Z.ai says an agent running on its GLM-5.3 model built the inference service that now serves GLM-5.3-Flash, reaching 3.22x baseline throughput in thirteen days on a cluster of more than 100,000 Chinese…
DeepSeek AI released DeepSeek-V4.1-Flash, a multimodal Mixture-of-Experts model with 552B backbone parameters, 196B additional Engram parameters, and a 1M-token context window, cutting global KV cache…
DeepSeek AI's open-source coding agent DeepSeek Harness has surpassed 206,000 GitHub stars within three weeks of its August 13, 2026 release, becoming the most-starred coding agent on the platform and…
DeepSeek AI released DeepSeek Harness, an open-source agent runtime under the MIT license, built on a plugin framework called Cordis. The project emphasizes a model-agnostic design where every compone…
DeepSeek AI released DeepSeek-Harness (dsh), an open-source agent harness in developer preview version 0.1.0-rc.5, where every feature—tools, commands, skills, and MCP connections—is implemented as a …
LongHorizon-Harness, an open-source loop engineering system for AI agents, reached #1 on the Hugging Face Daily Papers weekly ranking for 2026-W32 and released v0.1.6 on 2026-08-15, adding OpenCode CL…
Tianyi from DeepSeek AI's Harness team publicly asked GitHub why the trending page excludes the deepseek-ai/deepseek-harness repository despite rapid community engagement, noting a colleague's recolle…
DeepSeek AI released DeepSeek-V4-Pro-0813, the official non-preview version of its open-weight coding and agentic model, reporting gains over the preview on every listed benchmark. On Terminal Bench 2…
DeepSeek AI and Peking University published a paper formalizing spatiotemporal composability for agent architectures, introducing the Cordis micro-kernel that powers DeepSeek Harness (dsh). The Cordis…
DeepSeek AI has released DeepSeek Harness (dsh), an open-source agent harness currently in developer preview, built on a plugin-based architecture powered by Cordis. Developers can install it via npx …
China's military AI push prioritizes unmanned technology, signaling a transition from research to procurement and fielding, according to an analysis of the country's defense strategy. The focus on unm…
DeepSeek AI released DeepSeek-V3.2, a 642.2 GB text-generation model under the MIT license on Hugging Face, which has garnered over 2.5 million downloads. The model requires large hardware such as mul…
A comparative analysis of scale-out RoCE network traffic patterns in training large language models reveals that hyperscale deployments like xAI's Grok 4.0 generate network loads orders of magnitude h…
A developer built a weekend project called "Achu" (meaning "print" in Tamil), a coding tool that iterated through multiple AI models. The project started with Google Flash 3.5, consuming 1,000 credits…
By May 2026, China’s AI ecosystem has shifted from a "model war" to an "agentic revolution" focused on real-world execution, cost efficiency, and full hardware independence from Western supply chains.…