All news

MMarkTechPost03:00Product release

AI score52

Liquid AI releases real-time inference engine RIPL, with end-to-end latency below 50 ms

Liquid AI has released the real-time inference engine RIPL (Real-time Inference Pipeline), used to serve the LFM2.5 series models (230M to 2.6B parameters) in real-time request paths, generating end-to-end responses in under 50 ms. According to Liquid AI's own benchmark data, on a single B200 its p95 latency is 5.6 to 28.4 times lower than that of Qwen3.5 running on vLLM, with LFM2.5-1.2B at 155.8 ms and Qwen3.5-0.8B at 4431 ms. The engine is aimed only at enterprise deployments and supports NVIDIA H100, H200, B200 and RTX PRO 6000; Shopify already runs LFM in search scenarios with inference latency below 20 ms.

#Product updates#Reasoning#English text is being prepared.

RRohan Paul02:47Product release

AI score55

HUD launches DataVendor data marketplace, where AI labs can buy private data at listed prices

HUD releases the public beta of DataVendor, an online marketplace that lets companies sell private data to AI labs; labs can browse data, view automated quality results, try samples, and purchase at listed prices. The marketplace does not include or price public code repositories, because frontier models have already been trained on that code; buyers can see quality results and a desensitized file tree, and can randomly run 1 task from the task set within their own HUD workspace. Previously, labs procuring private data for RL environments typically had to go through intermediaries, NDAs, and weeks of testing.

Quoted postDataVendor · @datavendor_ai

We're announcing the public beta of https://t.co/YgMb6QF79k We spent the last two years building the best solution for people buying and selling data, whether you're a researcher in a lab or a company trying to sell data. Data is the fundamental bottleneck for AI progress, and DataVendor will be the infra that solves it.

Same story · 2 reports

#Product updates#English text is being prepared.#Agents

ZZhidx02:38Product release

AI score79

Anthropic Launches Personal AI Agent Claude Duet

Anthropic has launched Claude Duet, a personal AI agent app built on Claude Opus 5.5 that can proactively identify pending tasks and continuously execute them even when the user has not given instructions. It can call on more than 6,000 connectors and plugins in the Claude Marketplace, covering scenarios such as financial analysis, recruitment management, customer communication, and payroll processing, and it has its own virtual machine and browser, so tasks can continue even after the user turns off their device. The app is currently available as standalone iOS, Android, and web apps, with the first batch of research preview waitlist applications open to Pro and Max subscribers in the United States; when it involves sending messages, making payments, signing, or deleting content, user confirmation is required, and securities trading and transfers of funds between accounts will not be executed even if approved.

Same story · 44 reports

#Product updates#Agents#English text is being prepared.

UUnite.AI02:36Product release

AI score61

NVIDIA Releases Minerva AIOps Platform for AI Factory Observability

On October 22, NVIDIA released Minerva, an AIOps platform for AI factories that integrates telemetry data from infrastructure and AI workloads, using purpose-built AI models to turn data into actionable insights. The platform builds a dynamic knowledge graph that maps GPUs, SuperNICs, switches and cables along with the jobs running on them, and can merge a large number of related events into a single event with a probable root cause. Minerva includes a multi-agent AI assistant named Sherlock and provides an MCP server, supporting troubleshooting through natural language; IREN, Bitdeer AI and others are already using it, and Dell, Lenovo and Supermicro will incorporate it into their AI factory operations product portfolios.

Same story · 2 reports

#Product updates#English text is being prepared.#Agents

UUnite.AI02:36Product release

AI score61

Cohere launches confidential computing public beta on Model Vault

On October 22, Cohere announced that the confidential computing feature of Cohere Model Vault has entered public beta, running on NVIDIA H100 GPUs on Microsoft Azure, targeting enterprises that handle sensitive data such as clinical records, financial information, privileged documents, and proprietary source code. Through encrypted request paths, remote attestation, and client-side verification, the feature prevents operators at Azure, NVIDIA, and Cohere itself from reading the memory of protected workloads. Cohere also stated that physical and side-channel attacks, availability, prompt injection, and other application-layer risks, as well as metadata such as message size and timing, are outside the scope of protection, and it plans to open-source the confidential inference stack.

#Product updates#English text is being prepared.#English text is being prepared.

99to5Mac02:33Product release

AI score58

1Password Launches 1Password for Agents, Letting AI Agents Use Login Credentials Securely

1Password has released the 1Password for Agents platform, enabling AI agents to complete operations requiring login credentials without viewing users' passwords. The platform initially supports Meta's Muse and SpaceXSI's Grok Bot, and says it will later integrate more partners including OpenAI and Anthropic. 1Password says the platform can connect to a user's single or multiple personal agents, with credentials and keys managed by 1Password.

Same story · 13 reports

#Product updates#Agents#English text is being prepared.

99to5Mac02:33Product release

AI score59

Apple adds Intelligent Assistant and other AI features to Pages, Keynote, and Numbers

Apple has rolled out an AI update for its iWork suite—Pages, Keynote, and Numbers—adding an Intelligent Assistant, currently available as a beta in US English and requiring iOS 27 or later. The assistant can understand the current document and supports drafting content in natural language, adjusting tone, and generating slides, speaker notes, images, and shapes; in Numbers it can generate tables, offer chart suggestions, and help write formulas, and it can also retrieve the latest information and write it into documents. Keynote also adds an open file format, Interchange Format (.keyx), that can be imported and exported. After updating all iWork apps, the author had not yet seen the assistant and speculates that a server-side component may be awaiting activation.

#Product updates#Agents#OpenAI

AAI Business02:30Product release

AI score65

Anthropic Launches Personal Agent Claude Duet, but Enterprises Remain Wary of Adoption

This week, Anthropic released Claude Duet, a personal agent app, as a research preview. Built on Claude Opus 5.5, it can continue completing tasks after the user steps away and seeks confirmation before acting on the user's behalf. The app targets professionals and small business owners, offering standalone iOS and Android apps as well as a Claude.ai web version, and is included in Pro and Max subscriptions. A Forrester analyst said Duet's distinguishing feature is the restraint it introduces: built-in classifiers and safety prompts block requests to share sensitive information such as credit card numbers and government IDs, and restrict securities trading and transfers between accounts. A GAI Insights analyst said B2B customers may not deploy Duet for months or even years, and many enterprises have yet to enable even Claude Cowork, which was released in January.

Same story · 44 reports

#Product updates#Agents#Anthropic

AAWS Machine Learning02:16First-partyProduct release

AI score56

Amazon SageMaker AI launches Training Sessions API, enabling custom training loops without managing GPUs

Amazon SageMaker AI has added a Training Sessions API to serverless model customization, turning operations such as forward_backward(), optim_step(), and sample() into API calls. The model stays loaded on the GPU, and users only need to write the training loop in a local Python process, while GPU allocation, distributed execution, fault recovery, and checkpoint storage are handled by the service. The API is billed by tokens processed, and steps such as local reward computation do not count toward GPU time. The accompanying cookbook provides SFT and RL recipes; in the example, the GRPO loop for GPT-OSS 20B trains for 10 steps on GSM8K, raising held-out set accuracy from 56% to 75%; the Code RL recipe raises test accuracy from 40% to 50.5% after 50 training steps.

#Product updates#English text is being prepared.#English text is being prepared.

AAWS Machine Learning02:16First-partyProduct release

AI score61

AWS launches training sessions GA for SageMaker AI serverless model customization

AWS announced that training sessions for SageMaker AI serverless model customization are generally available; users can write and orchestrate their own training loops, including custom loss functions and reward signals, while SageMaker AI handles GPU allocation, distributed execution, fault recovery, and checkpoint storage. The feature supports SFT, DPO, RLVR, RLAIF, and multi-turn RL, with a choice of LoRA or full fine-tuning, and currently supports gpt-oss-20b, gemma-4-31B-it, Qwen3.6-27B, and Qwen3.8-27B. Billing is per token processed by SageMaker AI, billing stops when the reward function scores in the user's environment, and no reserved capacity is required.

#Product updates#English text is being prepared.#English text is being prepared.

TThe Decoder02:04Product release

AI score69

Google launches Agent Compute Engine, allocating a hardware-isolated microVM to each Gemini agent

Google announced Agent Compute Engine, which provides each AI agent with an independently isolated virtual machine. The service already powers Gemini and will open in preview over the coming months. Each agent runs in a microVM protected by Google Titanium chips, automatically suspends when idle to free up machines, and shared memory between virtual machines lowers per-machine costs. Google says its open-source Kubernetes runtime Agent Substrate can host ten times as many agents on a single server, and in a test with 200,000 agents, suspend and resume took under 500 milliseconds. It also adds token tracking, model routing, and support for third-party identity providers such as Ping Identity and SailPoint; these controls will extend to Cloud Run, GKE, and managed endpoints.

Same story · 6 reports

#Product updates#Agents#English text is being prepared.

DDavid Singleton01:14Product release

AI score59

Muse integrates 1Password; agent logins require user approval each time

Muse announced that it will integrate 1Password directly into its product; users must approve each login themselves, and passwords remain stored in their own vaults. The feature will roll out across all platforms over the coming days. 1Password released the 1Password for Agents platform the same day: when an agent encounters a login form, it requests credentials from the password vault, and after the user approves, 1Password autofills directly, so the agent cannot see or hold the credentials. The feature is free and optional for individual and family accounts; when the connector is not enabled, agents cannot access the password vault.

Quoted post1Password · @1Password

Today we announce 1Password for Agents platform, building with companies @Muse, SpaceX Super Intelligence Grok @Bot, @AnthropicAI’s @Claudeai, @OpenAI, @TownAI, @Browserbase, @Browser_use, @heydana_ai, Instinct, @usekernel, @Perplexity_ai, @tryramp, @Tomo, and @wajo_ai, with more integrations on the way. 🎉 We've heard from so many of our customers who want to use these new personal AI agents to help them get more done, but want more control over what the agent can access with the same trust and ease of use they get when using 1Password directly. 1Password for Agents gives your agents access without giving up your passwords. When an agent reaches a sign-in form, it requests a credential from your vault. If you approve the request, 1Password directly auto-fills the credential in the destination site so the agent can’t see or take possession of the credentials. The agent gets the task done, but it doesn’t get your passwords.. You stay in control. 1Password for Agents free and optional for personal and family accounts. Unless you opt-in to an agent connector, agents cannot touch your vault. ➡️Learn more: https://t.co/KxDcCum2iE

Same story · 13 reports

#Product updates#Agents#English text is being prepared.

TTesting Catalog01:14Product release

AI score62

Google Cloud launches Agent Compute Engine, providing dedicated VMs for AI agents

At AGNTCon + MCPCon, Google Cloud announced the launch of Agent Compute Engine (ACE), a VM type built specifically for AI agents and aimed at large-scale operation scenarios. The system comprises four components—Agent Gateway, Agent Registry, Agent Identity, and Agent Observability—and also expands Gemini's capabilities for enterprise agent governance.

Quoted postGoogle Cloud · @googlecloud

Announced at AGNTCon + MCPCon: the new Agent Compute Engine plus expanded Gemini for business agent governance deliver the control you need to run, govern, and scale your agents. Learn more about Agent Compute Engine and expanded agent governance → https://t.co/z1fW5Mxs0I https://t.co/TTUmhe98Yt

Same story · 6 reports

#Product updates#Agents#Google

DData Center Knowledge01:13Product release

AI score55

Nvidia unveils DSX POD modular data center solution at GTC Berlin

Nvidia unveiled its prefabricated modular reference design DSX POD at GTC Berlin, aimed at partners including Dell, HPE, Schneider Electric and Vertiv, based on Vera Rubin infrastructure, and claims it can cut deployment time from about six months to as fast as 14 days. Jensen Huang said a gigawatt-scale data center costs $50 billion, and a three-month delay would mean losing $10 billion in revenue. Before the event, Greenpeace protesters interrupted the keynote, raising questions about data center environmental issues and reports of Nvidia technology appearing in Russian drones.

Same story · 20 reports

#Product updates#English text is being prepared.#English text is being prepared.

CCohere Product and Research Blog01:07First-partyProduct release

AI score58

Cohere Model Vault Launches Confidential Computing Public Beta

Cohere announced that the confidential computing feature of Model Vault has entered public beta, providing hardware-level isolation for hosted model inference based on the NVIDIA H100 GPU confidential computing capability on Microsoft Azure. This feature extends the scope of encryption from data at rest and in transit to prompts, documents, responses, and model execution during inference; after end-to-end encryption on the application workload side, decryption can only occur in a confidential environment that has passed remote attestation. Customers can cryptographically verify the hardware, firmware, and software configuration before sending sensitive data; if it does not comply with the security policy, the client will refuse to send the request. The existing Model Vault API and request format remain unchanged.

#Product updates#English text is being prepared.#English text is being prepared.

CCohere Product and Research Blog01:07First-partyProduct release

AI score58

Cohere Model Vault Launches Verifiable Confidential AI Inference

Cohere has launched confidential computing capabilities on Model Vault, currently in public beta, running on NVIDIA GPUs on Microsoft Azure. The feature combines confidential CPU and GPU execution, OHTTP end-to-end encryption of request content, and remote attestation, allowing customers to verify the target environment before releasing sensitive data. Cohere plans to open-source the confidential inference stack and advance support for NVIDIA Blackwell and B200 GPUs.

#Product updates#Safety / alignment#English text is being prepared.

NNVIDIA Developer · Generative AI01:00First-partyProduct release

AI score55

NVIDIA DGX Spark supports eight-node clusters, unified memory expands to 1TB

At GTC Berlin, NVIDIA announced that DGX Spark officially supports eight-node clusters, with unified memory expanding to 1TB, enabling local execution of frontier models such as Z.ai's full GLM 5.3 (753B). Official benchmarks show that Qwen3.8-27B (NVFP4) decoding throughput rises from about 12 tokens/s on a single node to 52 tokens/s on eight nodes, and with the Spark-optimized RoCEnante collective communication backend, achieves a 4.43x decoding speedup on eight nodes. From-scratch pretraining of the nanochat model on eight nodes achieved a 7.79x speedup and 97.4% scaling efficiency. NVIDIA Sync's eight-node cluster support will launch before the end of the year.

Same story · 3 reports

#Product updates#English text is being prepared.#Reasoning

MMicrosoft Research00:15First-partyProduct release

AI score54

Microsoft Research releases Flint-interactive, enabling agents to generate interactive data visualizations

Microsoft Research has released Flint-interactive, an interactive data visualization language built on top of Flint that enables agents to generate charts supporting clicking, filtering, zooming, highlighting, and cross-chart linking. It uses the Flint compiler to parse raw triggers such as clicks and drags into semantic events, so users only need to describe update behaviors with high-level operators without writing low-level visualization code. The language is already usable in conversations through Flint-MCP and MCP Apps, and it also offers a programming API for developers to build dashboards and data stories.

#Product updates#Agents#English text is being prepared.

LLMSYS Blog00:09First-partyProduct release

AI score52

LMSYS Releases SGLang AWS Trainium Plugin Beta

LMSYS, in collaboration with AWS, has released the SGLang Trainium plugin Beta, built on the newly released TorchNeuron, supporting LLM inference serving and RL rollout generation on Trainium2 and Trainium3. Initial coverage includes Qwen3 (dense), Qwen3-MoE-30B-A3B, DeepSeek-V2-Lite, GLM-4, and GLM-5.2/5.3, and supports data parallelism, tensor parallelism, expert parallelism, Radix Cache, BF16/FP8, and asynchronous scheduling. In the first end-to-end heterogeneous RL run, SGLang generated rollouts for Qwen3-MoE-30B-A3B on Trainium3, while Miles trained a GRPO policy with Megatron on a single H200 node. The Beta is currently limited to a small number of teams.

#Product updates#English text is being prepared.#Open-source ecosystem

GGoogle Developers Blog23:59First-partyProduct release

AI score52

Google API Gateway supports streaming and enters public preview

Google announced that streaming support for API Gateway has entered public preview. SSE, chunked transfer, WebSocket, and gRPC streaming endpoints can share the same gateway and policies with existing APIs, with no changes required to API specifications. Streams can remain open for up to one hour; during the preview period, this must be enabled when creating a gateway, and existing gateways cannot be converted in place. Combined with model routing, clients can use a single OpenAI-compatible endpoint to route to Gemini, Claude, and OpenAI gpt-oss models based on the model field in the request and return streamed responses. The gateway's default domain name has changed to a fixed format composed of the gateway ID, project number, and region, which can be written into the configuration before creation.

#Product updates#English text is being prepared.#English text is being prepared.