aws
-
01
Reduce ops and risk: SageMaker AI runs custom training loops without GPUs
Amazon made SageMaker AI’s serverless customization options generally available for 'training sessions', so teams can control post-training logic while AWS handles GPUs, checkpoints and…
-
02
Prevent CloudFormation failures when updating SageMaker AI endpoints
Add a reusable AWS CDK construct that deregisters and re-registers Application Auto Scaling so you can change endpoint settings and autoscaling in one deployment.
-
03
Smartsheet MCP lets AI agents act on live projects; AWS and OpenAI follow with toolkit and safety updates
Agents can now read and write Smartsheet data; Polly adds voices; Bedrock KBs remove RAG ops; Amazon Quick targets HIPAA; OpenAI tightens training containment; battledrones adds four-player…
-
04
Moonshot's Kimi K3.1 Pro lands, and inference costs push ops decisions
Kimi K3.1 Pro appeared with promised open weights and frontier benchmarks; AWS and Bayer ship agent tooling; Spotify tightens family-plan checks; inference storage climbs.
-
05
You can now get OpenAI models through Google Cloud Marketplace
OpenAI models join Google Cloud’s Model Garden; AWS shows a pattern for running routine agent work on local small models; Xbox starts paid cloud playtime; security and new models arrive.
-
06
Get cross‑system supply‑chain answers in seconds with Amazon Quick and AgentCore
AWS shows Amazon Bedrock AgentCore as a tool layer and Amazon Quick as a conversational front end that pulls from live systems and documents, replacing manual stitching of data.
-
07
Prevent AI-agent impersonation with X.509 PKI and Keyfactor EJBCA on AWS
AWS shows how to give each AI agent a cryptographic identity using X.509 PKI and Keyfactor EJBCA alongside AWS services, closing a gap in agent authentication.
-
08
Laplace 3B cuts per‑sequence memory and raises concurrent throughput on Trainium2
i14’s Laplace 3B runs on AWS Trainium2 with a fixed 29.5 MB recurrent state and scaled to 1,280 concurrent sequences; pricing was not disclosed.
-
09
AWS Agent Registry lets agents discover tools at runtime, cutting coordination
This week: AWS adds runtime discovery for agent tools; Smartsheet and Eon push MCP into production; a franchisee reports big time savings with ChatGPT Work; watch AI unit economics and a…
-
10
A few sign‑up steps now deliver a preconfigured SageMaker Studio
New AWS sign‑up flow lets new SageMaker AI customers use existing sign‑in methods and start in a ready Studio; new accounts also get $100 in Free Tier credits.
-
11
Amazon Bedrock adds cross‑Region GPT‑6 inference for five Asia Pacific markets
Teams in Thailand, Malaysia, Singapore, Indonesia and Taiwan can call OpenAI GPT‑6 models via Bedrock from other AWS Regions; setup, permissions and monitoring are required.
-
12
Gemini Argon access and price window force near-term budget choices
Google’s Gemini 4 Argon is available to paying users at a lower rate until 1 Dec; AWS added agent payments, licence automation, presentation templates and a security query agent.
-
13
Run a large LLM across two GPU nodes on EKS to avoid single‑GPU limits
AWS shows how to split one model across two GPU nodes with Ray LLM Deep Learning Containers, using separate prefill and decode nodes behind an OpenAI‑compatible endpoint.
-
14
Cloudflare lets teams filter and reshape logs before export
Cloudflare made Transformers in Logpush generally available and opened Logpush to all self-serve plans; Meta, AWS and others also pushed model and tooling updates worth testing.
-
15
SpaceX buys low‑band spectrum, accelerating Starlink Mobile plans
New spectrum and satellite approvals raise the prospect of a terrestrial Starlink mobile network; AWS, OpenAI and Oracle released adoption tools that shorten workflows and test agents.
-
16
GPT-6.1 Sol on Bedrock reshapes agent costs and performance
Bedrock gets GPT-6.1 Sol with OpenAI’s pricing claim; Anthropic, Cloudflare, AWS and Nvidia updates change model speed, search billing, migration time and device costs.
-
17
Cloudflare makes enterprise features available to every team
Cloudflare opens enterprise capabilities to all teams; Cloudflare and Microsoft publish models and frameworks for deterministic decisioning and smaller agentic models; AWS and Google offer…
-
18
Cloudflare acquires Deno; runtime supported for one year, then development ends
Six updates today: Cloudflare buys Deno with a one-year maintenance window; AWS recaps Bedrock, AgentCore and Strands; Postman runs Agent Mode on Bedrock for 40M devs; Asana cuts browser…
-
19
Add automated, pre‑validated remediation after AWS DevOps Agent investigations
Use Lambda Durable Functions, Amazon EventBridge and Amazon Bedrock to turn AWS DevOps Agent investigation output into an automated remediation workflow that reduces on‑call toil.
-
20
Using hours-saved ROI will undercount agentic automation
AWS says the RPA-era formula (transactions × minutes × rate − build cost) misses maintenance, exceptions and outcome value; change your business case.
-
21
Using Claude Haiku 5.5 on AWS lowers model costs for high-volume workloads
Claude Haiku 5.5 is available on Amazon Bedrock and Claude Platform on AWS, offering faster, more efficient inference and about 75 percent lower cost than Haiku 4.5.
-
22
Anthropic Claude models in AWS GovCloud let teams run AI coding for regulated workloads
Anthropic Claude Opus 5.5 and Sonnet 5.5 are available in AWS GovCloud on Amazon Bedrock with Claude Code; Sonnet 5 holds FedRAMP Class D and DoD IL4/IL5 authorisations.
-
23
You can add a voice travel concierge using three managed AWS services
AWS published a pattern for a spoken travel assistant using Bedrock AgentCore, Nova Sonic for real‑time voice, and Bedrock Knowledge Bases to ground answers.
-
24
Keep conversations across sessions by adding AgentCore memory and OpenClaw
AWS shows how to give assistants persistent, taggable memory using OpenClaw on the AgentCore runtime, so teams can stop re-entering context and prototype a week-long proof of concept.
-
25
Coding agents can benchmark and generate SageMaker SDK v3 deployment code
Install the aws‑ai‑ml skill in the Agent Toolkit to let Kiro, Claude Code or Codex benchmark SageMaker endpoints, recommend deployment settings and emit executable SDK v3 code.
-
26
Share one Claude subscription to cut environment sprawl and centralise access
AWS's October 1, 2026 guide shows how to run production, developer laptops and external CI from one Claude subscription while isolating traffic by workspace.
-
27
GitHub Taskflow Agent found Android flaws; plan for model costs
GitHub’s open-source Taskflow Agent found Android vulnerabilities; teams must budget for a Copilot licence, premium-model calls and triage time.
-
28
Route Claude Desktop web searches through Bedrock AgentCore to keep answers current
A concise, hands-on sequence to connect Claude Desktop to Amazon Bedrock AgentCore Web Search, secure it with JWT and keep queries inside your AWS boundary.
-
29
Microsoft offers AI converter to cut Salesforce-to-Dynamics migration work
Today: Microsoft and AWS add tools that shorten migrations and model development; OpenAI shares formalised math proofs and shows ChatGPT used in quant workflows; Instinct adds group agents.