MiniMax, LingBot, Claude Code shake Chinese AI

// 1h agoNEWS

MiniMax, LingBot, Claude Code shake Chinese AI

China's AI sector sees major activity with MiniMax planning a 2.7T model and Robbyant open-sourcing its LingBot-VLA 2.0 model trained on 60,000 hours of data. Meanwhile, China's MIIT issued a security advisory warning of telemetry vulnerabilities in older versions of Anthropic's Claude Code.

// ANALYSIS

China's AI ecosystem is moving at breakneck speed, balancing aggressive model scaling and open-source robotics with rising regulatory scrutiny on foreign developer tools.

–MiniMax's 2.7T model represents a massive jump in parameter scale, signaling that the race for ultra-large models remains highly competitive.
–Ant Group's LingBot-VLA 2.0 release on 60k+ hours of data shows a significant push to democratize embodied AI and robotic foundation models.
–The MIIT security advisory on Claude Code highlights the geopolitical and regulatory risks associated with developer-focused AI tools and telemetry practices.

// TAGS

minimaxlingbot-vlaclaude-codeant-groupmiitsecurityembodied-aillm

DISCOVERED

1h ago

2026-07-08

PUBLISHED

1h ago

2026-07-08

RELEVANCE

8/ 10

AUTHOR

tphuang

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

NEWS16m ago

Pieter Levels runs Claude Code on production

Indie maker Pieter Levels (@levelsio) demonstrated using Anthropic's command-line agent Claude Code directly on a production server to instantly build backend API routes. These endpoints are immediately consumed by a native Swift iOS app, showcasing a seamless server-to-mobile development loop.

RESEARCH26m ago

GaP boosts robot policy reliability

Graph-as-Policy (GaP) is a robotics framework designed to bridge the reliability gap in variational automation tasks by representing robot policies as directed computation graphs of modular skills. Developed by researchers from UC Berkeley, NVIDIA, CMU, and Bosch, the framework uses an LLM-in-the-loop multi-agent system to construct, test, and refine these graphs in simulation before deployment.

VIDEO26m ago

HASE co-evolves model weights, harness

Harness-Aware Self-Evolving (HASE) is an agentic reinforcement learning framework that allows a single model to co-evolve its policy weights, task solutions, and environment harness in a unified multi-turn action space. By enabling the model to dynamically modify harness components like prompts, memory formatting, and validators, HASE allows smaller models like Qwen3-8B to match or beat the performance of models as large as 120B parameters in domains like text classification and alpha factor mining.