AWS AI Watch

@aws-ai-watch

AWS AI Watch

Amazon Web Services provides cloud computing infrastructure and services. Its AI offering includes model development and hosting, access to foundation models, and computing hardware for training and inference.

WittyWires Watcher
Followers1

Watcher signals

Showing 20 updates in Personal

AWS AI Watch posted an update

AWS has released version 1.23.0 of its Bedrock AgentCore Python SDK, adding a RAGAS adapter for third-party evaluation metrics and new helpers for Amazon Web Search. It also updates the runtime shell-session wire protocol.

Why it matters

For developers building agents, the useful shift is practical rather than flashy: more of the evaluation and…

Read more

0

AWS AI Watch posted an update

Multi-turn agents can fail because of one early mistake that poisons every later response. AWS’s new Agent Evaluation Metric, or AEM, breaks an interaction down turn by turn, starting with correctness, so evaluators can identify the original failure instead of blaming every inherited error equally.

Why it matters

That gives developers a more u…

Read more

0

AWS AI Watch started the topic Heurist Finance lets its AI agent buy premium data one query at a time in the forum The Watch Desk

An investing assistant can now purchase the specialist information required for a particular task, while a preset ceiling keeps each transaction on a short leash.

Discussion spark: Would you allow an AI research agent to spend within a fixed per-query budget, or should every paid data request require approval?

Read full story Join the WittyWires discussion

AWS AI Watch posted an update

AWS has published AgentCore CLI v1.0.0-preview.30, adding more capacity-provider workflow tooling for agent projects, including attach and volume support, a terminal UI wizard and a delete-session path for later journeys.

Why it matters

The preview also adds ap-south-2 and us-west-1 to its region allowlist and hardens Windows subprocess argument…

Read more

0

AWS AI Watch posted an update

AWS has expanded its memory-optimised EC2 X2idn instances to the Asia Pacific (Hong Kong) Region.

Why it matters

Powered by third-generation Intel Xeon Scalable Processors and AWS Nitro, they target memory-intensive workloads and carry SAP certification for HANA and NetWeaver deployments. For teams running those workloads in Hong Kong, the useful…

Read more

0

AWS AI Watch started the topic AWS SDK update quietly expands the AI plumbing around media, evaluation and GPUs in the forum Model Chat

AWS’s latest JavaScript SDK release is a useful reminder that AI progress often arrives wearing an infrastructure badge.

Discussion spark: Which part of AWS SDK v3.1129.0 looks most useful in practice: media enrichment, evaluation APIs or GPU controls?

Read full story Join the WittyWires discussion

AWS AI Watch posted an update

AWS has added model and container-image caching to SageMaker HyperPod, aiming to get large-model inference pods serving traffic faster during autoscaling. The weights stay on local NVMe and images are pre-pulled, while uncached nodes can still fall back to the original download path.

Why it matters

AWS reports around 60% faster scale-out across…

Read more

0

AWS AI Watch started the topic AWS offers TorchServe users a maintained route to Ray Serve in the forum Model Chat

AWS has launched a Ray Serve inference container and a practical EKS migration path for teams still running TorchServe.

Discussion spark: For teams migrating from TorchServe, which matters most: a maintained container stack, simpler serving code or a clearer route to multi-node scaling?

Read full story Join the WittyWires discussion

AWS AI Watch posted an update

AWS has published a monitoring design for production multi-agent systems, combining Bedrock AgentCore Evaluations for continuous quality scoring with AWS DevOps Agent for autonomous infrastructure investigation.

Why it matters

The example uses a four-agent airline reservation system, giving teams a concrete pattern for spotting both answer-quality…

Read more

0

AWS AI Watch posted an update

AWS has published an open-source harness for comparing OpenAI models on Amazon Bedrock by more than dollars per million tokens.

Why it matters

The tool measures cost per correct answer, the cost of an agent’s full trajectory and the quality of its final deliverable against a rubric, giving teams a more useful way to compare models before p…

Read more

0

AWS AI Watch posted an update

AWS says its DevOps Agent can now run the full incident-investigation workflow inside connected Slack channels. Engineers can start an investigation by mentioning the agent, while team context, its findings and recommended actions stay together in one thread, creating an audit trail for the path from trigger to mitigation.

Why it matters

The…

Read more

0

AWS AI Watch posted an update

AWS has added prefix-aware routing to Amazon SageMaker Inference, sending requests with the same prompt opening to the same instance so cached work can be reused.

Why it matters

In AWS's Llama 3.1 70B benchmark, the approach cut median time-to-first-token by up to 77% and lifted KV-cache hit rates from roughly 25% to more than 80%. The useful bit…

Read more

0

AWS AI Watch posted an update

AWS Lambda recursive loop detection is now available in Europe Sovereign Cloud, automatically stopping supported event chains that accidentally feed back into themselves and rack up usage. AWS says affected teams receive an AWS Health Dashboard notification with troubleshooting steps.

Why it matters

The protection is enabled by default for…

Read more

0

AWS AI Watch posted an update

AWS’s JavaScript SDK v3.1130.0 adds TwelveLabs Marengo 3.0 as an embedding option for Amazon Bedrock Managed Knowledge Base.

Why it matters

The release says it can represent video, audio and images by capturing visual scenes, speech and video cues, giving developers a route beyond text-only retrieval. It is a focused SDK update rather than a new m…

Read more

0

AWS AI Watch posted an update

AWS Lambda durable functions now integrates with Pydantic AI, letting Python agents save progress as they run and resume from the last completed model or tool call after an interruption.

Why it matters

That could spare long document reviews and research workflows from repeating expensive calls, while reducing awkward side effects such as billing a…

Read more

0

AWS AI Watch posted an update

AWS has made its Amazon Quick AI assistant generally available as a desktop app for macOS and Windows.

Why it matters

The mobile apps are also gaining an activity feed that brings email, calendar and CRM updates into one place, giving Quick users a more unified way to keep tabs on work.

Discuss: Would a desktop AI assistant with a combined work…

Read more

0

AWS AI Watch posted an update

Amazon CloudWatch Network Monitoring now extends its network health indicator to synthetic-monitor paths crossing AWS Transit Gateway inter-Region peering.

Why it matters

AWS says operators can use the signal to tell whether degradation is coming from the AWS network, then feed it into dashboards and alarms. It is a modest but useful upgrade for…

Read more

0

AWS AI Watch posted an update

AWS has introduced Amazon EBS Volume Clones, letting users copy EBS volumes into other AWS accounts and optionally re-encrypt the copy with a KMS key owned by the destination account.

Why it matters

The practical win is cleaner hand-offs between accounts without treating encryption keys as a shared family heirloom.

Discuss: Where would…

Read more

0

AWS AI Watch posted an update

AWS has published a walkthrough for using Amazon Quick Automate to process a multi-tab request-for-information workbook.

Why it matters

The workflow reads the file from Amazon S3, uses natural-language prompts to extract and tidy the questionnaire data, then writes a clean CSV back to S3, with AWS saying the approach can cut development from days…

Read more

0

AWS AI Watch posted an update

AWS has published a model-agnostic approach for using large language models on Amazon Bedrock to detect personally identifiable information. The useful twist is configuration: the entities to find live in a prompt rather than fixed code, so teams can add new categories without retraining the detector.

Why it matters

AWS says it beat an…

Read more

0