Last 24 hours · updated
11 things shipped in AI in the last 24 hours
Builders got a multi-provider routing endpoint with failover and unified billing from Perplexity, voice cloning in Amazon Polly, and a Bedrock AgentCore payments capability that lets agents buy AWS Marketplace products on the developer's AWS bill. Gemini API Tool Retrieval hit preview for calling hundreds of tools with fewer tokens, and Kimi Code CLI and Claude Code both shipped updates.
What it means for builders
Kimi Code CLI v2.7.0 can now connect directly to Amazon Bedrock models with standard AWS credentials, removing a separate key setup for AWS-based coding agents 1. Perplexity's Orchestrator API serves OpenAI, Anthropic, Google, xAI and open-weight models through one endpoint, with same-model provider failover and a perplexity/auto router that picks a model per request 2. On the agent side, Bedrock AgentCore payments now lets agents discover and buy x402-enabled Marketplace products and pay in USD on the AWS bill, with no stablecoin wallets needed 3, and the AgentCore harness picked up Strands defaults, prompt caching, context management and new built-in tools 4. In preview, Gemini API Tool Retrieval lets agents access hundreds of tools while using fewer tokens 5. Cost pressure is real: token prices for AI models fell 21 percent in six weeks while usage hit a record high 6, which matters as enterprises scale token spend on Bedrock even as per-token prices fall 7.
Shipped
Google AI chats
AI chats in a collapsible sidebar on Google homepage
Tools & APIs · The Verge +1 · 2h ago
Nemotron-Labs-Sigma
Continuous diffusion language model with 14 open checkpoints
Models · MarkTechPost · 35m ago
Kimi Code CLI 2.7.0
Connects to Amazon Bedrock models with AWS credentials.
AI coding · AWS · 14h ago
Amazon Bedrock AgentCore
Agents can buy AWS Marketplace products and pay on their AWS bill.
Agents & automation · AWS · 2 releases · 16h ago
Claude Code 2.1.329
Adds trust dialog warning and hook load failure alerts
AI coding · Claude Code releases · 17h ago
SageMaker AI
Monitors inference endpoints for latency, idle GPUs and optimization opportunities.
Tools & APIs · AWS · 17h ago
Amazon Polly voice cloning
Creates custom voices from speaker recordings.
Tools & APIs · AWS · 19h ago
Orchestrator API
Route requests to multiple models and providers with failover and unified billing.
Tools & APIs · MarkTechPost · 21h ago
Dialog-RSN-1
Deploy a reasoning model for complex dialog to a SageMaker endpoint.
Models · AWS · 21h ago
Gemini co-marketing
Gemini-powered co-marketing in partners' workflows.
Business · Google Cloud · 22h ago
Amazon SageMaker HyperPod
Diagnose GPU training issues faster with enhanced observability.
Chips & infra · AWS · 22h ago
Models
PolyAI Dialog-RSN-1, a reasoning model built for real-time complex dialog, is available on AWS Marketplace and Amazon SageMaker JumpStart 8.
- NVIDIA AI Releases Nemotron-Labs-Sigma, a Continuous Diffusion Language Model with 14 checkpoints at 3B/8BShippedMarkTechPost · 1 outlet
- Meta FAIR Self-Improving Memory Hits 56.0 on Continual Learning BenchMarkTechPost · 1 outlet
- PolyAI Dialog-RSN-1 is now on Amazon SageMaker JumpStartShippedAWS · 1 outlet
AI coding
Kimi Code CLI v2.7.0 adds a native Amazon Bedrock provider so the open source coding agent connects with your standard AWS credential chain 1, and Claude Code v2.1.329 adds a trust dialog for hook declarations and warnings when managed hooks fail to load 9.
- Use Kimi Code CLI with models on Amazon BedrockShippedAWS · 1 outlet
- Claude Code v2.1.329ShippedClaude Code releases · 1 outlet
Tools & APIs
Perplexity's Orchestrator API routes to multiple model providers through one endpoint with failover and unified billing 2, Amazon Polly voice cloning is generally available for custom brand voices 10, SageMaker AI now monitors inference endpoints by default for latency drift, idle GPUs and optimization opportunities 11, and Google put collapsible AI chats on its homepage 12.
- Google is putting AI chats right on the homepageShippedThe Verge · 2 outlets
- How Salesforce is advancing global-scale model serving with cross-Region inference routing in Amazon SageMaker AIAWS · 1 outlet
- Introducing harness adapters: adapt your favorite model to your harness to improve performance and costAnnouncedHugging Face · 1 outlet
Agents & automation
Bedrock AgentCore payments now lets agents buy x402-enabled AWS Marketplace products and pay on the AWS bill 3, the AgentCore harness added Strands harness defaults, built-in tools and lifecycle hooks 4, Hermes Agent announced a Collective Wisdom feature that shares AI skills across teams in Microsoft Teams 13, and Google expanded access to Gemini Tasks for calls, logins and purchases 14.
- Hermes Agent's Collective Wisdom feature turns one employee's AI trick into the whole team's skillAnnouncedThe Decoder · 1 outlet
- Google expands access to Gemini Tasks, which can help with calls, purchasesTechCrunch · 2 outlets
- Enable AI agents to pay for AWS Marketplace x402 products with Amazon Bedrock AgentCore paymentsShippedAWS · 1 outlet
Business
Amazon and Google will together own nearly 30% of Anthropic after its IPO 15, Apple is hiring engineers to build its own AI agent 16, Meta plans to price the Muse Charm gadget at $300 17, and token prices for AI models fell 21 percent in six weeks while usage hit a record high 6.
- Apple is hiring engineers to build its own AI agentThe Verge · 3 outlets
- Nvidia bets at least $500 million on Inception's diffusion models to strengthen its open-source Nemotron lineupThe Decoder · 1 outlet
- Amazon and Google will together own nearly 30% of Anthropic after its IPOTechCrunch · 4 outlets
Policy & safety
The California AG sued OpenAI over rogue agent incidents in one of the first attempts to hold an AI lab legally accountable 18, and China reportedly recruited an AI executive to spy on Nvidia CEO Jensen Huang 19.
- China reportedly recruited AI exec to spy on Jensen HuangThe Verge · 6 outlets
- Signal is blocking Gemini and Siri AI from reading your screenThe Verge · 1 outlet
- California state AG sues OpenAI over rogue agent incidentsTechCrunch · 3 outlets
Chips & infra
NVIDIA and Microsoft announced expanded accelerated computing from Azure to Windows PCs 20.
- NVIDIA and Microsoft Expand Accelerated Computing From Azure to Windows PCsAnnouncedNVIDIA · 1 outlet
- Introducing enhanced infrastructure observability in Amazon SageMaker HyperPodShippedAWS · 1 outlet
- Using NVIDIA Run:ai GPU Fractioning on Amazon EKSAWS · 1 outlet