In May 2025, delivered targeted API improvements for the Banner component in Khan/wonder-blocks and fixed a release tooling issue, focusing on immutability, type safety, and release correctness. The work enhances stability and downstream adoption while reducing risk in future changes.

In November 2024, Khan/perseus delivered a substantial update across widget prompt utilities, refactoring, and test quality, delivering tangible business value through broader widget support, improved reliability, and stronger typing. Key features delivered span 10+ widgets with standardized prompt utilities, enabling data-driven prompts and consistent UX across Video, Sorter, Python program, Plotter, PhET simulation, passage/passage-ref, Orderer, NumberLine, Measurer, and Matrix widgets. Interactive prompt JSON utilities were introduced for Graphs and Group widgets, with additional support for graded widgets and a getPromptJSON utility for GradedGroupSet, expanding the platform’s ability to render and test prompts from JSON definitions. API and typing enhancements improved developer productivity and reliability: manual type definitions for prompt utilities, consolidated typings, and the export of WidgetPromptJSON type; notable API change for getUserInputFromProps accompanied by targeted tests. Quality and maintenance improvements included a directory refactor with updated test names, adoption of React.createRef for renderer refs, and fixes for type errors and duplication (TeX onLoad relocation reversion and duplicate definitions). Testing and reliability were strengthened via tests updated to assert literal values, expanded test coverage for new utilities, and new tests around getUserInputFromProps. Overall impact: accelerated widget onboarding and feature parity across the widget family, reduced maintenance overhead through better typings and clearer structure, and improved reliability and test coverage for future changes.

October 2024 achieved a major milestone in Perseus by implementing a unified widget state serialization layer via PromptJSON, enabling reliable save/load, AI integration, and centralized state management across all widgets. The initiative included broad API coverage, extensive testing, and type safety improvements, with targeted fixes to unsupported widget handling and input validation.