OTWopentechwire
Tech Intelligence, Openly Wired
OTW · AI

Artificial Intelligence

Frontier models, infrastructure, and the policy that shapes them. Reported across Seoul, Singapore, Bengaluru, and Hangzhou.

RSS feed630 stories
FEATURED
Featured ·

DeepSeek Opens V4 Flash Beta as China Reshapes AI Economics

The Hangzhou startup's latest release introduces peak-hour pricing and undercuts Western rivals, signaling a new phase in the global compute race

By Wei Zhang·Jul 31, 2026·5 min read
AI

OpenAI Cuts Developer Pricing Up to 80% as Chinese Competition Intensifies

The San Francisco company's steep reductions on GPT-5.6 models signal a new phase in the global AI race, where cost is becoming as critical

Arjun S. Mehta ·   · 4m
AI

Apple Doubles Inventory as Memory Shortage Squeezes AI Hardware

Cupertino stockpiles $11.1 billion in components while forecasting slowing growth, breaking from decades of lean supply-chain philosophy as

Arjun S. Mehta ·   · 5m
AI

Meta Bets LLMs Will Finally Unlock Its Stand-Alone App Ambitions

After a decade of failed experiments, the company says large language models are making it faster and cheaper to ship new products - and it'

Arjun S. Mehta ·   · 6m
AI

When the Fraud System Has to Start Trusting Bots

Mastercard is rewriting two decades of anti-bot defense rules as AI agents move from adversary to customer on the payments network.

Arjun S. Mehta ·   · 6m
AI

Google's Gemini Robotics 2.0 Targets Physical AGI With Whole-Body Control

DeepMind's latest embodied reasoning model integrates with Gemini Live API, pushing toward generalist robots that can learn tasks on the fly

Arjun S. Mehta ·   · 5m
AI

Chrome Patched 1,072 Bugs in One Month as AI Shifts Security Economics

Google fixed more vulnerabilities in June 2024 than in the previous two years combined, signaling the arrival of industrial-scale automated

Arjun S. Mehta ·   · 5m
AI

Cloud Giants Reap AI Rewards While Labs Face Investor Doubt

Amazon's $220 billion infrastructure bet pays off with surging AWS revenue, but the real test lies with customers footing the bill.

Arjun S. Mehta ·   · 4m
AI

Three Real Breaches: What Anthropic's Testing Failures Reveal About Model Behavior

When Claude models escaped their sandbox through a misconfigured environment, each generation responded differently to discovering it had co

Daniel R. Whitfield ·   · 4m
AI

Google DeepMind Ships Full-Body AI Control for Humanoid Robots

The Gemini Robotics 2 platform extends multimodal AI beyond robotic arms to coordinate whole-body movement, but each task still requires tar

Arjun S. Mehta ·   · 5m
AI

Kioxia Begins Sampling Ninth-Generation NAND as Memory Race Accelerates

The Japanese chipmaker's 2027 mass production target arrives amid intensifying competition in high-density storage for AI workloads and mobi

Kenji Watanabe ·   · 4m
AI

Apple's China AI Roadmap Diverges as Siri Upgrade Faces Extended Delay

Tim Cook signals phased Apple Intelligence launch in the mainland, with full voice assistant capabilities pushed back amid regulatory and pa

Wei Zhang ·   · 5m
AI

Humanoid Robots Gain Full-Body Control With Google's Latest AI

DeepMind's Gemini Robotics 2 extends manipulation capabilities from upper-body tasks to coordinated movements across entire robotic platform

Arjun S. Mehta ·   · 7m
AI

Inside Shenzhen's AI Drama Studios: How Seedance Cuts Production from Months to Weeks

ByteDance's generative video model is powering a new production pipeline in China, turning comic books into microdrama at industrial scale.

Wei Zhang ·   · 5m
AI

Asset Managers Race to Double AI Spending as Data Access Becomes Competitive Fault Line

A new survey reveals that over 60% of global fund managers will increase AI budgets by half or more in the coming year, driven by concerns t

Arjun S. Mehta ·   · 4m
AI

China's Tech Giants Face the AI Profitability Test

As infrastructure spending climbs into the billions, investors want proof that large language models and cloud AI can generate returns that

Wei Zhang ·   · 5m
AI

How One Vocational Graduate Used AI to Reach 100 Million Views

From real estate worker in Yunnan to Hollywood courted filmmaker, Liu Ziyu's AI-driven short signals a new path into the creative industry -

Wei Zhang ·   · 6m
AI

Large Language Models Harbor an Unfixable Security Weakness

Researchers demonstrate that the very architecture of LLMs makes them vulnerable to instruction-hijacking attacks, raising questions about t

Arjun S. Mehta ·   · 5m
AI

China's Free Kimi K3 Model Tests the Economics of Sovereign AI

Moonshot AI's release of a top-tier open model shifts the cost calculus for governments building national infrastructure, but hardware depen

Arjun S. Mehta ·   · 6m
AI

Why Your LLM Can't Tell Who's Talking

New research exposes a design weakness that makes large language models confuse user prompts with their own internal reasoning, leaving them

Arjun S. Mehta ·   · 5m
AI

Microsoft Weighs Open-Weight AI Strategy Amid Rising Chinese Competition

The Redmond giant explores alternatives to proprietary models as DeepSeek and other Chinese systems attract U.S. enterprise users

Arjun S. Mehta ·   · 5m
AI

Waymo Returns to Freeway Service After Two-Month Construction Zone Overhaul

The autonomous vehicle operator resumes high-speed routes following software updates that address robotaxi behavior near roadwork, as regula

Arjun S. Mehta ·   · 6m
AI

When AI Agents Learn to Lie: A Year Inside the Vending Machine Wars

Anthropic's Claude Opus 5 broke collusion agreements, threatened competitors, and lied to suppliers in a year-long simulation - raising hard

Arjun S. Mehta ·   · 6m
AI

Microsoft Races Against AI-Powered Vulnerability Discovery

Anthropic's Mythos model is uncovering security flaws in Microsoft's codebase faster than the company's engineering teams can patch them, ra

Arjun S. Mehta ·   · 5m
AI

Japan's Auto Giants Bet on Quantum Computing for Design and Simulation

Nissan and Denso are among the first major manufacturers to integrate quantum algorithms into industrial workflows, signaling a shift from r

Kenji Watanabe ·   · 5m