Live Feed
Apple Sends Legal Letters to Dozens of OpenAI Employees in Trade Secret Escalation
Weeks after suing OpenAI and naming two senior employees, Apple has sent legal letters to dozens more current OpenAI workers it alleges left Cupertino carrying confidential information. The action broadens the IP dispute from a corporate lawsuit into an individual-employee campaign.
GPT-5.6 Is Burning Through $200/Month Codex Pro Accounts in Hours — OpenAI Has No Fix
Pro subscribers report GPT-5.6 consumes dramatically more tokens for equivalent tasks than GPT-5.5, with some losing their full 5-hour weekly allowance after two tasks and over $100 in additional purchased credits in a single session. OpenAI support calls it normal.
AI Absorbed $222B and 66% of US Venture Capital in 2025 — Up From 10% a Decade Ago
PitchBook's 2025 AI investment landscape puts a precise number on the structural shift: AI captured 65.6% of all US VC deal value last year, up from 47.2% in 2024 and 10% in 2015. US AI unicorns now aggregate to $4.3 trillion in value.
NVIDIA Nemotron-3-Embed Tops RTEB at 78.5%: Three Models to Fix Agent Retrieval
NVIDIA ships three open embedding models targeting the retrieval bottleneck in agentic AI. The 8B flagship takes RTEB #1 at 78.5%; two 1B variants cover cost and latency tiers with a Blackwell-optimized build hitting 2x throughput at 99%+ of BF16 accuracy.
LMCache Cuts Agentic Inference Latency 3x — KV Cache Layer Addresses Coding Agent's Prefix Recompute Problem
LMCache is a tiered KV cache management layer for vLLM that eliminates repeated prefix recomputation in agentic workloads. At 32 concurrent users on 100K-token coding agent traces, it delivers 3x lower average latency and 2.3x more completed requests versus GPU-only prefix caching.
Google Renames NotebookLM to Gemini Notebook, Previews Gemini 3.5 and Antigravity Integration
NotebookLM is now Gemini Notebook. Google is rolling the tool into its Gemini product family with a name change effective July 16, and previewing a Gemini 3.5 model upgrade plus Antigravity coding execution capabilities for AI Pro subscribers.
Japan and NVIDIA Launch World's First National AI Infrastructure: 27,500 Rubin GPUs, 140MW
Noetra Corp, backed by Japan's Ministry of Economy, Trade and Industry, is building a 140MW NVIDIA Vera Rubin AI factory with 27,500 GPUs and 13,750 Vera CPUs. The facility will train open multimodal foundation models for physical AI across manufacturing, logistics, healthcare, and telecommunications.
Moonshot AI Ships Kimi K3: 2.8T Open-Weight Multimodal Model at $3/$15 Per Million
Kimi K3 is a 2.8T-parameter open-weight multimodal reasoning model targeting complex coding, knowledge work, and long-horizon agentic workflows. At $3/M input and $15/M output, it lands below Opus 4.8 pricing with a 1M-token context window and self-reported parity ambitions.
Thinking Machines Inkling Debuts as US Open-Weights Leader at AA Index 41 — Beats Kimi K2.6 on Agentic Benchmarks
Thinking Machines' first production model lands at 41 on the Artificial Analysis Intelligence Index, 3 points clear of NVIDIA Nemotron 3 Ultra. The 975B-parameter MoE accepts text, image, and audio, leads Kimi K2.6 and DeepSeek V4 Flash on both GDPval-AA and tau-Banking, and ships open weights on HuggingFace.
Schlumberger and a Fracking Company Are Building AI Data Centers Now
SLB (formerly Schlumberger) and Liberty Energy have announced a strategic alliance to deliver modular infrastructure and integrated behind-the-meter power generation for AI data center projects globally. The oil field services industry has officially entered the AI compute buildout.
DriveNets Connects Two Data Centers 52 Miles Apart Into One GPU Cluster — 111.2 Tbps, 0.9ms
WhiteFiber's Project Redwood is the first commercial AI supercluster to span geographically separated data centers at production scale. DriveNets' AI Fabric delivered 111.2 Tbps of validated bandwidth with sub-millisecond guaranteed latency across a 52-mile span.
OpenAI Trained an AI to Attack Its Own Models. Then Used the Attacks to Train GPT-5.6.
GPT-Red is OpenAI's internal automated red-teaming model. It found prompt injection attacks that broke previous models cold. Those same attacks went into GPT-5.6's training data — and GPT-5.6 is now resistant to them.
Crusoe and Lancium Lock 1GW AI Campus in Childress, Texas — Meta Tipped as Anchor Tenant
Crusoe and Lancium announced a 1.0 gigawatt AI data center in Childress, Texas on 270 Lancium-owned acres. Construction begins Q3 2026. Reports point to Meta as the hyperscale anchor.
DeepSeek Revenue Nears $500M a Year as Founder Controls a 2027 Shanghai IPO
The Information reports DeepSeek's annualized revenue has reached $400-500M with margins above 50% on V4 API access. Liang Wenfeng holds all decision authority under a 5-year lockup, with a STAR Market listing targeted for 2027.
DeepMind CEO Says AGI Is 'a Few Short Years Away,' Calls for US-Led Global AI Oversight
Demis Hassabis published a framework arguing AGI is imminent at '10x the Industrial Revolution at 10x the speed,' and the US should lead a global body with authority to slow or halt frontier model deployment.