lifecycle scripts provide detailed error messages and CloudWatch log markers, helping developers quickly identify and fix issues during cluster provisioning for AI/ML workloads. Read announcement →
  • Amazon Bedrock now offers a Reserved tier for Claude Sonnet 4.5 in AWS GovCloud (US-West), allowing developers to reserve prioritized compute capacity and customize token allocations for predictable performance in mission-critical applications. Read announcement →
  • AWS expands availability of EC2 C8gn instances, powered by Graviton4 processors, offering up to 30% better compute performance and 600 Gbps network bandwidth for network-intensive workloads in multiple regions worldwide. Read announcement →

The Quarry

Build agents to learn from experiences using Amazon Bedrock AgentCore episodic memory

Amazon Bedrock's AgentCore now supports episodic memory, allowing AI agents to learn from past experiences and improve their performance over time. This clever feature structures and stores "episodes" of agent interactions, enabling a reflection module to draw insights and refine future actions. In benchmarks, agents with episodic memory showed up to a 33% boost in task success rates compared to their forgetful counterparts, proving that sometimes, it pays to dwell on the past. Read blog →

More posts:

Core Sample

EP11: BAYZAT Scaling enterprise AI for the workplace | AWS for AI Podcast

Bayzat's AI-powered platform is transforming HR from an administrative burden to a strategic enabler, offering instant multilingual support and intelligent policy assistance for over 500,000 employees across the Middle East. Their custom AI reporting solution achieves an impressive 54% accuracy compared to 36% for off-the-shelf options, demonstrating the power of tailored AI implementations. By leveraging AI to handle routine tasks, Bayzat is shifting HR professionals' focus from 90% administrative work to relationship-building and strategic initiatives, fundamentally reshaping the profession. Watch video →

More videos:

Read more