Claude Code settings files expose 31,000-plus secrets on public GitHub
Two security firms have independently catalogued thousands of credentials — database passwords, API tokens, and root logins — exposed in Claude Code settings files committed to public GitHub repositories.
Berkeley's self-reorganizing agent teams beat production GPU kernels at 1.21× FlashInfer speed
A multi-agent coding system from UC Berkeley that restructures its own team mid-task produced GPU kernels faster than expert-written production code, achieving a geometric mean speedup of 1.21× over FlashInfer across tested workloads.
ExecuTorch WebGPU backend runs PyTorch models on the GPU inside the browser
PyTorch models can now run on the GPU inside any WebGPU-enabled browser through a new ExecuTorch backend, eliminating the need to convert models to formats such as GGUF, ONNX, or TFLite.
Airbnb open-sources Skipper, its embedded durable-workflow engine for JVM services
Airbnb released Skipper, the embedded workflow engine it runs in production across multiple internal teams, as an open-source project under the Apache 2.0 license on November 10.
ALL BRIEFS
RANKED · 6 THIS DAYSGLang Megakernel Serves 816B Muse Spark Open on Four TPU v7 Chips
A PyTorch blog post details serving Meta's 816B-parameter Muse Spark Open on a single four-chip TPU v7 host using SGLang's megakernel, halving decode time per token.
Meta Deploys TLX GPU Compiler Across Production ML Infrastructure
Meta disclosed Triton Low-Level Extensions (TLX), an extension to Triton that adds explicit warp-group and cluster-level orchestration for GPU optimization while preserving Triton's programming model.