Espresso

Sign in Subscribe

Artificial Intelligence

DeepSeek’s mHC Architecture Boosts LLM Efficiency, OpenAI Enters Hardware with AI-Pen, and A-QCF-Net Clears FDA Threshold for Tumor Segmentation

DeepSeek’s mHC Architecture Boosts LLM Efficiency, OpenAI Enters Hardware with AI-Pen, and A-QCF-Net Clears FDA Threshold for Tumor Segmentation

TL;DR * DeepSeek releases mHC architecture, boosting LLM performance by 4.4% on MMLU with 6.27% hardware overhead * OpenAI acquires io for $6.5B, launching first hardware product—an AI pen—expected in early 2026 * Google Gemini surpasses ChatGPT in daily active users and tokens consumed in Q1 2026

Intel, AMD, and HP Redefine Laptop AI with 12+ TOPS NPUs, Triple-Chip Strategy, and Copilot-Optimized Performance

Intel, AMD, and HP Redefine Laptop AI with 12+ TOPS NPUs, Triple-Chip Strategy, and Copilot-Optimized Performance

TL;DR * Intel and AMD Launch AI-Optimized Laptop Chips (Panther Lake, Ryzen AI 300) at CES 2026 with NPU Integration for Copilot * OpenAI’s $1.5M Average Stock Compensation Per Employee Drives AI Talent War Amid 50% Revenue Allocation to Equity * Cisco and Intel Unveil Joint AI Workload Platform Using

MAI-UI 235B Dominates AndroidWorld Benchmark as Gumdrop AI Peripheral and Tesla Optimus Struggle Highlight AI Agent Divide

MAI-UI 235B Dominates AndroidWorld Benchmark as Gumdrop AI Peripheral and Tesla Optimus Struggle Highlight AI Agent Divide

TL;DR * MAI-UI 235B AI agent achieves 76.7% success on AndroidWorld benchmark via MCP tools and self-evolving navigation pipeline * SK hynix to build $3.9B HBM module plant in Indiana with CHIPS Act funding to supply Nvidia Rubin AI servers by 2028 * OpenAI and Jony Ive’s 'Gumdrop&

Meta Buys Manus AI for $2B, China Enforces AI Safety Rules

Meta Buys Manus AI for $2B, China Enforces AI Safety Rules

TL;DR * Meta acquires Manus AI for $2 billion, integrating 147T-token AI agent platform with 80M virtual computers into its ecosystem * MiniMax releases M2.1 model with 42.8% HLE benchmark improvement, surpassing GPT-5 on mathematical reasoning tasks * China drafts AI rules requiring emotional chatbots to disclose bot identity, ban

China Drafts AI Companion Rules with 2-Hour Reminders, U.S. Influence, and Global Compliance Shifts Amid NVIDIA Feynman

China Drafts AI Companion Rules with 2-Hour Reminders, U.S. Influence, and Global Compliance Shifts Amid NVIDIA Feynman

TL;DR * SK Telecom unveils A.X K1, South Korea’s first 519B-parameter AI model optimized for Korean language, enabling teacher-model knowledge distillation to smaller models * LG Electronics launches DQ-C2 in-house AI semiconductor with enhanced CPU and memory, enabling complex AI computations in smart home appliances * China's Cyberspace

NVIDIA Launches RTX 5090 with 15x AI Training Boost; Olympus AI Migrates 40K Enterprises to AWS Bedrock in 6 Hours

NVIDIA Launches RTX 5090 with 15x AI Training Boost; Olympus AI Migrates 40K Enterprises to AWS Bedrock in 6 Hours

TL;DR * NVIDIA Launches RTX 5090 GPU with 10x–15x Performance Leap Over Hopper Architecture, Dominating High-End AI Training Market * Amazon Unveils Olympus LLM with 5T Parameters and 16M-Token Context, Deploying 100K Trainium3 Chips to Power Enterprise AI Workloads * MIT Releases DisCIPOL Method, Reducing LLM Reasoning Costs by 80.2%

Meta's Llama 4 Falters Against Gemini 3 and GPT-5.2, Triggering $14.3B Pivot and Layoffs Amid AI Race

Meta's Llama 4 Falters Against Gemini 3 and GPT-5.2, Triggering $14.3B Pivot and Layoffs Amid AI Race

TL;DR * Meta's Llama 4 model launch underperforms in coding and reasoning, triggering internal restructuring and job cuts amid pressure from Google's Gemini 3.0 and OpenAI's GPT-5.2 * Broadcom's AI semiconductor revenue surges 74% YoY to $8.2B in Q4 2025,