65 related articles

China's Ministry of Commerce is weighing restrictions on overseas access to top open-weight AI models from Alibaba, ByteDance, and DeepSeek. Meanwhile, DeepSeek is quietly building its own inference chip and raising ~$7B in first external funding.

From DeepSeek to Kimi K3 and Qwen 3, Chinese open source AI models are closing in on OpenAI and Anthropic at stunning speed. A deep dive into narrowing gaps, IPO valuation risks, the "open source decelerationism" debate, and why Google may be the biggest winner.

Qwen launches a massive open-weight model, DeepSeek V4 priced at $0.0028/M tokens, Kimi K3 hits compute limits — Chinese AI labs are aggressively disrupting OpenAI and Anthropic's frontier model dominance.

Chinese open-source models rapidly close the capability gap with top closed-source AI. DeepSeek shocks the industry while Qwen matches global benchmarks.

A hands-on guide to LLM fine-tuning: from understanding model weights to local Qwen3 deployment, dataset preparation, and domain-specific training. Build a complete AI engineering skill set.

Fields Medalist Terence Tao on AI and mathematics: within a decade, AI will handle much of mathematicians' routine work — but that work was never the core of the discipline.
Three Core Gaps in Multimodal LLMs: Fr…
Microsoft Research India reveals three core gaps in multimodal LLMs: visual perception blindspots, cognitive hallucination, and architectural limitations. Explores Faithful GRPO, behavior modeling, and model alignment breakthroughs.
4 Alternatives for Running CUDA on Non…
A deep dive into running CUDA on non-NVIDIA hardware (AMD, Intel): comparing ROCm/HIP, ZLUDA, SYCL/oneAPI, and OpenCL across principles, use cases, and limitations.

Diana Deutsch's Tritone Paradox proves that identical sounds can be heard in opposite ways. Explore the acoustics of Shepard Tones, why perception varies by culture, and what this means for AI.
Causal Theory Cracks Open the LLM Blac…
How can we solve the LLM black box problem? This article explores how causal theory powers mechanistic interpretability research — from causal intervention and activation patching to circuit discovery — and its implications for AI safety and alignment.

Learn how Skill-powered AI Agents can automatically log in to WeChat Official Account backends, scrape data, sort it, and generate Excel reports — eliminating manual analytics work.

DeepSeek is entering AI chip development, targeting compute autonomy. This article analyzes its motivations, software-hardware synergy, chip R&D challenges, and impact on China's AI vertical integration.

Tencent Hunyuan HY3 official version is open-sourced under Apache 2.0, priced as low as 1 yuan per million input tokens, with major gains in agents, reasoning, coding, and long context. On the same day, Meituan open-sourced its trillion-parameter LongCat 2.0.

Zhipu AI's GLM open-source model approaches top closed-source models at 15% of GPT-5.5's cost. HBM shortages fuel AI inflation, and space data centers emerge as a serious compute play.

A deep dive into the Vibe Coding four-module framework: paradigm cognition, open-source customization, SDD, and project rules for real engineering delivery.

A Texas A&M philosophy professor resigned his tenured position after being told he couldn't teach Plato's Symposium, exposing a deepening academic freedom crisis at U.S. public universities.

Deep dive into Zhipu's GLM-5.2: truly usable 1M-token context, MIT open-source strategy, full-stack Huawei Ascend training, and how it compares to Claude Opus. Includes benchmarks, use cases & pricing.

Huawei HDC unveils Pangu 2.0 full open source and HarmonyOS 7 system-level Agent capabilities. Deep analysis of sparse architecture efficiency, on-device 30B models, and the Agent gateway battle.

Xiaomi releases open-source MIMO Code while Huawei enters the Agent era with Pangu. Compare their AI strategies: Xiaomi's Android-like open ecosystem vs. Huawei's iOS-like vertical integration.

Deep dive into Huawei Kunpeng's AI Agent solution — why general-purpose computing matters as much as AI computing for agent architecture, from memory systems to tool calling and secure runtimes.