Fable writes first GPU megakernel; AI online work automation quadruples in 8 months

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

Fable submitted the first genuine GPU megakernel on KernelBench-Mega, achieving an 18.71x speedup over an optimized PyTorch baseline with a single cooperative kernel launch per decoded token. Claude Opus 4.8 reached 14.4x and GPT-5.5 only 4.34x. This benchmark measures AI systems writing their own low-level kernels, a signal for recursive self-improvement. Separately, the Remote Labor Index shows AI end-to-end success on online freelance projects rose from 2.5% in October 2025 to 16.1% in July 2026, with Fable 5 hitting 16.1%. Tasks span 3D modeling, animated ads, and architectural renders, with a median human completion time of ~1.6 hours. The post does not disclose specific model scores on OSWORLD 2.0, only noting poor performance so far.

Why it matters: Fable submitted the first genuine megakernel to KernelBench-Mega, hitting 18.71x speedup with a single cooperative kernel launch — cleaner than Claude Opus 4.8 and GPT-5.5 entries. It's an early signal of AI improving its own low-level kernels, directly relevant to people doin...