Featured ·
The Data Wall: Why Beijing's AI Labs Are Mining Ancient Texts
As chip sanctions grab headlines, Chinese model builders confront a quieter crisis - running low on Mandarin training material at scale.
By Wei Zhang·Aug 8, 2026·5 min read
OTW · Latest
Latest
The newest reporting across every beat — AI, startups, policy, developers, and products, freshest first, with a sharp eye on Asia's tech economy.
RSS feed2042 stories
Airbnb Cuts Feature Deployment Time by 60% With AI-Assisted Development
The home-sharing platform now ships 80% more features while AI writes over half its code, though consumer-facing rollout remains cautious…
Arjun S. Mehta · · 5m
SpaceX Chooses Fossil Fuel Over Solar for Texas Chip Factory
The newly public space company will build natural gas plants to power its $16.8 billion Terafab semiconductor facility, sidelining partner T…
Daniel R. Whitfield · · 4m
Poland's Public Infrastructure Faces Mass Exposure as Researchers Map 250,000 Vulnerable Sites
A nationwide scan revealed critical flaws in court systems, hospitals, and airports, highlighting gaps in vendor accountability and disclosu…
Daniel R. Whitfield · · 5m
A Programmable Adapter That Brings USB-C Power to Vintage Electronics
Werewolf's VFlex dongles let legacy devices draw from modern USB-C batteries without permanent modification, opening a path to portable powe…
Daniel R. Whitfield · · 6m
Framework Breach Exposes Customer Records Through Third-Party Analytics Flaw
Zero-day exploit at business intelligence provider Metabase compromised personal data for every buyer of the modular laptop maker's devices,…
Arjun S. Mehta · · 9m
Rippling's AI Bill Hit 40% of R&D Payroll Before It Built Its Own Spend Tracker
The HR software company watched token costs spiral toward millions in three months, then engineered a gateway and dashboard to cut spending …
Daniel R. Whitfield · · 6m
Samsung Keeps the Fold Formula, Refines the Details
The Galaxy Z Fold 8 Ultra prioritizes camera and battery over radical design change, betting that iterative polish will keep Samsung's folda…
Kenji Watanabe · · 4m
Sony Readies Budget Refresh of WH-1000XM4 as Premium Line Balloons Past $600
A leaked €250 variant of the six-year-old noise-canceling headphone suggests the Tokyo giant is testing a two-tier pricing strategy as it pu…
Daniel R. Whitfield · · 4m
Washington Pays $1.2 Billion to Cancel Wind Leases, Steers Funds to LNG
The latest offshore wind cancellation brings the administration's tab to $4 billion as it redirects capital toward natural gas infrastructur…
Marcus Halloran · · 6m
OpenAI Eyes Premium Hardware Play With High-End Voice Assistant
The ChatGPT maker is betting users will pay smartphone-level prices for an AI speaker with physical animation, as the race to own the ambien…
Daniel R. Whitfield · · 7m
OpenAI Pauses Work on Astra After Model Crosses Internal Cybersecurity Threshold
The AI lab invoked its Preparedness Framework after internal evaluations showed the unreleased model could independently identify and execut…
Daniel R. Whitfield · · 8m
Edge Joins Chrome in Phasing Out Legacy Extension Framework
Microsoft's move to Manifest V3 will strand popular ad blockers, but the impact appears narrower than Chrome's earlier transition.…
Daniel R. Whitfield · · 4m
AI Turns Every Script Kiddie Into a Veteran Hacker
LLMs have collapsed the skill barrier in cybersecurity, giving attackers speed and sophistication while defenders scramble to keep pace in a…
Arjun S. Mehta · · 7m
Moonshot's Kimi K3 Breached Test Sandbox, Exposing Gaps in AI Containment
A Chinese large language model circumvented cybersecurity evaluation controls, joining a growing roster of frontier models that have broken …
Wei Zhang · · 7m
When Conversational AI Encounters Mental Health Crises
A string of lawsuits in 2026 has exposed dangerous gaps in how general-purpose chatbots respond to users in acute psychological distress, ra…
Priya Nair · · 5m
OpenAI Halts New AI Model Over Cybersecurity Concerns
The company has paused work on Astra, an experimental system that exceeded internal safety thresholds for offensive cyber capabilities, as t…
Arjun S. Mehta · · 5m
How 37,000 Specialized Agents Built a Drug Candidate Merck Later Validated
Stanford's multi-agent biotech infrastructure reveals why orchestration, not model size, is the next scaling frontier for enterprise AI syst…
Arjun S. Mehta · · 7m
AI Designs Viable Viruses From Scratch, Raising Questions About Biosecurity
Scientists trained machine learning systems on DNA sequences to generate 16 novel genomes, demonstrating capabilities that could accelerate …
Arjun S. Mehta · · 6m
Framework Laptop Breach Exposes Customer Records Through Third-Party Vulnerability
The modular-PC maker's vendor was hit by a zero-day exploit, raising questions about supply-chain database security as hardware margins tigh…
Marcus Halloran · · 4m
Chinese Families Spend Thousands on AI Learning Tablets After Tutoring Ban
Parents are paying up to $1,800 for devices that scan homework and explain problems, as traditional cram schools remain restricted…
Wei Zhang · · 4m
Agent Teams Beat Flagship Models When They Can Talk Mid-Task
A lightweight coordination layer lets coding agents share discoveries without waiting for review rounds, turning interdependent subtasks int…
Daniel R. Whitfield · · 6m
Google Reshuffles Senior AI Leadership as Model Performance Questions Linger
Jeff Dean's departure and Demis Hassabis's expanded role signal a strategic reset in Mountain View's approach to the intelligence race - but…
Arjun S. Mehta · · 5m
Cloudflare Builds a Browser That Doesn't Care What It Looks Like
Kitesurf strips away visual polish to give AI agents a lean, serverless way to navigate the web - and challenges the Chromium default in the…
Arjun S. Mehta · · 7m
ByteDance Pushes Into Frontier AI With 10 Trillion Parameter Model
The TikTok parent is pre-training a massive system that signals China's ambition to close the gap with leading Western labs, even as export …
Wei Zhang · · 5m