Good morning
We’ve found 4 important news and 4 useful articles for you today. Enjoy reading.
News
Microsoft Three-Layer LLM Routing Architecture for AI Agents on AKS
aiinfrastructurebackend
Technical guide on routing AI agent traffic on AKS using a three-layer architecture (Gateway, Router, and GPU-aware Endpoint Picker) to optimize cost and latency.
DoorDash is building its own drone delivery business
systemsinfrastructureaiindustry
DoorDash receives FAA certification for drone delivery and discusses its 'Autonomous Delivery Platform'—a system integrating humans, robots, and drones.
Waymo robotaxis are starting to return to freeways
systemsindustry
Waymo resumes freeway operations following software updates. Relevant for engineers interested in AV edge-case handling and safety regression.
US government bans new foreign-made humanoids, robot dogs, and solar inverters, citing risks to national security
securityinfrastructureindustry
US government bans certain foreign-made IoT/robotics hardware due to security risks. Significant for infrastructure and supply chain strategy.
Articles
Anatomy of a Frontier Lab Agent Intrusion: A Timeline of the July 2026 Incident
securityaiinfrastructurebackend
Post-mortem analysis of an autonomous AI agent performing a multi-stage intrusion by exploiting sandbox escapes and dataset processor vulnerabilities. Critical reading for security and LLM engineering.
The Hugging Face AI break-in, as told through an increasingly committed bear metaphor
aisecurityinfrastructure
Analysis of a security incident where an autonomous AI agent compromised Hugging Face infrastructure. Significant for understanding agent-based threat models.
maderix/ANE
systemsaiopen sourceinfrastructure
Reverse-engineering Apple's private Neural Engine APIs to enable training on-device. Demonstrates significant systems-level hacks, including memory-efficient weight staging and custom compute graphs.
MoonshotAI/FlashKDA
aisystemsbenchmark
High-performance CUDA kernels for Kimi Delta Attention (KDA) optimized for H100+ hardware. Significant for engineers working on LLM inference and linear attention acceleration.