Snorkel AI

Experts doubt Fable distillation explains Kimi K3

U.S. officials have alleged that Chinese company Moonshot copied Anthropic’s Fable LLM and used advanced chips banned for export to China to build Kimi K3, but independent AI researchers say simple distillation of Fable is unlikely to explain Kimi K3’s rapid, high-end capabilities. Experts argue that supervised fine-tuning (SFT) alone would not produce such performance in weeks and that reinforcement learning and substantial compute infrastructure are more plausible contributors. Anthropic previously accused Moonshot, DeepSeek and MiniMax of distillation activity, while U.S. policymakers have discussed export controls and know‑your‑customer rules for data centers. The reporting notes ongoing uncertainty about alleged “watermarks” and the provenance of hardware claims; several organizations either did not respond or have publicly raised concerns about illicit chip transfers.

Debate Over Banning Chinese Open-Weight LLMs

TechCrunch reports a heated debate after Chinese lab Moonshot released Kimi K3, described as the largest open-weight large language model. OpenAI strategist Dean W. Ball suggested the U.S. government could create regulatory pressure around such models to protect frontier labs’ economic returns, a position he later retracted. Axios reported the Trump administration is considering banning K3 and other advanced Chinese models, though Politico said the Department of Commerce is unlikely to act immediately. Supporters of open models — including researchers and executives from Snorkel AI and Hugging Face — argue that open-weight models lower costs, accelerate innovation, and broaden participation, while some security and geopolitical voices suggest focusing on chip export controls (e.g., restricting Nvidia H200 sales to China) instead of model bans. The article frames tensions between open-source proliferation and protection of U.S. frontier AI firms.

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.