AI News
Daily AI industry news — funding, products, policy, and major moves from global AI companies and startups.
Claude Platform on AWS Bridges the Gap Between Native APIs and Cloud Scale
Anthropic launched Claude Platform on AWS with native API capabilities. The service integrates AWS IAM and CloudTrail for enterprise securit
Anthropic Open-Sources 10 Financial AI Agents to Automate Workflows
Anthropic released 10 open-source financial AI agents under Apache 2.0. The agents integrate with 11 data providers via the Model Context Pr
TML-Interaction-Small and the 200ms Micro-Turn for Real-Time AI
Thinking Machines Lab released TML-Interaction-Small for real-time collaboration. The model achieves a 0.40-second turn-taking latency on FD
Gemini Intelligence Turns Android into an Autonomous AI Agent
Google is integrating Gemini Intelligence into Android for multi-step automation. The system introduces generative widgets created via natur
Claude Code Agent View Now Manages Parallel AI Sessions in One Screen
Claude Code introduces Agent View to manage multiple background sessions. The system uses git worktrees to isolate file changes between agen
Googlebook: How Gemini AI is Replacing the Traditional Laptop OS
Google announced Googlebook, a Gemini-integrated laptop category. The device launches this autumn through five major hardware partners. Andr
Claude Code's /goal Command Shifts AI from Assistant to Agent
Anthropic added a /goal command to Claude Code for autonomous iteration. A fast model evaluates progress based on chat history to drive the
Why MacSync Malware is Now Hiding in Claude Shared Chats
Attackers are using Claude shared chats to spread MacSync malware. The malware steals macOS Keychain data using memory-resident execution. m
Israel's AI Targeting System Cuts Analysis Time From Weeks to Seconds
Israel uses AI to turn massive data pipelines into real-time targets. Palantir Maven and Stingray devices automate the identification proces
Needle: The 26M Parameter Model Mastering Gemini Tool Calling
The Needle model enables Gemini-level tool calling with only 26M parameters. It outperforms larger models like Qwen-0.6B in single-shot func
Gemini AI Pointer Turns the Mouse into a Contextual Interface
Google introduced an AI-powered pointer driven by Gemini. The tool replaces manual prompting with visual context awareness. Magic Pointer ai
Supertonic 3 Brings 31-Language TTS to Local CPUs
Supertonic 3 enables high-quality TTS on local CPUs without GPUs. The model supports 31 languages with a lightweight 99M parameter design. I
Why Google Integrated Rambler AI Dictation Directly Into Gboard
Google introduced Rambler AI dictation for Gboard at the Android Show. The Gemini-powered tool removes filler words and supports code-switch
Why Amazon Employees Are Token-Maxing to Hit AI Performance KPIs
Amazon employees are intentionally inflating AI usage to meet performance KPIs. This practice of token-maxing prioritizes usage metrics over
Perceptron Mk1 Slashes Video Analysis Costs by 90%
Perceptron released the Mk1 model for high-efficiency video analysis. The model reduces operational costs by 90% compared to frontier LLMs.
Anthropic Deploys MCP Connectors to Automate Legal Workflows
Anthropic launched legal-specific plugins and MCP connectors for Claude. The tools integrate with Westlaw, DocuSign, and Box for paid users.
Anthropic Voids Stock Trades Across 8 Unofficial Platforms
Anthropic declared stock trades on eight unofficial platforms void. The company banned the use of SPVs for acquiring its shares. Retail inve
Why Google and SpaceX are Planning Orbital Data Centers for AI
Google and SpaceX are discussing the creation of orbital data centers for AI. SpaceX aims to lower AI compute costs ahead of a $1.75 trillio
Googlebooks and Gemini Integration Transform the Android Ecosystem
Google launched Googlebooks laptops with deep Gemini AI integration. Android updates now enable AirDrop compatibility and iOS migration. Gem
Meta AI Brings Real-Time Web Search to Threads via @meta.ai Mentions
Meta is testing @meta.ai mentions for real-time info on Threads. The beta is currently active in five countries including Singapore. Meta cl
Tessera Brings Persistent Workspace Management to Claude Code and CLI Agents
Tessera provides an open-source workspace for CLI AI coding agents. The tool integrates Git worktrees to isolate AI-driven code changes. It
OpenOSINT Integrates 9 Intelligence Tools Into One AI Agent
OpenOSINT automates nine security tools into a single AI agent. The framework uses Anthropic's API to eliminate AI hallucinations. It integr
Kuku Transforms Local Markdown Folders Into AI Knowledge Bases
Kuku is a macOS app that turns local Markdown folders into an AI second brain. The tool uses a diff-based editing system to keep users in co
The LoPE Technique That Raised Qwen3-4B Math Performance by 22%
Researchers introduced LoPE to improve small model reasoning. Qwen3-4B saw a 22% performance boost on AMC 2023 benchmarks. The method uses L
axon Brings Local LLM Orchestration to Low-Spec Hardware
A developer released the alpha version of axon for local AI orchestration. The tool uses a multi-agent hierarchy to automate software develo
Why AI is Turning Software Engineering Into a Professional Sport
AI tools increase short-term productivity but may erode long-term skills. The lack of labor protections forces engineers to prioritize immed
Gemini API File Search Now Supports Multimodal Queries and Page Citations
Google updated Gemini API File Search with multimodal and metadata support. New page citation features allow users to verify PDF sources dir
LiteRT-LM-Unity Enables GPU-Accelerated On-Device LLMs for Android
LiteRT-LM-Unity allows Unity developers to run LLMs on Android GPUs. The tool replaces slow CPU-based inference with Model Tensor Parallelis
Why Mythos Reported 5 curl Vulnerabilities But Only Confirmed One
Anthropic's Mythos AI reported five vulnerabilities in the curl codebase. Only one low-severity flaw was confirmed and fixed in version 8.21
Why Qwen 3.5 9B is the Sweet Spot for M4 MacBook Pro 24GB
Qwen 3.5 9B runs at 40 tokens per second on M4 MacBook Pro 24GB. The model supports a 128K context window and integrated thinking mode. Loca