KO EN

AX BRIEF

AI news, benchmarks & engineering blog curation

● LIVE
지능1. Claude Fable 5.1 (max with fallback) 100 pt2. Claude Fable 5 (with fallback) 96 pt3. Claude Opus 5 (max) 95 pt4. GPT-6 Astra (max) 95 pt5. GPT-5.6 Sol (max) 91 pt코딩1. Claude Fable 5.1 (max) 100 pt2. Claude Opus 5 (xhigh) 94 pt3. GPT-6 Astra (max) 91 pt4. Muse Spark 1.3 (xhigh) 81 pt5. Grok 4.5 (high) 81 pt이미지1. GPT Image 2 (high) 100 pt2. MAI-Image-2.6 93 pt3. GPT Image 1.5 (high) 91 pt4. Reve 2.1 88 pt5. Muse Image 85 pt비디오1. Gemini Omni Flash 100 pt2. Minimax H3 Max (post-trained by fal) 94 pt3. Dreamina Seedance 2.0 720p 91 pt4. Wan 3.0 85 pt5. gemini-omni-1.1-flash 85 pt가격1. Claude Fable 5.1 (max with fallback) 100 pt2. GPT-6 Astra (max) 100 pt3. Claude Fable 5 (with fallback) 100 pt4. Claude Opus 5 (max) 49 pt5. GPT-5.6 Sol (max) 39 pt속도1. Gemini 3.5 Flash-Lite 100 pt2. Gemini 3.8 Flash (high) 66 pt3. Muse Spark 1.3 (max) 46 pt4. gpt-oss-120b (high) 42 pt5. GPT-5.6 Luna (max) 27 pt

AI Engineering Blogs

Curated deep insights from tech leaders and researchers — engineering, product, and strategy.

MIT and Meta Develop Flexible Gel with 400x Light-Triggered Conductivity

MIT and Meta Develop Flexible Gel with 400x Light-Triggered Conductivity

MIT and Meta Reality Labs created a flexible gel that boosts conductivity 400 times using light. The material utilizes photo-ion generators

DiffusionBlocks: Sakana AI's Method to Slash Training Memory to 1/B

DiffusionBlocks: Sakana AI's Method to Slash Training Memory to 1/B

Sakana AI introduced DiffusionBlocks to reduce training memory to 1/B. The method treats residual networks as discretized diffusion ODEs. It

Why Delta Line Integrated Motion Control into the IBIX57 Motor

Why Delta Line Integrated Motion Control into the IBIX57 Motor

Delta Line launched the IBIX57 BLDC motor with an integrated controller. The 57mm design reduces wiring complexity and system footprints. Fo

NVIDIA Vera Rubin: Shifting the Enterprise Focus to AI Factories

NVIDIA Vera Rubin: Shifting the Enterprise Focus to AI Factories

NVIDIA introduced the Vera Rubin platform to build full-stack AI factories. The system uses Omniverse digital twins to optimize power and co

Verizon Connect Deploys Agentic AI to Process 500M Daily Data Points

Verizon Connect Deploys Agentic AI to Process 500M Daily Data Points

Verizon Connect deployed Agentic AI to 100,000 Reveal platform users. The system processes 500 million daily data points from 1.2 million ve

DeepThe Autonomous Loop Driving OpenAI's Tax AI to 86% Field Accuracy

The Autonomous Loop Driving OpenAI's Tax AI to 86% Field Accuracy

OpenAI developed Tax AI to automate complex tax filing processes. The system uses an autonomous loop to turn expert feedback into code. Fiel

DeepNVIDIA's Polar Framework Enables RL Training Without Code Modification

NVIDIA's Polar Framework Enables RL Training Without Code Modification

NVIDIA released Polar to perform reinforcement learning on AI agents without code changes. The framework uses a gateway proxy to capture tra

DeepReachy Mini Moves to Fully Local Voice AI Without Cloud APIs

Reachy Mini Moves to Fully Local Voice AI Without Cloud APIs

Reachy Mini now supports a fully local four-stage voice processing pipeline. The system eliminates cloud latency and token-based costs for r

Amazon Quick Turns Day-Long Reporting into 45-Minute Tasks

Amazon Quick Turns Day-Long Reporting into 45-Minute Tasks

Amazon Quick automates document creation using AWS data sources. The tool generates editable files in Word, Excel, and PowerPoint. Reporting

DeepAWS Strands Agents Cuts AI Agent Development to 30 Lines of Code

AWS Strands Agents Cuts AI Agent Development to 30 Lines of Code

AWS released Strands Agents to simplify autonomous AI development. The framework allows building complex agents with only 30 lines of code.

DeepHaier's 1.75kg W3 Exoskeleton Leverages AI for Precision Mobility

Haier's 1.75kg W3 Exoskeleton Leverages AI for Precision Mobility

Haier launched the 1.75kg W3 exoskeleton for elderly mobility support. The device uses AI algorithms to provide 16Nm of torque per leg. Haie

MEMO Framework Updates LLM Knowledge Without Changing Model Parameters

MEMO Framework Updates LLM Knowledge Without Changing Model Parameters

The MEMO framework separates knowledge storage from reasoning. It outperforms RAG on NarrativeQA with 53.58% accuracy. TIES merging reduces

Google I/O 2026: 10 Changgoo Alumni Pivot Toward Android XR and AI

Google I/O 2026: 10 Changgoo Alumni Pivot Toward Android XR and AI

Google invited 10 Changgoo alumni startups to I/O 2026 in Mountain View. The startups explored Android XR and Google Omni for spatial comput

DeepAgentWatch Automates AWS Infrastructure Monitoring via Ambient AI

AgentWatch Automates AWS Infrastructure Monitoring via Ambient AI

AWS introduced AgentWatch to automate infrastructure monitoring. The agent uses Claude Sonnet to send 15-minute summaries to Slack. A human-

DeepNVIDIA Vera CPU Challenges x86 Dominance With 1.2TB/s Bandwidth

NVIDIA Vera CPU Challenges x86 Dominance With 1.2TB/s Bandwidth

NVIDIA's new Vera CPU delivers 1.2TB/s memory bandwidth for AI workloads. The 88-core Olympus architecture outperforms x86 chips by 1.5x in

DeepAmazon Bedrock AgentCore Payments Removes the Paywall for AI Agents

Amazon Bedrock AgentCore Payments Removes the Paywall for AI Agents

Amazon Bedrock AgentCore payments enables autonomous micro-payments for AI agents. The system uses stablecoins to allow transactions below o

DeepEngineAI Scales T800 Humanoid Production to 10,000 Units Per Year

EngineAI Scales T800 Humanoid Production to 10,000 Units Per Year

EngineAI launched a mass production line for the T800 humanoid robot. The facility produces one robot every 15 minutes with 40% higher effic

DeepAmazon Quick's New Observability Solution Collapses AWS Logs Into One View

Amazon Quick's New Observability Solution Collapses AWS Logs Into One View

AWS launched an observability solution for the Amazon Quick AI platform. The system centralizes fragmented logs into an S3-based data lake.

DeepAI Drone Detects 0.1mm Cracks at Changwon NC Park Stadium

AI Drone Detects 0.1mm Cracks at Changwon NC Park Stadium

SierraBase deployed its Sirius Wing drone for autonomous facility inspection at Changwon NC Park. The drone uses 360-degree LiDAR SLAM to ma

Unitree's June 1 IPO Review Could Trigger a Humanoid Robot Rush

Unitree's June 1 IPO Review Could Trigger a Humanoid Robot Rush

Unitree faces a critical IPO review on June 1, 2026, in Shanghai. The company uses vertical integration for humanoid and quadruped robots. A

DeepOmniVoice Studio Brings 646-Language Local TTS to the Desktop

OmniVoice Studio Brings 646-Language Local TTS to the Desktop

OmniVoice Studio provides local voice cloning and TTS for 646 languages. The tool integrates an MCP server to connect with Claude and Cursor

DeepOpen-MM-RL Introduces Verifiable Rewards for Multimodal Reasoning

Open-MM-RL Introduces Verifiable Rewards for Multimodal Reasoning

TuringEnterprises released the Open-MM-RL dataset and pipeline. The system uses SymPy to provide mathematically verifiable rewards. It enabl

DeepNVIDIA FLARE: Why FedProx Beats FedAvg for Non-IID Data

NVIDIA FLARE: Why FedProx Beats FedAvg for Non-IID Data

NVIDIA FLARE compares FedAvg and FedProx using non-IID CIFAR-10 data. FedProx uses a proximal term to stabilize training against client drif

Electromate's Dobot Rollout Bridges Canada's Robotics Skills Gap

Electromate's Dobot Rollout Bridges Canada's Robotics Skills Gap

Electromate is introducing Dobot educational robots to the Canadian market. The initiative provides a structured ecosystem for mechatronics

DeepTogether AI's OSCAR Breaks the 2-Bit KV Cache Accuracy Barrier

Together AI's OSCAR Breaks the 2-Bit KV Cache Accuracy Barrier

Together AI released OSCAR to compress KV caches down to 2 bits. The system increases throughput by up to 7.83x on H100 GPUs. OSCAR maintain

DeepOpenAI Taps Folha de S.Paulo and UOL for Brazil's First Media Partnership

OpenAI Taps Folha de S.Paulo and UOL for Brazil's First Media Partnership

OpenAI partnered with Folha de S.Paulo and UOL to integrate Brazilian news. The deal uses grounding to reduce hallucinations for 50 million

DeepWhy AI Agents Require a Harness to Move Beyond LLM Text Generation

Why AI Agents Require a Harness to Move Beyond LLM Text Generation

LLMs require a harness and scaffold to function as autonomous agents. A harness manages the execution loop while a scaffold provides the blu

The Functional Fear and Sadness Found Inside Anthropic's AI Models

The Functional Fear and Sadness Found Inside Anthropic's AI Models

Anthropic co-founder Chris Olah revealed that AI models exhibit functional states of fear and sadness. The company is aggressively expanding

How Mimesis Exposes AI Bias Through Counterfactual Synthetic Data

How Mimesis Exposes AI Bias Through Counterfactual Synthetic Data

Mimesis allows developers to audit AI bias using synthetic data. The library creates counterfactual pairs to isolate protected attributes. T

DeepZest: The Dual-Arm Humanoid Solving the Chaos of Commercial Kitchens

Zest: The Dual-Arm Humanoid Solving the Chaos of Commercial Kitchens

Robotable unveiled Zest, a dual-arm humanoid robot specialized for F&B. The system uses a VLA model and multimodal sensors to handle cooking