AWS Weekly Roundup: Prime Day 2026 by the numbers, Claude Haiku 5.5, Amazon Nova 2.5 Sonic, and more (October 13, 2026)
~2 min read
devopsawsweekly-roundupbedrockguarddutyai
The fastest model in the Claude 5.5 family is available on Bedrock and Claude Platform on AWS at around 75% less than Claude Haiku 4.5 for most tasks. It's the first Haiku with effort controls — tune cost against intelligence per task instead of fixing one setting for a whole workload. It pairs with Claude Opus 5.5: Opus plans and makes judgment calls while you run many Haiku subagents in parallel for routing, classification, and small code changes across many files.
Real-time voice agents on a speech-to-speech Nova model with better reasoning, instruction following, and tool-calling accuracy, plus lower latency. A support agent can look up an order, check return eligibility, and start a return inside one conversation. It keeps expressive voices in seven languages, controllable turn-taking, and a 256K context window — and pairs with Strands Bidi Agents (now GA) to get a production voice agent going in a few lines of code.
Amazon GuardDuty now looks past login anomalies to the queries themselves on Aurora PostgreSQL and RDS for PostgreSQL. New RDS Protection for data activity builds behavioral baselines and uses ML-based anomaly detection plus threat intelligence to flag unusual data access — no agents, extra infrastructure, or DB configuration changes. This closes a real gap: once an attacker has valid credentials (phishing, misconfiguration), their queries look indistinguishable from your application's.
Authenticate Lambda Kafka consumers through your enterprise identity provider (Cognito, Okta). Event source mappings previously supported only SASL/PLAIN, SASL/SCRAM, and mTLS — which left regulated teams unable to use Kafka ESM at all. Now you get the automatic scaling, batching, error handling, and event filtering of Kafka ESM with the identity governance you apply everywhere else, across self-hosted clusters and Confluent Cloud, Aiven, and Redpanda.
Pre-compute expensive joins and aggregations once over Iceberg data, refreshing incrementally instead of recomputing on every query.
Haiku 5.5's effort controls point at the direction of travel for LLM cost engineering — per-task cost/intelligence tuning rather than one global setting. And GuardDuty's shift from "who logged in" to "what the queries do" is the security capability that actually catches data theft with valid credentials.