Turning a real-world archive into a verifiable AI index.
ChatGPT Work with GPT-6 Astra was tasked with structuring a real-world archive of 1,225 files across 97 folders, protecting existing numbering, assigning consecutive numbers to previously unnumbered records, identifying duplicates, and building a verifiable central index. The long-running task became a real-world test of autonomous work, persistent checkpoints, interruption recovery, and state transparency.
Results:
1,225 files across 97 folders were processed. 628 planned renamings were completed while existing numbering was protected. Previously unnumbered records received consecutive numbers, duplicates were identified without deletion, and a verifiable central index was created. A persistent checkpoint became essential for reliable recovery after interruptions.