The Information Machine

Import AI 464: Fables writes GPU kernels; AI automation; and analog computation

Import AI · Jack Clark · 2026-07-06

Jack Clark's Import AI newsletter reports that Fable AI set a record 18.71X GPU kernel speedup on KernelBench-Mega, that the Remote Labor Index shows AI online-task automation quadrupled in eight months, and that OSWorld 2.0 reveals frontier AI agents remain below 21% accuracy on long-horizon computer tasks.

Open original ↗

Appears in

Extraction

Topics: gpu-kernel-optimizationai-automationcomputer-use-agentsai-benchmarkslabor-economics

Claims

  • Fable achieved an 18.71X speedup on KernelBench-Mega using CUDA on an RTX PRO 6000 Blackwell, outperforming Claude Opus 4.8 at 14.4X and GPT 5.5 at 4.34X.
  • The Remote Labor Index shows AI success rate on online freelance tasks rose from 2.5% to 16.1% between October 2025 and July 2026, more than quadrupling in under eight months.
  • OSWorld 2.0 introduces 108 long-horizon computer tasks with a 1.6-hour median human completion time, and the strongest model (Claude Opus 4.8) achieves only 20.6% binary accuracy.
  • AI capability expansion on economically relevant tasks is outpacing humans' ability to adapt and maintain comparative advantages, and person-light AI-heavy organizations will displace un-augmented workers.
  • JD's Oxygen AI Item Center processes hundreds of millions of inventory updates daily using self-evolving LLMs and VLMs on Huawei Ascend NPUs, illustrating how AI is being woven into business back-office functions at country scale.

Key quotes

Being able to autonomously develop and improve kernels is one of the fundamental input tasks for being able to do AI research and development. The better AI systems at doing tasks like kernel design, the better they get at the kinds of tasks required for AI development, and that means the better they get at things that could lead to recursive self-improvement.
The frontier has more than quadrupled in under eight months, a concrete signal of how quickly economically capable AI agents are advancing.
AI systems are expanding their economically relevant capabilities faster than humans are expanding their comparative advantages relative to AI systems.