JAX SPEEDS UP SNAKE, DROPS ROCM WHEEL LOOKUP
JAX's planar_snake is now up to 10% faster for power-of-two tiles, and the team is killing a painfully slow ROCm wheel discovery path.
read --wire →
$ tail -f topics/ai.log
Daily updates from the AI and machine-learning ecosystem - PyTorch, Hugging Face, the Anthropic and OpenAI SDKs, local LLMs, and the agent frameworks.
256 updates across 10 projects this week.
One calm review of what shipped across AI / ML - the commits, releases, and security advisories that matter. Every Monday, with security advisories same-day. Free, unsubscribe in one click.
We'll start you on the top AI / ML projects - refine anytime. · Read a sample issue →
JAX's planar_snake is now up to 10% faster for power-of-two tiles, and the team is killing a painfully slow ROCm wheel discovery path.
read --wire →
Meta
React's server compiler just killed a silent payload bloat that's been fattening every client reference in the framework.
read --wire →
OpenAI
OpenAI rolled back the Plan mode prompt in Codex, quietly removing the 'Create a plan?' nudge from the composer and the friction it brought to drafting.
read --wire →
Anthropic
Anthropic's plugin marketplace just grew by one, and for ecommerce teams watching checkout failures, that's the story to track today.
read --wire →
AI Agents
Cline's new hub drain and upgrade flow means you can restart a hub mid-session and lose nothing, thanks to durable event replay.
read --wire →
Local LLMs
The biggest news overnight: SGLang is patching a production-breaking Pixtral bug, and vLLM engineers found a way to shave up to 25% off time-to-first-token on Mamba models.
read --wire →
Hugging Face
Harbor-hf shipped five fixes overnight that close the gaps where stuck Sandboxes, terminal jobs, and invalid campaign results could quietly block recovery.
read --wire →
PyTorch
PyTorch's attempt to shed its Caffe2 legacy just got blocked by internal dependencies, and the resulting reverts ripple across the cleanup effort.
read --wire →
Google's JAX team is renaming a core Mosaic GPU attribute, and it will break your cross-process parameter code.
read --wire →
Meta
React's server rendering stack just got a memory fix that matters for long-lived signals.
read --wire →
OpenAI
Codex now requires content classification for all thread sources and extension fragments, while the Node SDK introduces caller-attested X.509 transport for hardened auth flows.
read --wire →
Anthropic
A coordinated wave of releases pushed Claude Code to 2.1.241 across the entire agent SDK stack, and you need to know where it matters.
read --wire →
AI Agents
Cline just shipped rollover-grade hub commands while Langchain exposes gateway internals, and OpenHands quietly broke its own monitoring contract.
read --wire →
Local LLMs
The biggest release of the week just landed: SGLang v0.5.18 adds new models, fixes a silent disagg bug, and defaults DeepSeek V4 FP4 checkpoints to the right MoE runner.
read --wire →
Hugging Face
LeRobot's FastWAM policy gets an opt-in compiled path that slashes Python and CUDA kernel-launch overhead on cached action inference.
read --wire →
PyTorch
Pallas stagers just gained the ability to align HBM windows on any tensor axis, not just the last one, and a critical autotuning crash is fixed.
read --wire →
If you thought Google's Python ecosystem was quiet overnight, you missed a flurry of fixes that touch everything from Bigtable filters to Spanner's DBAPI driver.
read --wire →
Meta
Overnight, Jest's resolver got 3x faster on warm paths and React Native fixed a pair of Android ViewManager crashers.
read --wire →
OpenAI
OpenAI shipped a wave of SDK refactors overnight, moving webhook signature checks and file polling into SDK-owned helpers while Codex expands browser and computer-use policy controls.
read --wire →
Anthropic
Anthropic's protobuf toolkit just eliminated thousands of redundant panic locations and dropped AnalyticsEvent size by 63 percent.
read --wire →