Terry Tao on Modern Coding Agents for Mathematical Visualization

AI Coding Agents Enable Domain Experts to Build Software

Modern coding agents significantly lower the barrier to entry for domain experts—such as mathematicians—to create functional software without needing deep professional software engineering expertise. By reducing the "activation energy" required to start and finish projects, these tools allow experts to translate complex theoretical concepts into interactive visualizations and tools quickly.

Modernizing Legacy Educational Software

Terry Tao has utilized AI agents to modernize educational tools he created decades ago. Specifically, he has been rewriting legacy Java applets—originally written in Java 1.0 around 1999 for courses in complex analysis and linear algebra—into modern web-based formats.

These applets were used to visualize mathematical objects, such as honeycombs or Besicovitch sets. The transition from obsolete Java applets to modern JavaScript allows these educational materials to remain accessible and interactive in a modern browser environment.

Balancing Risk and Utility in AI-Generated Code

While AI agents accelerate development, they are not a substitute for mission-critical engineering. Tao notes that for interactive supplements to a mathematical paper, the risk associated with using guided interaction with LLM agents is acceptable because these tools are not central to the core mathematical proofs or the paper's primary findings.

"as such [LLM-coded interactive] supplements are not mission-critical to the core of the paper, I again feel that the downside risk of using guided interaction with LLM agents to generate such visualizations is acceptable."

Community Insights on the Impact of Coding Agents

Discussion among developers and educators suggests that the impact of AI coding agents extends beyond simple hobby projects into broader educational and professional shifts:

Expanding Software Accessibility

There is a significant amount of "latent demand" for software in non-software-focused fields. AI agents enable a large group of highly intelligent people who are experts in their respective domains but lack formal coding training to finally build the tools they need.

Reviving Legacy Code

The ability to port old code to new languages is a practical application of these agents. For example, community members reported using Claude to port 30-year-old Java applet games to JavaScript, effectively bringing dead code back to life.

The Shift in Software Moats

Some observers suggest that as AI becomes capable of disassembling and rewriting sophisticated software, the competitive advantage (or "moat") for software companies may shift away from the code itself and toward massive data storage and hardware assets.

Educational Applications

Educators have noted that LLMs provide a major boost for creating classroom visualizations. One example includes the design of a simplified 8-bit computer to complement a 16-bit teaching computer, created in a few days with the help of Claude.