Topic: #amazon

The Amazon SageMaker AI Spaces add-on for Amazon EKS runs managed JupyterLab and Code Editor environments on the cluster your ML team already operates. This post shows how to install and configure the add-on, connect from the browser and from VS Code over SSH-over-SSM, and move your team to OpenID Connect sign-in with Amazon Cognito.

Amazon is investing in a new gas-fired power plant in Pecos County, Texas, built to supply its own data center with up to 7.65 gigawatts of electricity. The facility's 35 natural-gas turbines would initially run disconnected from the Texas grid, tying output directly to the data center's demand. The New York Times reports it could rank among the largest single sources of greenhouse gas emissions in the US.

TReNDS, a research center at Georgia State University, built an agentic AI pipeline on Amazon Bedrock and the open-source Strands Agents SDK that automatically investigates production errors in real time, reducing root-cause analysis from 15 to 30 minutes of manual work to under 60 seconds.

Ahead of the US midterms, rules for AI deepfakes in political campaigning differ sharply from state to state. Twenty-nine states have deepfake election laws in effect, while California's and Hawaii's have been permanently enjoined, according to the National Conference of State Legislatures. Minnesota and Texas ban political deepfakes only in the days before an election, while Maryland's ban runs year round.

Writing in the Guardian, Alan Finkel proposes three laws for AI modelled on Isaac Asimov's laws of robotics. One trigger is Elon Musk's prediction that AI-powered robots will shape the physical world and may eventually stop taking orders from people. Musk's own counter-vision is a shared objective that aligns AI with a love of truth and human flourishing, enforced by governments if necessary.

AI data center company Firmus has raised $2 billion in a new funding round, according to the company. Backers include Coatue Management, Nvidia, Blackstone vehicles and Jane Street. The size of the round shows how hard investors are betting on additional compute capacity for AI.

Learn how to configure rate limits on Amazon Bedrock AgentCore gateway to enforce per-user and per-target traffic controls. Define request, token, and connection limits scoped by JWT claims or IAM identity to protect downstream models, tools, and agents from traffic spikes.

Learn about new capabilities in Amazon Bedrock AgentCore: temporal policies powered by Dogwood, a new open source policy language for AI agents, and rate limiting on the gateway. These features give you deterministic control over sequences of agent actions and cost ceilings that hold regardless of agent behavior.

A regulated customer needed all Claude Code inference processed in a single AWS Region (London), not just in-geography. This post shows two ways to pin Claude Code on Amazon Bedrock to one Region: an application inference profile or the Mantle endpoint, paired with an IAM Region condition, plus how to verify compliance in AWS CloudTrail.

OpenAI has asked a federal judge to dismiss Apple's trade secrets lawsuit, calling the allegations meritless. In its motion, OpenAI argues that Apple mischaracterises ordinary employee conduct as theft and generic product development information as trade secrets. It also claims Apple took no reasonable steps to keep that information secret in the first place.

AI agents on Amazon Bedrock AgentCore run in the cloud, but users' tools and files live on their laptops. Learn how to build a secure MCP bridge that lets a cloud-hosted agent call local MCP servers by tunneling signed messages over the existing WebSocket connection through a browser extension and Chrome native messaging, with no open ports or VPN required.

OpenAI calls Apple's lawsuit careless, aggressive, and oddly personal, and is going on the offensive. In a blog post titled Apple is getting this wrong, the company published iMessage and email exchanges meant to undercut central allegations in the case. This is not a legal response but an attempt to shift public opinion.

Formula 1 partnered with AWS to build the Data Accelerator, using agentic AI on Amazon Bedrock AgentCore to rebuild its MarTech data platform. Onboarding a new data source dropped from up to eight weeks to roughly 40 minutes, schema evolution is handled automatically, and the fan engagement data estate now has end-to-end observability.

An internal presentation revealed that a failed AI deployment cost Amazon $1.8 million on a menial coding task, running 860% over budget. A couple of other projects resulted in hundreds of thousands of dollars in extra AI expense, and the issue went undetected for several months.

OpenAI GPT-5.6 Sol, Terra, and Luna are now generally available on Amazon Bedrock, along with explicit prompt caching that gives you precise control over which parts of your prompt are cached and reused. Learn how to get started, set up explicit caching, and migrate existing GPT workloads to reduce inference cost.

Amazon Bedrock AgentCore delivers autonomous, cross-system business intelligence through configuration rather than custom code. Using pre-built MCP server connectors, fine-grained access control, and persistent memory, enterprises can query multiple data sources with natural language while role-based boundaries are enforced automatically.

Open-source questions stir frank discussion – and both sides have clear economic incentives for where they land Hello, and welcome to TechScape. This week we’ll be looking at a debate over the future of artificial intelligence that’s dividing the tech industry, as well as how the European Union gave Google a slap on the wrist for anti-competitive behavior.

Traditional RAG hits a ceiling on analytical tasks that span hundreds of documents. This post shows how to use task-aware knowledge compression (TAKC) on AWS to pre-compress entire knowledge bases into task-specific representations, cache them at multiple fidelity tiers, and route each query to the right tier, with an open-source implementation you can deploy.

In this post, we cover why Deepgram built on IAM temporary delegation, how the integration works end-to-end, and what it unlocks for customers running Deepgram speech models on SageMaker AI. With this integration, Deepgram has reduced the time for initial investigation on a SageMaker AI support ticket from days to minutes.

Learn the architecture and design decisions behind an explainable next-best-product recommendation system for banking, built with Amazon SageMaker AI and PyTorch. A multi-tower neural network with learned attention delivers accurate, per-customer recommendations while providing the explainability that banking regulators require.

Amazon is rolling out an Alexa Plus update that connects the assistant to smart home devices from Bosch, Delta, Ecovacs, iRobot, Yale Home, Whirlpool, Tapo and Eufy, automatically routing requests to the right device. In Amazon's example, a user can say their kid's soccer jersey needs a deep clean but the tag says cold wash only, and Alexa Plus picks the correct washer cycle.

Together, Motorway and AWS built an end-to-end evaluation pipeline that reduced incorrect results from 1 in 8 queries to 1 in 50 and cut issue detection time from hours to minutes. The pipeline combines the Strands Agents SDK with Amazon Bedrock AgentCore, a fully managed service for deploying and operating AI agents at scale. This post shows you how to build the pipeline for your own agents.

This post explores how Jefferies solved its front-office trading challenges with a solution built on Strands Agents, an agent harness SDK for AI agents that reason, plan, and act by orchestrating calls to foundation models and external tools. It uses LLMs, Amazon Bedrock, Bedrock Knowledge Bases, and the Model Context Protocol (MCP), covering the architecture, technology choices, lessons learned, and business impact.

The settlement was finally approved by a U. federal judge, with a majority of the plaintiffs accepting the amount. A few members of the group refused, citing the small amount compared to the number of infringed titles, and are pursuing a separate lawsuit of their own.

Sony Music Entertainment has filed another lawsuit against Udio, accusing the AI music generator of infringing the copyright of more than 30,000 of its songs, ranging from Elvis Presley's Hound Dog to Beyoncé's Say My Name, and Harry Styles' As It Was. The lawsuit, filed in a New York court on Monday, claims that this list represents only a small portion of the plaintiffs' works that Udio infringed.

Excel data analysis can become challenging, particularly when dealing with extensive or disorganized datasets. Kenji highlights how AI, specifically Claude, can simplify this process by following a structured six-step framework. Tasks such as profiling, cleaning and exploratory analysis are addressed, with a World Cup dataset serving as an example.

Company aims to develop AI software that cuts research times and use of rare metals in chipmakers’ supply chains Business live – latest updates Amazon’s founder, Jeff Bezos, and the UK government have invested in a £2bn British artificial intelligence startup that is aiming to become the “search engine for rare materials” that accelerates the next wave of technological breakthroughs. The Cambridge-based CuspAI has raised $450m (£330m) in funding from in…

In this post, we show you how to build a voice ordering system that answers a phone number and takes the order from greeting to confirmation. The system uses Amazon Bedrock AgentCore to host and run the agent and Amazon Nova 2 Sonic for real-time speech, connected to a restaurant backend through the Model Context Protocol (MCP).

When the PM talks about new laws applying to the ‘next generation of large-scale datacentres’ what does he mean? Albanese’s AI blueprint sparks calls for datacentre moratorium until new regulations in place Expectations were high as the prime minister took the stage at the University of Sydney on Wednesday to outline a pivot in his government’s approach to artificial intelligence.

Built partnered with the AWS Generative AI Innovation Center (GenAIIC), AWS Partner AND Digital, and AWS account teams to create a scalable, AI-powered document processing engine that can classify, split, extract, evaluate, and reason over complex real estate finance documents. It reduces workflows that previously took days to minutes, supports hundreds of document types, and gives technical teams and industry experts a shared environment for building a…

Group of major publishers accuses the tech giant of ‘one of the most prolific infringements of copyrighted materials in history’ A group of major publishers have filed a lawsuit against Google, accusing the company of illegally using millions of copyrighted books to help build its Gemini artificial intelligence models, in “one of the most prolific infringements of copyrighted materials in history”. The case, filed in federal court in New York, has been…

In this post, we extend that foundation to demonstrate how QA Studio addresses batch regression testing and pipeline integration through test suites that organize and parallelize execution, and a command-line interface that brings agentic testing into automated CI/CD pipelines.

The AI revolution is here, and with it a fear that soon it will replace many of us in the workplace. The Australian government is grappling with how to deal with the multi-layered disruption, but so far reform has been slow as it weighs up regulation against the claims of investment opportunities an AI boom presents.

Anthony Albanese will deliver a landmark speech on AI this week as MPs are torn between attracting datacentre investment and protecting the rights of creatives Get our breaking news email, free app or daily news podcast When Anna Funder stood before a pack of journalists at Parliament House earlier this month, she presented herself not just as a writer but also a “victim of crime”. The Stasiland author was using the analogy to illustrate how technology…

This post describes how Henry Schein One closed that gap by building Image Verify, an AI-powered quality verification system on Amazon SageMaker AI that evaluates dental X-ray quality at the point of capture, in real time, across thousands of locations. The system went from concept to over 10,000 active locations within months and has already processed over 11 million X-rays and growing at 1.5 million per week.

In this post, we show you how to combine case management with agentic automation capabilities in Quick Automate. We introduce case management and explore the lifecycle of cases in an agentic workflow from case creation through processing to resolution.

- Microsoft’s 2026 environmental report shows the cost of the AI buildout: total greenhouse gas emissions rose 25% in fiscal 2025, mainly due to new data-center infrastructure and a shift in electricity procurement. - The sharpest signal is purchased-electricity emissions: reported figures jumped 945%, while electricity use rose 24%.

- AWS shows an end-to-end blueprint for an ecommerce MCP server on Amazon Bedrock AgentCore, connected to Mistral AI Studio Vibe. - The server uses Python and FastMCP, runs as a stateless container in AgentCore Runtime, and exposes tools for product search, orders, reviews, returns, and order history. - Data sits in five DynamoDB tables; Cognito handles OAuth 2.1 identity.

- Google and Amazon are reporting rising emissions despite net-zero pledges: The Guardian cites Google up 25 percent year over year and Amazon up 16 percent. - Microsoft had already reported a 23 percent rise versus its 2020 baseline in 2025; Meta showed a 64 percent year-over-year jump despite its 2030 target.

- Three residents of Sturtevant, Wisconsin, have filed a class-action lawsuit against Microsoft over its $7.3 billion Fairwater data center in nearby Mount Pleasant. - The complaint alleges private nuisance and negligence. Residents say diesel generators, HVAC systems, chillers, cooling towers and fans create constant, excessive noise on their properties.

- Australian digital content designer Jodie Heenan made „Guardians of the Burrow“, a short wildlife documentary that is fully AI-generated while aiming to feel like traditional nature TV. - The film shows an Amazonian tarantula and a tiny dotted humming frog inside an underground burrow.

- Anthropic’s new Mythos and Fable models were offline for about 20 days after Amazon flagged a possible jailbreaking flaw to the U. government, Axios reports. - The Trump administration responded with broad export controls.

- Australia is facing pressure over a reported package in which Big Tech would invest more than $50bn in datacentres and create a $350m fund for creatives. - In return, tech companies want weaker copyright rules so they can use Australian music, journalism and books to train AI models. - Anthony Albanese’s government says it has no plan to weaken copyright.

- AWS outlines a Bedrock workflow for AI-generated phishing emails: SPF, DKIM and DMARC run first, then a model checks word choice, communication style and whether the request fits the context. - The system builds sender baselines: how a contact usually writes, what they normally ask for and who they communicate with. A first-ever payment change request gets treated as higher risk.

- AWS outlines how to make multi-turn reinforcement learning in SageMaker AI more reliable: build a reproducible sandbox first, set up external evaluation, then design rewards and train. - The post focuses on agents that use tools across several steps, such as support or moderation workflows. AWS argues that live systems are a bad training target because rollouts can cause side effects and unstable metrics.

- Amazon Bedrock now offers OpenAI GPT OSS and NVIDIA Nemotron in AWS GovCloud (US): gpt-oss-120b, gpt-oss-20b, plus Nemotron Nano 9B v2, Nano 12B v2, Nano 30B, and Super 120B. - The models run inside the GovCloud boundary. In-Region inference is available in us-gov-west-1, while Geo Cross-Region routing spans us-gov-west-1 and us-gov-east-1 without using commercial AWS Regions.

- Meta is reportedly planning a cloud infrastructure business called Meta Compute, selling access to AI compute and potentially hosted models. - That would put Meta up against AWS, Google Cloud, and Microsoft Azure, while following SpaceX/xAI’s move to lease excess data center capacity to Anthropic, Google, and Reflection AI.

- Google’s new Google Home Speaker is its first smart speaker in six years: $99.99, compact, good-looking, solid sound for its size, with Matter controller and Thread border router support. - The hardware is the stronger part of the product. The speaker hears commands well, fits into rooms easily, and can pair with the Google TV Streamer, though its bass and overall audio trail larger speakers.

- AWS outlines five resilience patterns for GenAI apps on Amazon Bedrock: cross-Region inference, multiple AWS accounts, an LLM gateway, model fallback, load balancing, and multi-tenant quota isolation. - Cross-Region Inference automatically spreads requests across available Regions to reduce the impact of regional quotas and traffic spikes.

- PAR outlines a production-grade text-to-SQL analytics agent on AWS for restaurant businesses, designed to separate tenants, businesses, admins, and location-level permissions. - The system uses three independent layers: AWS SigV4 for signed requests, Amazon Bedrock for semantic validation, and Split-Plane SQL for deterministic row-level data isolation. - The LLM never sees the raw Databricks schema.

- Axios reports that the Trump administration is close to allowing Anthropic to restore access to Fable 5. The model has been offline for 15 days after government security concerns. - One source says the limits could be lifted as soon as next week. Talks are expected to continue over the weekend, and Anthropic reportedly expects access to return soon.

- AWS presents Cara as a domain-specific AI platform for large insurance brokerages, built to automate back-office work such as applications, coverage comparisons, proposal creation, and renewals. - The system runs on Amazon EKS across multiple Availability Zones and uses Amazon Bedrock for LLM inference, avoiding the need to manage dedicated GPU infrastructure.

- Geeky Gadgets reports speculation that Fable 5 could return, but the article says there is no official confirmation yet. - The potential comeback is framed mainly around enterprise needs: cloud integration, access controls, scalability and data security.

- Amazon-owned MGM dropped Luca Guadagnino’s OpenAI drama Artificial while it was nearly finished. The film reportedly portrayed Sam Altman unfavorably, raising questions about how tightly Amazon, OpenAI, and Hollywood interests now overlap.

- AWS presents „agentic overlays“ as thin wrappers that make existing REST services usable in A2A interactions while exposing REST endpoints as MCP-compatible tools. - The main idea is retrofit over rebuild: keep business logic unchanged, add agent-facing routes such as /. json and /a2a, and reuse the existing deployment path.

- In a sponsored IEEE Spectrum article, Capital One explains why Prem Natarajan moved from leading Amazon Alexa AI to becoming Chief Scientist at a bank: AI in finance is framed as research, not just model deployment. - The article points to real-time fraud detection, personalized customer guidance, agentic customer service and privacy-preserving model training as areas where generic foundation models fall short.

- AWS explains how to tune Amazon SageMaker AI training jobs for NVIDIA Blackwell: batch size, sequence length, precision format and activation checkpointing are the main levers. - The examples use P6-B200 instances with 8 Blackwell GPUs and PyTorch FSDP, focused on transformer models from 1B to 64B parameters.

- AWS shows how to run SeedVR2 video super resolution on Amazon SageMaker AI: upload video to an S3 input bucket, trigger Lambda, run a SageMaker processing job on an ml. g5.4xlarge GPU, and write the output back to S3. - The architecture is deployed with AWS CDK and includes VPC, IAM, KMS, encrypted S3 buckets, an ECR container, CloudWatch logs, and ComfyUI as the SeedVR2 inference layer.

- AWS shows an end-to-end workflow that loads movie review data from Amazon S3 into Snowflake and turns it into a semantic view with SQL. - That view defines relationships, dimensions, and metrics at the data layer, so BI dashboards and AI queries use the same business logic.

- NVIDIA and AWS are packaging several production AI infrastructure pieces: Amazon EC2 G7 with NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs, OpenSearch Serverless with NVIDIA cuVS, and validated GB300 training performance. - EC2 G7 is positioned as a step up from G6, with up to 4.6x AI inference performance, up to 2.1x graphics performance, and faster GPU analytics via cuDF on Amazon EMR.

- Netflix, A24, Focus Features, and Warner Bros. ’ Clockwork have reportedly passed on distribution deals for Luca Guadagnino’s Artificial, a Sam Altman biopic. - Amazon MGM dropped the nearly finished film from its slate, despite reported plans for a 2026 awards run and a wider 2027 release.

- AWS presents Ampersend as a pay-per-intelligence stack for AI agents: an agent chooses a model tier through Ampersend, pays per request, and receives the result. - The payment layer uses Amazon Bedrock AgentCore Payments, x402, USDC on Base, and wallet providers such as Coinbase CDP or Stripe Privy. The agent never handles private keys.

- AWS outlines a search architecture for Vexcels aerial imagery: tiles are embedded via Amazon Bedrock, indexed in Amazon OpenSearch Serverless, and queried with natural language. - The benchmark used OpenStreetMap as ground truth for Grant Park in Chicago and compared about 100 configurations across two query types: swimming pools as discrete objects and roads as distributed infrastructure.

- AWS published a CDK blueprint for running ComfyUI as a batch pipeline on SageMaker AI processing jobs: Lambda triggers the job, ECR serves the container, and outputs stream to S3. - The sample uses Z-Image Turbo in a ComfyUI workflow, six ml. xlarge GPU instances, 125 GB volumes, VPC private subnets, KMS encryption, and CloudWatch logs.

- The US government forced Anthropic on June 12 to shut down Claude Fable 5 and Claude Mythos 5 worldwide. The official framing was export control and national security, not just access by specific foreign users. - The trigger was reportedly a non-public Amazon paper claiming researchers had found a way to use Fable 5 for security tasks despite its guardrails.

- Amazon MGM has reportedly dropped Luca Guadagnino's OpenAI film Artificial and is working with the filmmakers to find another studio. - The movie was set to cover the five chaotic days in 2023 when Sam Altman was fired as OpenAI CEO and then reinstated. - Andrew Garfield was attached as Altman, with Monica Barbaro as Mira Murati, Ike Barinholtz as Elon Musk, and Yura Borisov as Ilya Sutskever.

- AWS is adding more than 100 detailed inference metrics for SageMaker AI in CloudWatch, covering GPU use, GPU memory, KV cache pressure, token latency, traffic distribution across Availability Zones, cold starts, and inference component placement. - New SageMaker endpoint configurations enable detailed observability by default.

- Amazon Bedrock AgentCore harness became generally available on June 18, 2026 and promises production agents through two API calls: CreateHarness to define one, InvokeHarness to run it. - The agent runs in an isolated environment with a filesystem and shell, can read files, execute commands, write code, and call external tools through Gateway or MCP.

- Amazon software engineers Patrick Schloesser, Darius Irani, and Liesl Wigand testified before Seattle City Council in favor of limits on large data centers, citing a local law that protects employees from discrimination over political speech. - On June 10, one day after Seattle passed a one-year moratorium on new large data centers, Amazon called the three into Employee Relations meetings.

- AWS announced inline payload support for Amazon SageMaker AI Async Inference on June 17, 2026, letting small inference inputs travel directly in the InvokeEndpointAsync request body. - For eligible jobs, teams no longer need to upload every input to Amazon S3 first and then pass an InputLocation reference into the async endpoint. - The raw inline payload limit is 128,000 bytes.

- AWS used its New York Summit to announce AWS Context, a coming service that maps relationships across data lakes, warehouses, databases, streams, and internal knowledge into a managed knowledge graph. - Agents are meant to query that graph at runtime through agentic search and MCP. Access is tied to IAM and Lake Formation permissions, so queries can be governed and audited.

- AWS introduced InvokeGuardrailChecks for Amazon Bedrock Guardrails, letting developers call individual safety checks inside agentic workflows without creating or versioning guardrail resources first. - The API is detect-only. It does not block or mask content by itself, but returns scores that apps can use to decide whether to block, retry, escalate, log, or allow a step.

- SpaceX rose another 4.8% on Tuesday and reached a $2.659 trillion market value, according to FactSet. - In its first two full trading days, the company added about $537 billion in market cap. - That put SpaceX just ahead of Amazon.

- AWS is bringing P-EAGLE into SageMaker JumpStart, letting compatible models run as real-time endpoints with a pre-trained drafter head and no custom containers or manual drafter training. - Launch support covers GPT-OSS-120B, GPT-OSS-20B, Qwen3-Coder-30B-A3B-Instruct, and Gemma-4-31B-IT. The walkthrough uses Qwen3-Coder.

- SpaceX has agreed to buy Anysphere, the company behind AI coding tool Cursor, for $60 billion shortly after its Nasdaq debut. Anysphere investors are set to be paid in SpaceX stock, with closing expected in the third quarter of 2026. - SpaceX shares jumped after the IPO.

- AWS is adding Google DeepMind's Gemma 4 family to Amazon Bedrock: three instruction-tuned Apache 2.0 open-weight models named Gemma 4 31B, 26B-A4B, and E2B. - All three variants support text and image input, built-in reasoning mode, and native function calling. The larger 31B and 26B-A4B models offer context windows up to 256K tokens.

- Geeky Gadgets frames GPT-6 as OpenAI’s attempt to regain momentum after Microsoft and Google allegedly shift more work to their own AI stacks. Microsoft is tied to Polaris, while Google is linked with Apple’s Gemini plans for Siri. - The article highlights larger context windows, redesigned reward pipelines and stronger long-term memory as the main GPT-6 promises.

- Axios reports that Anthropic took its top models Mythos and Fable offline after the Trump administration imposed strict export controls. Officials say the trigger was jailbreak risk and Anthropic's alleged failure to respect a cyber executive order. - On June 11, Amazon CEO Andy Jassy warned Treasury Secretary Scott Bessent that the models could be jailbroken.

- Anthropic launched Fable 5 on June 9; only days later, access was cut off. Axios says the trigger was an Amazon report claiming parts of Mythos 5 could be jailbroken and posed a national security risk. - Amazon CEO Andy Jassy and other company contacts reached US officials Thursday night and Friday.

In this post, we explore how Rocket Close built a solution using Strands Agents, large language models (LLMs), Amazon Bedrock, Amazon Bedrock Knowledge Bases, and Model Context Protocol (MCP) tools. We cover solution features, the rationale for the technology stack, lessons learned, and the business impact at Rocket Close.

This post outlines the development of a cost-effective and scalable intelligent document processing pipeline on AWS, powered by Amazon Bedrock and its features. BDA is a managed service within Amazon Bedrock that automates the extraction of insights from documents.

AWS Professional Services (AWS ProServe) compressed engagement timelines from months to days, not by adding artificial intelligence (AI) tools to an existing process, but by fundamentally rebuilding how we deliver from the inside out. In this post, we share how AWS ProServe became a frontier team, the practices that enabled it, and what your engineering organization can take from our experience.

Suit filed in US alleges chatbot told Alice Carrier, 24, ‘maybe this is just the end’ as she struggled with suicidal thoughts A Canadian mother sued OpenAI and its CEO, Sam Altman, in US court on Thursday, alleging that ChatGPT encouraged her daughter to kill herself. The lawsuit is the latest in a slew accusing the company of failing to address dangerous conversations between users and the company’s chatbot.

Agent-EvalKit is an open-source toolkit (Apache 2.0) that makes this evaluation infrastructure available by integrating with AI coding assistants, including Claude Code, Kiro CLI, and Kilo Code. This post walks through how Agent-EvalKit works across its six evaluation phases, using a travel research agent built with the Strands Agents SDK and Amazon Bedrock as a running example.

Today, we’re announcing the Neuron Agentic Development capabilities: a collection of AI agents and skills that make this possible for developers building on AWS Trainium and AWS Inferentia. In this post, we explain how the Neuron Agentic Development capabilities accelerate the kernel development workflow.

This post shows engineering teams how to apply that principle to one of the most time-sensitive workflows in engineering: incident triage. You will build a custom incident triage assistant agent using Amazon Quick that orchestrates a response with the New Relic Model Context Protocol (MCP) Server and Asana through native integrations.

Amazon Bedrock AgentCore Runtime gives each agent session its own isolated microVM with a persistent workspace, secure tool access through Gateway, and built-in observability—so you can run Claude Code, Codex, Kiro, and Cursor in parallel without sharing secrets, ports, or filesystems. Close the lid, go to dinner, and pick up where you left off tomorrow.

This blog has previously discussed FHE for ML inference in the post Enable fully homomorphic encryption with Amazon SageMaker endpoints for secure, real-time inferencing, but this post goes a little further. That previous post showed how to implement FHE-based inference 'from scratch' by hand-crafting a linear-regression algorithm using a low-level library called SEAL.

Judges like Colorado magistrate Maritza Braswell increasingly face stacks of filings drafted with AI tools by people without a lawyer. Many can't afford one or have cases too small to interest one. The catch: AI-written briefs often contain fabricated rulings and false citations, adding strain to already overloaded courts.

Amazon has unveiled a new version of its fully autonomous warehouse robot, Proteus, that can be directed with natural language instead of specialized code. Workers can now assign it tasks the way they'd talk to a colleague. The upgrade is part of Amazon's broader automation push, as the e-commerce giant increasingly replaces human workers with robots designed for heavy lifting.

Amazon introduces Bedrock Ops Alert, a three-layer automated monitoring solution that proactively detects operational issues, dynamically tunes alarm thresholds, classifies alarms, auto-creates context-aware support cases, prevents duplicates, and sends targeted notifications to AI SRE teams. The post walks through the architecture and how to deploy it yourself.

In this post, you learn how to use Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO) together to improve the tool-calling accuracy of a small language model (SLM). The example uses Amazon SageMaker AI training jobs, so you can focus on training code instead of managing your own training infrastructure.

Fine-tuning a model for domain-specific tasks means boosting performance in one area without degrading its general capabilities, and that balance is harder to strike than it looks. This guide walks through picking the right customization strategy for your data and task, configuring the training parameters that matter most (learning rate, batch size, checkpointing), and catching the common mistakes that waste training runs and burn compute.

In this post, we walk through how to use Amazon Quick Research to integrate biomedical data sources for rare cancer research. The walkthrough uses pediatric sarcoma as the research domain and draws on publicly available datasets from PubMed and other open biomedical repositories.

Amazon’s approach to artificial intelligence (AI) has faced significant challenges, as highlighted by Brendan Dell. One notable issue is the company’s reliance on strict adoption metrics, such as requiring 80% usage of its internal AI coding system, Kira.

Speaking at Amazon’s AI on the Lot event, the Rogue One film-maker Gareth Edwards said ‘it’ll do anything you ask’ and ‘it’s going to be better than CGI’ Jurassic World Rebirth and Rogue One director Gareth Edwards has enthusiastically endorsed the use of generative AI in film-making, saying “it is a fucking genius at helping you” and “it’s going to be better than CGI”. Edwards was speaking at AI on the Lot, an event in Culver City, California, organise…

Claude Opus 4.8 introduces practical updates for development workflows, including dynamic workflows with parallel sub-agents for tasks like code migration and bug detection. The release also reintroduces manual effort control so developers can allocate compute based on task complexity.

Azercell Telecom partnered with AWS to build a production-ready Azerbaijani LLM on Amazon SageMaker AI for telecom use cases and a customer-facing chatbot. The team had to adapt foundation models to a morphologically rich language with limited training data and no proven blueprint. A six-week collaboration with the AWS Generative AI Innovation Center delivered the production framework.

Walk through building a custom portal with embedded SageMaker AI MLflow Apps UI, using a React frontend and a Flask reverse proxy for AWS Signature Version 4 authentication. The full stack is deployed via AWS CDK, then validated end to end. The piece also covers security considerations and cleanup procedures for a production setup.

This guide combines LangChain's work on evaluating deep agents with Anthropic's eval playbook into a hands-on workflow. You'll learn five evaluation patterns, build offline evals with pytest and LangSmith, and configure online monitoring for production. A text-to-SQL deep agent on Amazon Bedrock serves as the running example from development through deployment.