Superpower Daily
xAI Publishes Procurement Bot Prompt That Requires Approval for...Artificial Analysis Clarifies Astra’s 67-Point Coding Score Is...Nvidia Agrees to Buy Hugging Face for $12.93B, Promising Hardwa...Google Opens Lyria 3.5 to Developers and Adds Song Generation t...Five State Lawmakers Urge Frontier AI Labs to Negotiate a Slowe...Arizona Study Finds Repeated Falsehoods Can Sway AI Models in L...xAI Publishes Procurement Bot Prompt That Requires Approval for...Artificial Analysis Clarifies Astra’s 67-Point Coding Score Is...Nvidia Agrees to Buy Hugging Face for $12.93B, Promising Hardwa...Google Opens Lyria 3.5 to Developers and Adds Song Generation t...Five State Lawmakers Urge Frontier AI Labs to Negotiate a Slowe...Arizona Study Finds Repeated Falsehoods Can Sway AI Models in L...

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 44Results matching "Agent Operating System"

Today

2 stories
LaunchTools

xAI Publishes Procurement Bot Prompt That Requires Approval for Vendor Messages—and Bars Purchases

The Haggle Bot template turns vendor savings into a persistent agent workflow: inspect live spend and usage, price alternatives, and prepare negotiations while keeping external outreach and financial commitments behind separate limits. The published configuration asks an organization to specify where spend, contracts, usage data, vendor dossiers, and colleagues can be reached.

3 min readRead story

Yesterday

3 stories
LaunchModels

GPT-6 Astra Hits 99.9% on ARC-AGI-3, but Scores 62.7% in a Shared Test

ARC Prize will label the two evaluation conditions separately after OpenAI’s context-management setup produced a near-perfect result that its shared interface did not reproduce. GPT-6 Astra reached 99.9% on ARC-AGI-3 ’s Semi-Private set at high reasoning effort through a Provider Adapter harness. The benchmark places agents in unfamiliar abstract, turn-based environments without explicit instructions.

3 min readRead story
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

Earlier this week

13 stories
LaunchProducts

CrowdStrike Wants AI Agents Rechecked Before Every Action

The company’s approach makes an agent’s authority temporary: access is scoped to a job and can be withdrawn when that job ends or risk conditions change. An AI agent’s access can now be treated as a series of decisions rather than a privilege that lasts for a full login session. CrowdStrike highlighted that model, called Continuous Identity for AI Agents, during its Fal.Con 2026 keynote program in Las Vegas.

2 min readRead story
LaunchProducts

Anthropic Gives Retailers Claude Agent Blueprints, With Checkout Still Off-Limits

The new guidance packages shopper recommendations and merchant decision support for a holiday-season retail push, while keeping agents from completing purchases. Anthropic has released blueprints for building shopper and merchant agents on Claude, giving retail, travel and ticketing companies guidance for systems that can recommend products, fill carts and advise internal teams.

2 min readRead story
LaunchTools

Google Launches Fairwind to Give Selected Defenders AI Tools for Finding and Fixing Flaws

The program draws a line between broadly available code-security tooling and restricted access to Gemini 3.8 Flash Cyber, reflecting the dual-use risk of AI that can identify and repair software weaknesses. Google has launched the Fairwind Program, a limited-access channel for governments and trusted partners to use AI systems that can find, verify and fix software vulnerabilities.

3 min readRead story
LaunchProducts

CrowdStrike’s SafeMind Pits Red and Blue AI Agents in an NVIDIA Infrastructure Twin

The new Falcon system combines CrowdStrike’s security data and agent harnesses with NVIDIA’s open Nemotron models. Early tests favor the optimized open pipeline on detection quality, but cover only one attack scenario family. CrowdStrike has introduced SafeMind inside its Falcon platform, pairing its cybersecurity models and agent harnesses with NVIDIA Nemotron open models.

3 min readRead story
LaunchProducts

Altruist Puts Six-Part Financial Planning in Hazel, With Separate Math Programs

The new agent gives advisors a way to build and adjust comprehensive plans during client meetings. Altruist’s key design claim is that AI handles client context while separate software produces the financial figures. Altruist has added a financial planning agent to Hazel that brings retirement, investments, cash flow, estate, tax, insurance, and risk work into one advisor workflow.

2 min readRead story
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
LaunchTools

Google Updates Antigravity Teamwork, Putting Long-Running Agent Critique in Paid Preview

The update packages critique, verification and synthesis into reusable agent-team patterns. Google’s results suggest the orchestration can produce useful work, while the preview’s lower-parallelism setup and model mix limit direct comparisons with its research runs. Google has updated Teamwork, its Antigravity framework for AI-agent teams that can work through complex tasks for hours or days.

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
18 more next