AI Agents Hit Wall on Human 'Feel' in Software Design
Kun Chen posted that agents cannot judge whether software feels right to users.
TLDR
Kun Chen wrote that agentic engineering hits a wall on assessing how software feels to users. He noted agents walk past design issues unless explicitly told about them. Several engineers replied that they notice the same off quality in agent output. One said the sensibility is scarce even among humans at large companies. Others described shifting their own role to reviewing feel rather than writing code line by line.
Combined views
—
13 Sources, first seen ago