Terence Tao uses AI agents to revive 1999 Java applets and build new math tools
Original: Old and new apps, via modern coding agents by Terry Tao
Why This Matters
Demonstrates how AI coding agents can lower barriers for expert-driven scientific software development and visualization.
Fields Medal-winning mathematician Terence Tao used AI coding agents to port ~24 Java 1.0 applets to JavaScript in hours, reviving 27-year-old math visualizations. He also completed an abandoned special relativity visualization project and built a new Gilbreath conjecture tool, all via 'vibe coding.'
Mathematician Terence Tao described on his blog how he used modern AI coding agents to migrate approximately two dozen Java 1.0 applets — originally written in 1999 for courses in complex analysis and linear algebra — into JavaScript. The applets, which had become non-functional as browser support for that version of Java ended, were ported in a matter of hours. The agent also applied graphical upgrades, such as colorizing a Besicovitch set applet that was originally monochrome. Tao noted that despite known risks of LLM-generated bugs, only one minor drag-event bug was found in the ported code, while the agent independently identified two pre-existing bugs in the originals. Encouraged by the results, Tao used the same AI-assisted 'vibe coding' approach to finish a special relativity visualization tool he had abandoned in 1999 due to code complexity. He described the concept as 'Inkscape, but in Minkowski space.' Separately, after writing a blog post about a Gilbreath conjecture paper, he commissioned the agent to build an accompanying interactive visualization — completed in a few hours. Tao published transcripts of both coding sessions and indicated he plans to add similar AI-assisted interactive supplements to future mathematical papers.