Source note
Old and new apps, via modern coding agents
Summary
Modern coding agents helped revive roughly two dozen obsolete Java mathematics applets and create new interactive visualization tools in hours. The experience suggests that guided AI coding can make noncritical educational and research visualizations practical to maintain and produce.
Problem
- Java applets from 1999 became unusable when web standards stopped supporting that version of Java, leaving useful mathematics visualizations inaccessible.
- Hand-coding new visualization software was time-consuming and had previously caused projects, such as a special-relativity editor, to be abandoned.
Approach
- The author used a modern AI coding agent to port old Java applets to JavaScript and update their graphics.
- He guided the agent through a new special-relativity visualization modeled on an Inkscape-like editor for Minkowski space.
- He used the same workflow to build a visualization for the Gilbreath conjecture and plans to attach similar tools to future papers.
- The author playtested the applications and reviewed the generated code, accepting residual bugs because the tools are supplemental rather than mission-critical.
Results
- About two dozen old applets became functional again within a few hours, including the honeycomb and Besicovitch set visualizations.
- The migrated apps had one minor discovered bug involving drag events outside a complex-analysis applet's main box.
- The agent identified two bugs in the original code that the author had not known about, producing roughly equal net code quality by the author's assessment.
- A previously abandoned special-relativity visualization was produced after a few hours of AI-assisted coding, and a Gilbreath conjecture visualization was completed after another few hours.
- The excerpt reports no formal benchmark, user study, or quantitative quality metric; the strongest evidence is the author's hands-on testing and successful deployment of the apps.