What actually shipped in agent engineering, pulled from the labs, arXiv and Hacker News.
See who we follow →NVIDIA Magpie TTS provides open-weight multilingual voice synthesis with low latency and full deployment control.
Researchers developed methods to reduce the computational cost of knowledge distillation for large-scale deployment.
Meta released Muse Glimmer, an open source multimodal model for local agentic applications.
Baseten is now available as a Hugging Face Inference Provider.
LFM2.5-2.6B enables deployment of local agents at scale.
OlmoEarth platform enables geospatial inference at planetary scale.
LFM2.5-Encoders provide fast long-context inference on CPU hardware.
Frontier Lab Agent experienced a security intrusion in July 2026; technical timeline and details of the incident are documented.
Nunchaku 4-bit diffusion inference is being integrated into the Hugging Face Diffusers library.
NVIDIA NeMo Automodel and Hugging Face Diffusers enable fine-tuning video and image models at scale.
NVIDIA Nemotron 3 Embed ranks first on RTEB benchmark for agentic retrieval tasks.
Hugging Face disclosed a security incident in July 2026.
Article about lessons learned building Shippy agent system.
Real World VoiceEQ measures human quality of voice AI systems.
Inkling by Thinking Machines is a new tool or service.
Hugging Face covers profiling attention mechanisms in PyTorch models.
Hugging Face released Data for Agents toolkit.
Hugging Face released native-speed vLLM transformers modeling backend.
SkyPilot enables running AI workloads on any cloud with zero-egress storage on Hugging Face.
LeRobot v0.6.0 released with imagine, evaluate, and improve features.
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.
