AI-generated content often feels generic and unreliable, resembling “toys” rather than professional tools. Replit CEO Amjad Masad identifies the core issue: Everything looks the same, from images to code, Taryn Plumb reports for VentureBeat.
The problem, known as “slop,” stems from lazy prompting and a lack of individual flavor. Masad believes platforms must expend more effort and developers should “imbue the agent with taste.”
Replit addresses this through specialized prompting, classification features, and proprietary RAG techniques. The company uses more tokens to generate higher-quality inputs. After creating an app, a testing agent analyzes its features and reports back to the coding agent. “If you introduce testing in the loop, you can give the model feedback and have the model reflect on its work,” Masad says.
The company pits different models against one another, with testing agents built on one LLM and coding agents on another. This approach capitalizes on their different knowledge distributions and generates more variety.
Masad predicts “vibe coding” will transform how enterprises use AI. This approach allows any employee to solve problems through automation without traditional software engineering skills. “I would say that the population of professional developers who studied computer science and trained as developers will shrink over time,” he says.
Traditional development roadmaps no longer work because AI capabilities evolve too rapidly. Replit’s team remains agile and evaluates new models immediately upon release.
