Superpower Daily
Google Signs Canada’s Data-Centre Principles, Pledges to Cover...GPT-6 Astra Hits 99.9% on ARC-AGI-3, but Scores 62.7% in a Shar...Crusoe Signs Reported $13 Billion, Five-Year AI Cloud Deal With...Microsoft Releases MAI-Transcribe-2 at 10 Cents an HourxAI Opens Grok Bot to Enterprises With Controls for Autonomous...Meta Lawsuit Adds Bystanders Who Say AI Glasses Recorded Them W...Google Signs Canada’s Data-Centre Principles, Pledges to Cover...GPT-6 Astra Hits 99.9% on ARC-AGI-3, but Scores 62.7% in a Shar...Crusoe Signs Reported $13 Billion, Five-Year AI Cloud Deal With...Microsoft Releases MAI-Transcribe-2 at 10 Cents an HourxAI Opens Grok Bot to Enterprises With Controls for Autonomous...Meta Lawsuit Adds Bystanders Who Say AI Glasses Recorded Them W...

Full archive

Launch: Latest AI News

Search reporting on AI markets, models, products, companies, policy, and culture.

Category

More filters
LaunchClear all
Showing 1-18 of 33Results matching "Scale AI"

Today

3 stories
LaunchModels

HUMAIN Releases Arabic AI Preview Built on MiniMax’s M3 Lineage

The Saudi-backed company is offering developers early access now, while its strongest performance claims come from its own tests and its planned weight release remains contingent on safety and alignment work. HUMAIN has put humain-m3, an Arabic-focused AI model built on the MiniMax -M3 lineage, into research preview through HUMAIN Node. Developers can use a no-code playground or an OpenAI-compatible API.

3 min readRead story
LaunchModels

OpenAI Releases GPT-6 Astra With Its First Advanced Cyber Safeguards

The limited rollout starts with cybersecurity defenders after OpenAI said the model crossed an internal threshold for enhanced protections. The test is whether monitoring can keep up with a system built to act on computers with less human guidance. OpenAI has released GPT-6 Astra to a limited set of customers, starting with participants in its Daybreak program for cybersecurity defenders.

3 min readRead story
LaunchModels

IFM Releases Six K2 Horizon Models With Weights, Code and Training Data

The new lineup gives developers models from 0.9B to 375B parameters under Apache 2.0, while putting IFM’s performance claims and unusually broad training record in public view. Developers seeking to move from a small local model to a larger enterprise system now have a single newly released family to test. A dense 32B model and sparse 36B-A4B model are intended for local hosting and on-premise servers.

3 min readRead story

Yesterday

2 stories
LaunchInfrastructure

Databricks Pushes Lakebase Postgres as the Backend for Apps and Agents

The developer push pairs managed PostgreSQL with Databricks Apps, handling application identities and connection details inside the workspace while leaving teams to choose their database environment and data-access model. Databricks is positioning Lakebase Managed Postgres as a managed PostgreSQL backend for web applications and AI agents. Connection details are injected into the application as environment variables.

3 min readRead story

Earlier this week

7 stories
LaunchProducts

Amazon Quick Pitches Connected Workflows as Vertiv Plans 25% User Growth in 2026

AWS is framing Quick as a conversational layer across business systems, but its customer examples describe deployments and intended uses rather than comparable measures of productivity or return. AWS is promoting Amazon Quick as a work assistant that uses agentic teammates for research, business insights and automation, while also turning conversations into finished work products. Respect your permissions?

2 min readRead story
LaunchModels

Skild AI’s S1 Uses One Human Video to Prompt Robots Through 10-Minute Tasks

The model’s central bet is that a demonstration can become immediate instruction rather than another task-specific training run. The harder question is whether that capability can turn into dependable deployments. The company says the approach can cover complex jobs lasting as long as 10 minutes, potentially reducing the task-by-task retraining that slows robotics development.

3 min readRead story
LaunchInfrastructure

Build American AI Takes Multimillion-Dollar Data-Center Ads to Three Election States

The campaign brings organized political advertising to a facilities fight that has become a flashpoint in contests for Congress and state and local government. Build American AI is launching a multimillion-dollar advertising campaign promoting data centers in Kansas, Ohio and Wisconsin. The nonprofit says the effort is meant to shift public opinion as disputes over the facilities enter statewide election politics.

2 min readRead story
LaunchProducts

AWS Agent Registry Reaches GA With Approval Gates for Enterprise Agent Discovery

The service makes approval status a condition of discovery, separating a complete administrative record from the capabilities available for enterprise reuse. AWS has made AWS Agent Registry generally available, giving organizations a searchable catalog for AI agents, tools, skills and custom resources. Administrators can configure discovery policies, compliance and security metadata, and custom metadata schemas.

3 min readRead story
LaunchProducts

Broadcom Releases AgentMinder to Check AI Agent Tool Calls at Runtime

The new control layer is designed to evaluate each requested agent action before it reaches company systems, shifting governance from static permissions toward live policy enforcement. Broadcom has made AgentMinder generally available, offering enterprises a control layer that checks an AI agent’s requested actions before they reach company resources. The product was introduced Aug.

2 min readRead story

Last week

6 stories
LaunchStartups

Jeff Dean Leaves Google to Build Discovery Loop, a Science AI Startup Backed by Radical Ventures

The new public-benefit company has an unusually experienced founding team and support from both Radical Ventures and Alphabet. Its central technical question is still unanswered: what system will turn the ambition to automate research into a usable product? Jeff Dean has left Google to co-found Discovery Loop, a new AI company with Sanjay Ghemawat, Quoc Le and Oriol Vinyals.

4 min readRead story
LaunchInfrastructure

OpenRelay Routes AI Inference Across GPUs, TPUs and Trainium Through One API

The early-access network aims to let developers buy performance rather than a particular accelerator, but its cost-savings claims and ability to deliver consistent service across mixed infrastructure remain unproven. OpenRelay has launched an inference API that sends AI workloads across GPU, TPU and Trainium capacity through a single endpoint. OpenRelay is in early access, though its /v1 API is listed as stable.

3 min readRead story
LaunchInfrastructure

Nvidia Rolls Out Vera Rubin as AI Competition Shifts to Data Traffic

The architecture packages compute with storage and networking hardware, betting that efficiently moving data through large AI systems is becoming its own competitive layer. Nvidia is rolling out Vera Rubin, an architecture that combines Rubin GPUs with Vera CPUs, Groq 3 LPX inference accelerators, and storage and networking racks. Related on X NVIDIA Vera is heading to @awscloud.

2 min readRead story
LaunchInfrastructure

Leiolai Launches Device-Run AI With 11M-Token Context and Output From $0.02 Per Million

The app and API depend on users supplying computation, making coordination, latency and trust central tests of its alternative to centralized AI infrastructure. Leiolai has launched leiolai-1 with a consumer app and developer API built around an unusual infrastructure promise: model inference runs across users’ existing devices rather than centralized data centers. That setup changes what expansion means.

3 min readRead story
15 more next