Agentic Artifact Creation: Systems, Evaluation, Principles, and Opportunities
The survey defines agentic artifact creation as stateful AI work that builds or revises a deliverable, then uses observations to steer the next steps.
The authors reviewed 259 works available through August 20, 2026, including 230 systems and 29 benchmarks. They frame the process around an artifact representation, a construction policy, and runtime verification that can trigger repair. The paper compares six artifact families and separates application settings from evaluation practice. It warns that decomposition can add coordination costs, and that learned judges may not add much evidence when they share a generator’s blind spots. HF Daily Papers' note
The authors reviewed 259 works available through August 20, 2026, including 230 systems and 29 benchmarks. They frame the process around an artifact representation, a construction policy, and runtime verification that can trigger repair. The paper compares six artifact families and separates application settings from evaluation practice. It warns that decomposition can add coordination costs, and that learned judges may not add much evidence when they share a generator’s blind spots. HF Daily Papers' note
score 4