Huawei

AI Advances: Fable GPU Kernel, Automation, and OSWORLD 2.0

Import AI reports several AI-research developments: Fable wrote a high-performance GPU 'megakernel' that achieved an 18.71x speedup on an RTX PRO 6000 Blackwell versus an optimized PyTorch baseline in KernelBench‑Mega; the Remote Labor Index (RLI) shows frontier models' success at end-to-end online freelance tasks rising from 2.5% (Oct 2025) to 16.1% (July 2026), with Fable 5 scoring 16.1%; OSWORLD 2.0, a long‑horizon benchmark created by multiple universities and organizations, evaluates agents on 108 multi-hour, multi-program computer-use tasks and finds current agents far from reliable (best settings ≈20.6% binary accuracy); and JD published the Oxygen AI Item Center, an industrial-scale LLM/VLM-centric inventory system running at massive scale on Huawei Ascend NPUs. The newsletter frames these items as signals that AI is improving at core R&D, complex computer use, and real-world operational automation, with implications for economic automation and large-scale enterprise systems.

Macquarie: Buy Chinese AI Chip Stocks; Favours Cambricon

Macquarie’s China Information Technology analysts said in a late-June report that now is the best time to invest in China’s AI chip makers, citing the rise of domestic LLM players, a growing token economy and PRC government support that limits imports of advanced Nvidia GPUs. The bank initiated coverage on five Chinese AI-chip companies and rated Shanghai-listed Cambricon its top pick (outperform) with a 2,060 yuan target. Among Hong Kong-listed names it prefers Biren Tech with a 140 HKD target. Other picks include Iluvatar CoreX and MetaX; Macquarie rated Shanghai-listed Hygon underperform over market-share concerns. The report cites IDC data showing Huawei led AI chip shipments, with Cambricon second and Hygon third.

Frontier Models, Z.ai's ZCode, and Deployment Arms Race

The newsletter reviews recent shifts in frontier AI: Anthropic’s Claude Fable 5 was pulled after a jailbreak and redeployed July 1 with a classifier fallback; Z.ai published ZCode, an agentic development environment packaged around GLM-5.2 with MIT-licensed weights and very large context; and Anthropic introduced Claude Science, a reproducibility-focused workbench for scientific workflows. The piece highlights a broader industry pivot: major cloud and AI vendors (Microsoft, AWS, OpenAI, Anthropic) are investing heavily in forward-deployed engineering and runtime integration — Microsoft announced a $2.5B, 6,000-person Microsoft Frontier Company and AWS created a $1B Forward Deployed Engineering org — arguing that deployment and integration, not model capability, are the new bottlenecks. The write-up also surveys related lab papers and funding/motion news across the AI stack.