October 20263

China's AI Benchmark Crisis: When Near-Perfect Scores Stop Meaning AnythingHOT

As large language models saturate traditional benchmarks, Chinese industry observers say buyers are shifting from leaderboard rankings to cost-per-task calculations. A new unified model from Tsinghua and Zhipu AI shows why: strong image generation scores paired with weak text performance.

🤖 AIOct 28, 2026·3 min read#Ai Models#Tsinghua

Nuclear Clocks Are Bringing Atomic Time Inside the Atom

Two teams, one in China and one in Europe, have independently made thorium-229 nuclear clocks tick in a stable loop — the first time a working time standard has been built on the atomic nucleus rather than its electrons.

🔬 ScienceOct 21, 2026·3 min read#Research#Astronomy

Tsinghua's Open-Source Agent OS Cuts AI Task Costs by 83%

Tsinghua's THUNLP lab, with ModelBest and OpenBMB, released PilotDeck 2.0, an open-source operating system for AI agents. The system natively supports MCP and gives agents searchable, editable memory.

🤖 AIOct 19, 2026·2 min read#Research#Innovation

You've reached the end