AI Code Debt, Disaggregated Inference, GPU‑Ops Teammates, and Robustness Insights

AI Code Debt, Disaggregated Inference, GPU‑Ops Teammates, and Robustness Insights AWS announced a new inference service that separates model compute from storage, enabling more flexible scaling of large language models. A YC startup launches an AI agent that manages GPU clusters, while researchers unpack the hidden costs of AI‑generated code and propose fresh theories on predictive robustness. These announcements reflect a broader push toward modular AI services, smarter automation, and stronger theoretical foundations. Introducing Disaggregated Inference on AWS powered by llm-d Amazon Web Services (AWS) What happened: AWS announced a new inference service that separates model compute from storage, enabling more flexible scaling of large language models. Why it matters: Developers can reduce latency and cost when running large models by provisioning compute and storage independently, making it easier to deploy LLMs at scale. Context: The approach could simplify deployment of LLMs acros