AI BriefingDaily AI briefing
Editor's picks come first, followed by stories repeatedly covered across days and sources.
Top stories
05- 01Hugging Face integrates TorchTPU for native PyTorch support on TPUs
Transformers, Diffusers, and Kernels now run natively on TPUs using the TorchTPU backend, eliminating the need for torch_xla or JAX.
HuggingFace Blog
- 02Mistral AI releases Voxtral 3 voice agent SDK and speech models with sub-1-second turn times
The release includes a LiveKit-based Voice Agent SDK and speech models priced from $0.0035 per minute for transcription and $13 per 1M characters for TTS.
Mistral AI
- 03Claude designs 17 protein binders against scorpion neurotoxins in 64 hours
Working autonomously for 64 hours, Claude produced 17 experimentally validated binders against four scorpion neurotoxins, with all 211 designs released as open data.
Anthropic Research
- 04Cloudflare and VoidZero Launch tsrs, a Rust Port of TypeScript Compiler
The experimental release is available on npm and incubated within the Oxc JavaScript compiler toolchain.
Cloudflare Blog
- 05DeepSeek open-sources DeepInfer inference framework for V4.1-Flash
The release supports NVIDIA SM100 and Huawei Ascend 950 hardware, featuring optimized kernels and asynchronous scheduling for offline inference.
Hacker News
Rising repositories and models
03- deepseek-ai/DeepInfer
High-performance inference framework dedicated to DeepSeek V4.1-Flash
- lodestones/MOSS-Image-SFT-8-steps-dmd
From 50 to 8 Steps: CFG-Free MOSS-Image Inference
- dealignai/DeepSeek-V4.1-Flash-1017-UNCENSORED-FP8
DeepSeek-V4.1-Flash FP8 with Refusal Responses Removed