The Research Assistant in the Room

Short summary

The author describes using an AI research assistant to adapt a 2010 academic paper's tree data model theory into Omnist, a tool that handles JSON, YAML, TOML, and XML interchangeably. Three key insights emerged: arrays are just repeated labels, schemas can use counting ranges instead of PSPACE-complete language containment, and value domains map to types. The R&D loop that once took days was compressed to an afternoon with AI assistance.

• •Adapted 2010 XML-centric tree model theory into a canonical model for JSON, YAML, TOML, and XML
• •Key insight: schema order constraints can be replaced with counting ranges, eliminating PSPACE-complete problems
• •AI research assistant compressed the hypothesis-test loop from days to an afternoon

Generated with AI, which can make mistakes.