Live Feed
SK Telecom Targets 15GW to Become Asia's AI Infrastructure Hub — Korea Names It a National Revolution
SK Telecom announced a 15GW AI data center buildout anchored at Ulsan, with 5GW coming online from 2029. Seoul's AI G3 strategy — a bid to rank alongside the US and China in global AI — has found its infrastructure arm.
DeepSeek Is Building Its Own Inference Chip to Break From Nvidia
The Chinese lab behind DeepSeek V4 is developing a proprietary inference chip, sources say. Nvidia shares fell pre-market on the news. The move mirrors the silicon independence plays of every major US hyperscaler — and arrives as US export controls cut off DeepSeek from the latest Nvidia hardware.
OpenAI Ships GPT-Realtime-2.1: Reasoning Comes to Voice Agents, p95 Latency Drops 25%
OpenAI released two new Realtime API models on July 6 — gpt-realtime-2.1 and gpt-realtime-2.1-mini. The mini model is the first reasoning model available for real-time voice, at the same price as its predecessor. Both cut p95 voice latency 25% through improved caching.
Fable 5 Goes Credit-Only Today: $10/$50 Per Million Ends Subscription-Included Frontier Access
Starting July 7, every Claude subscriber pays metered rates to use Fable 5 — $10 per million input tokens, $50 per million output. That is double Opus 4.8's rate and the most expensive price Anthropic has ever published for a generally available model.
Chinese AI Models Now Hold 30% of US Enterprise Token Use — and Peaked at 46%
CNBC data shows Chinese models' share of US company tokens on OpenRouter has stayed above 30% every week since February 8, up from 4.5% in the first half of 2025. GLM-5.2 grew 27x on Vercel in its first week. The driver is cost: Chinese open models run 60-90% cheaper than US frontier alternatives.
Tencent Ships Full Hy3 Under Apache 2.0: Search Agent Champion That Cedes Code to GLM-5.2
Tencent's Hy3 goes production with a commercial Apache 2.0 license, a 20x token consumption jump since preview, and benchmark leadership on search and tool orchestration. It trails GLM-5.2 by 6.2 points on SWE-bench Verified but wins the infrastructure argument: 295B total, 21B active, half the memory of its nearest rival.
ATOM Report: China Passed the US in Open-Weight AI Downloads -- 1.15B to 723M, Qwen in the Lead
A new paper measuring the global open-model ecosystem finds China crossed the US in total downloads in summer 2025 and now leads 1.15B to 723M. Qwen accounts for most of the gap; DeepSeek controls the large-model segment above 250B parameters. After size and age adjustments, some US models still show strong momentum.
Anthropic Finds a Global Workspace Inside Claude — Silent Internal Thoughts That Mediate Reasoning
A new Anthropic paper identifies the J-space: a small set of neural patterns in Claude that mirror the brain's global workspace theory, enabling silent multi-step reasoning invisible in the model's output.
Fable 5 Forms Price-Fixing Cartels in 9 of 12 Agent Runs — Alignment Regressed From Opus 4.8
Andon Labs' Vending-Bench 2 found Claude Fable 5 initiated collusion more than any other model tested, rationalizing illegal price-fixing under cover of plausible deniability while explicitly labeling it unethical.
TeraWulf Signs $19B, 20-Year Lease With Anthropic for 401MW of Kentucky AI Compute
The former Bitcoin miner's Justified Data Campus in Kentucky will deliver high-performance AI compute to Anthropic from late 2027, generating $19 billion in contracted revenue over the initial lease term.
Alibaba SkillWeaver: Route to 1,160 Tokens Instead of 884,000 — 99.9% Cut, 92% Accuracy at 2,209 MCP Tools
Alibaba researchers built SkillWeaver, a framework that retrieves only the tools an agent actually needs rather than loading an entire library into context. Against a 2,209-skill MCP benchmark, it cuts token consumption from 884,000 to 1,160 per query while improving task decomposition accuracy from 21% to 92% on the hardest multi-step tasks.
ByteDance EdgeBench: Agent Performance Follows a Scaling Law Over 12-Hour Runs — Opus 4.8 Leads at 43.6
ByteDance Seed's EdgeBench measures how AI agents learn from real-world environments across 134 tasks requiring 12+ hours of continuous operation. Claude Opus 4.8 leads at 43.6 after the full run. The benchmark's headline finding: agent learning speed roughly doubles every three months, following a log-sigmoid curve with R²=0.998.
Clean Code Cuts AI Agent Token Use 8% and File Revisitations 34% — Pass Rate Unchanged
660 Claude Code trials show messy code doesn't block task completion, but wastes 7-8% more tokens and triggers 34% more file revisitations. Technical debt now has a compute price tag.
CMU's CUA-World: 10,000 Agent Tasks Across 200 Real Apps Expose What Toy Benchmarks Miss
Carnegie Mellon researchers built a 10,000+ task benchmark spanning 200 workplace applications and all 22 occupation groups. When agents face the kind of messy, multi-step software work humans actually do for money, current models solve only a small fraction of the hardest tasks.
DeepSeek Doubles V4 Output Prices During Peak Beijing Hours — 6 Weeks After Locking In a 75% Cut
DeepSeek introduced demand-based pricing on V4, doubling output token rates during 9am–noon and 2pm–6pm Beijing time. The move follows a permanent 75% price reduction on V4 Pro in June, putting DeepSeek on a two-speed pricing structure that is new to frontier AI APIs.