AI News
Daily AI industry news — funding, products, policy, and major moves from global AI companies and startups.
NTSB Shuts Down Data System After AI Restores Pilot Voices from Images
The NTSB temporarily closed its docket system after AI restored pilot audio. Users combined spectrogram images and transcripts to recreate v
Cohere Command A+ Brings 218B MoE to Just Two H100 GPUs
Cohere released Command A+ as an Apache 2.0 open-source model. The 218B MoE architecture runs on two H100s using W4A4 quantization. Integrat
Google Pixel's Disco Ball Theme Shifts AI Icons From Color to Style
Google introduced a Disco Ball icon theme for Pixel smartphones. The update evolves AI customization from simple color matching to complex s
The $28 Trillion Market Vision Behind SpaceX's $1.75 Trillion IPO
SpaceX filed an S-1 for an IPO with a target valuation of $1.75 trillion. The company defines its total addressable market as $28 trillion.
Indirect Prompt Injection: The Invisible Commands Hijacking LLMs
LLMs are vulnerable to hidden instructions called indirect prompt injection. Attackers use invisible text to override system prompts in RAG
Why Google AI Overviews Fail the 'Disregard' Search Test
Google AI Overviews are pushing traditional search links further down. A search for 'disregard' results in a useless blank space at the top.
Why AI-Driven Headcount Reduction Erases Institutional Knowledge
AI-driven layoffs risk erasing critical institutional knowledge. True innovation comes from augmenting human capability, not cutting costs.
Android XR Glasses: Why Google is Outsourcing Hardware to Samsung and Gentle Monster
Google unveiled Android XR glasses developed with Samsung and Gentle Monster. The lineup includes an audio-only model and a prototype displa
Steve Wozniak on the Critical Gap Between AI and Actual Intelligence
Steve Wozniak argues that AI is a mimicry of human intelligence. He defines AI as the result of trillions of repeated routines. Wozniak urge
Why D&B Rebuilt Its 642 Million Record Database for AI Agents
D&B redesigned its 642 million record database for AI agents. The system uses MCP and A2A protocols to ensure sub-second latency. A new KYA
Antigravity 2.0 and Gemini 3.5 Flash Lead 3D Spatial Coding Benchmarks
Google released Antigravity 2.0 and Gemini 3.5 Flash for 3D spatial coding. The system achieved a top quality score of 4.5 in the Pantheon b
delta-mem Implements AI Working Memory With Just 0.12% More Parameters
Mind Lab researchers introduced delta-mem to improve AI working memory. The system uses only 0.12% additional parameters to store context. I
Why Developers are Fighting AI Proxy Communication on GitHub and Reddit
Developers are reporting a rise in AI-generated proxy answers on GitHub. Workplace communication is suffering from unverified ChatGPT screen
The 440MB Hy-MT2 Model That Outperforms Commercial Translation APIs
Tencent released Hy-MT2, a multilingual model supporting 33 languages. The 1.8B version is compressed to 440MB and beats commercial APIs. On
Why Google's AI-First Search is Driving Power Users to Alternative Engines
Google announced a massive AI-centric overhaul of its search engine at I/O 2026. Developers are rejecting forced AI summaries in favor of co
The 0.07 Score That Proves Humans Cannot Detect Deepfakes
A study shows US citizens have a 0.07 score in detecting deepfakes. Half of respondents overstate their ability to spot AI-generated media.
Why Spotify and UMG Chose Licensing Over AI Copyright Lawsuits
Spotify and UMG launched an AI cover tool for Premium subscribers. The model uses a revenue-share system to compensate participating artists
Google Gemini Ads: Turning AI Search into a Sales Engine
Google is integrating Gemini-powered ads directly into AI search results. New AI Max tools enable conversational discovery and native checko
Why Microsoft Revoked Claude Code Licenses for the E+D Team
Microsoft is revoking Claude Code licenses for its E+D engineering team. The company aims to consolidate agentic tools under GitHub Copilot
Why Prompt Caching Fails: The 46x Cost Trap in AI Coding Agents
AI coding agents often face a 46x cost surge due to prompt cache invalidation. Sending full transcripts triggers context window compaction a
Why AI Answer Engines Are Killing Traditional SEO
AI search engines now prioritize synthesized answers over simple link lists. Technical accessibility and structured data are the new gatekee
AgentMemory Solves the 200-Line Limit for AI Coding Agents
AgentMemory achieves a 95.2% R@5 score on the LongMemEval-S benchmark. The system reduces token usage by 92% through hierarchical compressio
Google Gemini Ultra: The $100 Barrier to an AI Agent Ecosystem
Google introduced a fragmented AI agent ecosystem at Google I/O. The Gemini Ultra plan requires a monthly subscription of 100 dollars. New a
Utilyze Exposes the 'Fake 100%' GPU Utilization Trap
Utilyze provides real-time GPU utilization by reading hardware performance counters. The tool introduces Attainable SOL to determine realist
Salesforce's Headless Bet: Moving the SaaS Moat to the Data Layer
Salesforce is launching headless products to prioritize API access for AI agents. The traditional SaaS moat of UI muscle memory is rapidly d
AI Scrapers Drive Wiki Server Costs Up 100x Through Stealth Crawling
AI scrapers are causing 95% of server instability for open wikis. Stealth bots use residential proxies to rotate one million IPs daily. Forc
Why a 40% Increase in AI-Generated Code Signals a Productivity Drop
Quantitative metrics often mistake AI verbosity for actual productivity. Real-world tasks show a 19% productivity drop for experienced devel
Studio by Spotify Labs Turns Personal Calendars Into AI Podcasts
Spotify launched Studio by Spotify Labs as a desktop AI application. The tool creates personalized audio briefings from emails and calendars
Google AI Overviews and the Shift Toward an Answer-Based Web
Google is shifting search from providing links to delivering AI answers. AI Overviews show a 10% error rate while stripping content context.
The AI Rebellion: Why Corporate Optimism Is Meeting User Resistance
A growing AI Rebellion is emerging as users reject low-quality AI slop. Pew Research shows 53% of US adults fear AI harms human creativity.