Einride Secures NHTSA Approval for Autonomous Truck Operations in Texas
Einride has received approval from the National Highway Traffic Safety Administration (NHTSA) to operate its autonomous trucks on public roads in Austin, Texas. This marks the fifth state where the company has received such approval, following Arizona, Colorado, South Carolina, and Tennessee. The company demonstrated its cab-less, SAE Level 4 fully autonomous electric truck in Austin on March 19th.
This approval is a significant step in Einride’s U.S. expansion and its plans for large-scale commercialization. The company is also in the process of a proposed business combination with Legato Merger Corp. III to become a publicly traded company in the first half of 2026.
Sources:
Google’s TurboQuant Slashes AI Memory Needs with Zero Accuracy Loss
Google Research has introduced TurboQuant, a new compression algorithm designed to reduce the memory requirements of large language models (LLMs) and vector search engines without losing accuracy. This training-free algorithm can quantize LLM key-value (KV) caches down to 3 bits, resulting in a memory reduction of at least six times.
On NVIDIA H100 GPUs, a 4-bit version of TurboQuant demonstrated up to an eight-fold performance increase in computing attention logits compared to unquantized keys. The technology addresses the significant memory bottleneck caused by KV caches, especially as context windows in LLMs grow larger. TurboQuant operates in a data-oblivious manner, meaning it doesn’t require dataset-specific calibration, which simplifies its integration into existing systems. The research will be presented at the upcoming ICLR 2026 conference.
Sources:
Atom Computing and Cisco Partner to Advance Distributed Quantum Computing
Atom Computing and Cisco have signed a Memorandum of Understanding (MOU) to explore the integration of neutral-atom quantum computers with quantum networks. The collaboration aims to create distributed quantum computing architectures by physically linking multiple quantum processors.
This partnership will leverage Cisco’s expertise in quantum networking hardware and software with Atom Computing’s neutral-atom quantum hardware. The goal is to accelerate the development of scalable and distributed quantum systems. Key areas of the collaboration will include extending Cisco’s network-aware distributed quantum computing compiler to support workloads on neutral-atom systems and investigating the necessary technologies to interface the two companies’ hardware. This initiative reflects a shared commitment to advancing the global quantum ecosystem towards utility-scale quantum computing.
Sources:
Microsoft Launches ‘AI Runway’ to Simplify AI Infrastructure on Kubernetes
At KubeCon + CloudNativeCon Europe 2026, Microsoft unveiled a new open-source project named AI Runway. This project provides a unified Kubernetes API for inference workloads, enabling platform teams to centrally manage model deployments. AI Runway is designed to simplify the process for users who may not have deep Kubernetes knowledge, featuring a web interface for deploying models.
The tool includes built-in discovery for HuggingFace models, indicators for GPU memory compatibility, and real-time cost estimations. It also supports various runtimes such as NVIDIA Dynamo, KubeRay, and KAITO. This initiative is part of Microsoft’s broader effort to bring the operational maturity of Kubernetes to AI workloads.
Solo.io Launches ‘agentevals’ Open-Source Project for AI Agent Reliability
At KubeCon + CloudNativeCon Europe, Solo.io announced the launch of a new open-source project called ‘agentevals’. This project is designed to help teams instrument, evaluate, and benchmark the behavior of agentic AI for quality and reliability across different models and frameworks.
‘agentevals’ addresses the challenge of evaluating non-deterministic AI agents by treating the agentic loop similarly to how observability platforms handle distributed systems, leveraging OpenTelemetry. The project offers both offline and online evaluation modes, allowing teams to analyze agents from recorded traces or live-streamed data. Alongside this, Solo.io has contributed its ‘agentregistry’ project to the Cloud Native Computing Foundation (CNCF) to address governance gaps in agentic infrastructure.
Sources: