Autonomous systems that plan, act, and verify across tool calls and multi-step workflows, with active debate over architecture choices, coordination costs, memory design, state management, and the governance infrastructure needed to make them reliable.

AI safety spans containment of agentic systems, epistemic harms from sycophancy, skill atrophy from unreviewed code generation, and macro-level risks from rapid capability growth — each requiring different mitigations.

Benchmarks measure model or system capability, but their results are only as meaningful as their design — a recurring problem across LLM, multi-agent, and vision tasks, where tests built for one context are routinely applied to contexts they cannot capture.

LLM engineering spans the full stack of building with large language models: training, inference optimization, agent architecture, harness design, and the operational tradeoffs that determine whether model capability translates into reliable software.