Tag
This article discusses how retries in AI systems, particularly with LLMs and agents, can exacerbate failures when underlying issues are not addressed, leading to repeated mistakes with increased cost and latency.
The blog post analyzes a recent GitHub outage caused by a misconfigured autoscaling policy, discussing the challenges of autoscaling in cloud infrastructure and referencing the component substitution fallacy to highlight systemic issues.
Discusses how cascading recovery and feedback loops can cause metastable failures in distributed systems, preventing proper recovery.
Explains why unbounded queues are a bug in software systems, using Little's Law and the bathtub analogy to show that queues only absorb variance, not sustained load. Discusses latency death spirals and advocates for backpressure instead.
The article discusses common failure patterns in agentic AI systems, specifically 'dumb AI loops,' citing issues like state poisoning and data leaks observed in Claude Code deployments.