Tag
A discussion sharing career advice for junior engineers, emphasizing systems thinking, measuring outcomes, and building evals when working with AI agents.
This post asks engineers to share their experiences with unexpected cost spikes when running AI models in production and offers advice on optimizing costs and setting up guardrails to avoid budget overruns.
An Anthropic lead engineer leaked a 12-page document detailing five common failure modes in agentic loops and was fired shortly after. The thread summarizes the key failure types including blind, tangled, nodding, amnesiac, and manual loops.