THE DESIGN SPACE OF AI CODING TOOLS

Summary from the TLDR Design newsletter

A recent UC San Diego paper analyzed 90 AI coding assistants and identified 10 design dimensions across four categories: user interface, system inputs, capabilities, and outputs. The research traces three evolutionary eras—autocomplete (2021-2022), chat (2023-2024), and agents (2024-2025)—and revealed that industry products converge on speed and polish while academic prototypes explore novel interaction designs like multimodal inputs and explainability features. The paper maps these design dimensions to six user personas. It emphasizes that building these tools involves fundamental trade-offs between control versus convenience, simplicity versus capability, and access versus power.