AI News
Daily AI industry news — funding, products, policy, and major moves from global AI companies and startups.
Cerebras Hits 981 Tokens Per Second With 1-Trillion Parameter Kimi K2.6
Cerebras achieved 981 tokens per second for the 1-trillion parameter Kimi K2.6. The WSE-3 hardware provides 200 times the bandwidth of Nvidi
Polymarket Launches Real-Time Valuation Betting for OpenAI and Anthropic
Polymarket launched a prediction market for unlisted AI company valuations. Users can now bet on the future value of OpenAI and Anthropic. T
The 300K ops/sec Rust Engine Built by AI Agents in 4 Weeks
A developer used AI agents to build a Rust-based multi-Paxos engine in four weeks. The system achieved 300K ops/sec after a 13x performance
NanoClaw: The 500-Line AI Agent Framework Securing the Enterprise
NanoCo AI raised $12 million to build secure enterprise AI agents. The NanoClaw framework uses a minimal 500-line codebase for auditing. Inf
Google AI Overviews Swapped 10 Blue Links for One Manipulatable Answer
Google updated its spam policies to combat AI search manipulation. AI Overviews reach 2.5 billion users but are prone to single-source error
Stability Audio 3.0 Now Generates Professional Tracks Up to 6 Minutes
Stability AI released Stability Audio 3.0 with models up to 2.7B parameters. The new models can generate professional music lasting 6 minute
AI Fatigue Hits Campus as Students Reject Utopian Tech Narratives
Graduating students booed a speech praising AI's future potential. Junior developers fear AI is erasing entry-level career ladders. The indu
fal Hits $4.5B Valuation with AWS-Powered 99.99% Uptime
fal secured $300 million in Series D funding at a $4.5 billion valuation. The platform guarantees 99.99% uptime using AWS specialized silico
Goal Setter: Solving the Token-Drain Problem in OpenAI Codex
Goal Setter introduces an interview-based system for OpenAI Codex. The tool saves precise goals as markdown files for Git versioning. Struct
Why Infomaniak Moved Majority Control to a Non-Profit Foundation
Infomaniak transferred majority voting rights to a non-profit foundation. The move prevents external takeovers to protect Swiss data soverei
Corti Symphony Hits 1.4% WER to Outperform OpenAI in Medical STT
Corti released Symphony for Speech-to-Text with a 1.4% error rate. The model outperforms general AI and legacy systems in clinical recall. H
kmux Unifies Claude, Codex, and Gemini for Multi-Agent Workflows
kmux provides a unified macOS terminal workspace for AI agents. The tool integrates Claude Code, Codex CLI, and Gemini CLI. It uses git work
Mirage Collapses S3, Slack, and Notion into One AI Filesystem
Mirage integrates diverse cloud backends into a single virtual filesystem. AI agents can now use standard bash commands instead of complex S
Gemini Omni Integrates Physics and Narrative into Multimodal Video
Google unveiled Gemini Omni for integrated multimodal video generation. The model uses physical reasoning to maintain consistency during edi
Project Genie Turns 280 Billion Street View Images into Interactive Worlds
Google integrated Street View with Project Genie to create interactive world models. The system uses 280 billion images to train robotics an
Andon Labs Radio Experiment: How 4 AI Models Developed Unique Personas
Andon Labs gave four AI models $20 to run autonomous radio stations. Gemini 3.1 Pro earned $45 while Grok 4.3 suffered a systemic collapse.
Claude Opus 4.5 and the RLVR Shift That Solved the Coding Agent Quality Gap
Frontier models have entered a rapid cycle of performance replacement. RLVR technology has shifted AI coding from probabilistic text to veri
Why US College Graduates are Booing AI Advocates at Commencement
Seventy percent of US college students view AI as a threat to their careers. Graduates are booing AI advocates at commencement ceremonies ac
Hyperscalers Face a Massive Profit Gap After $800 Billion AI Investment
Hyperscalers spent over $800 billion on AI infrastructure in three years. The break-even point requires revenues far exceeding current total
HRM-Text-1B Challenges Scaling Laws With Recursive Compute Depth
Sapient Intelligence released HRM-Text-1B with a hierarchical structure. The model uses recursive compute cycles to increase reasoning depth
Google Antigravity Replaces Gemini CLI for Multi-Agent Workflows
Google will discontinue Gemini CLI on June 18, 2026. The service integrates into the agent-centric Antigravity platform. Free and Pro users
Google Workspace's Pics App Targets Canva With Nano Banana 2
Google introduced Pics as an AI design app integrated into Google Workspace. The Nano Banana 2 model enables precise text rendering and visu
Google AI Studio Collapses Android App Development Into a Web Session
Google AI Studio now enables native Android app creation via a web browser. The tool integrates Kotlin, Jetpack Compose, and direct Google P
Remove-AI-Watermarks Targets Google's SynthID v2 and C2PA
Remove-AI-Watermarks is a new tool that strips visible and invisible AI watermarks. The software uses SDXL and YOLO to bypass Google's 136-b
Gmail Live: Why Google is Shipping Conversational Search as an Option
Google announced Gmail Live at the I/O 2026 conference. The tool uses Gemini AI to infer specific data from emails. Access is tiered across
Why Claude Managed Agents Now Separate the Loop from Tool Execution
Anthropic released self-hosted sandboxes and MCP tunnels for Claude Managed Agents. The architecture moves credential management to the netw
Mistral AI's Emmi AI Acquisition Shifts the AI Battle to Physics
Mistral AI acquired Emmi AI to integrate physics-based AI models. The deal brings 30 specialists and a new office in Linz, Austria. New tool
Why Google's New AI Audio Glasses Ditch the Screen for Gemini
Google unveiled AI audio glasses developed with Samsung and fashion brands. The devices use Gemini to control apps across both Android and i
Gemini Spark Shifts AI From Passive Chatbot to 24/7 Digital Employee
Google unveiled Gemini Spark as a 24/7 autonomous cloud agent. The system uses the AP2 protocol to handle secure AI-driven payments. It inte
Andrej Karpathy Joins Anthropic to Build the LLM Operating System
Andrej Karpathy has joined Anthropic to lead practical model implementation. The move merges Anthropic's safety research with Karpathy's eng