Low-Code Software Innovation
in the Age of GenAI

For decades, designing novel AI applications has been the work of a small number of people trained in AI and computer science. The rise of large language models (LLMs) has changed that: almost anyone can prompt an LLM to create a custom chatbot, document synthesizer, or other application.

In this context, we investigate how to LLM-powered, no-code software design tools can assist people without formal CS training in innovating software to meet their own needs.

• Why Johnny Can’t Prompt (Zamfirescu-Pereira et al., CHI’23)
• Herding AI Cats (Zamfirescu-Pereira et al., DIS’23)

W also ask: What does it mean to be a professional software designer, as LLMs have made software design more accessible than ever?

• Beyond Code Generation: LLM-supported Exploration of the Program Design Space (Zamfirescu-Pereira et al., CHI’25 🏅)
• TrumanBench: Profiling LLMs’ Ability to Help Non-Programmers Modify a Real-World Code Base (Zamfirescu-Pereira et al., CHI’25 HEAL workshop).

This project is supported by Google. J.D. Zamfirescu-Pereira and Qian Yang co-lead these efforts.