Why it matters
Tim Sweeney demonstrates ARIA launching GPU experiments, examining existing runs and producing visual reports inside Weights & Biases. Long-running training executes outside the conversation loop while the agent checks progress. He also describes turning reviewed traces into nightly evaluations. The live batch completes a research iteration without beating the earlier best result, illustrating that automated execution does not guarantee improvement.
My takeaway: Separate job execution from conversation handling, preserve experiment provenance and score correctness independently from insight quality and tool use. Use reviewed production traces to build repeatable release checks.