Why AI Feels Generic: Replit’s CEO Says “Taste” Is the Missing Ingredient

In a candid discussion on the Beyond the Pilot podcast, Amjad Masad, CEO of Replit, argues that much of today’s AI output feels bland, repetitive, and unreliable. He labels this phenomenon “slop” — a byproduct of shallow prompting, limited effort, and a lack of what he calls “taste.” Masad explains how Replit works to overcome this sameness and why “vibe coding” may define the future of software development as AI capabilities evolve faster than traditional planning can keep up.

Key Points

• Masad describes many current AI tools as “toys” — experimental, generic, and only marginally effective.
• The rise of “slop” is driven by sameness across AI-generated images, code, and content.
• Replit fights slop using specialized prompting, built-in classification systems, proprietary RAG techniques, and higher token usage for better inputs.
• Continuous testing is central: Replit runs outputs through testing agents that feed structured feedback back to coding agents.
• The company pits different LLMs against each other to increase variety and reduce low-effort outputs.
• Masad believes “vibe coding” will enable non-developers to build software solutions, shrinking the traditional developer population.
• Enterprises must abandon rigid roadmaps and adopt a more fluid, agile mindset as AI capabilities shift rapidly.

Key Quotes

• “There’s a lot of sameness out there… all the images, all the code, everything.”
• “The way to overcome slop is for the platform to expend more effort and for the developers of the platform to imbue the agent with taste.”
• “If you introduce testing in the loop, you can give the model feedback and have the model reflect on its work.”
• “That way the product you’re giving to the customer is high effort and less sloppy.”
• “Vibe coding… can make everyone in the enterprise the software engineer.”
• “You need to be very zen about it and not have an ego about it.”

Implications

Masad’s perspective suggests that AI’s biggest limitation isn’t intelligence, but craftsmanship. As models commoditize, differentiation will come from systems that apply judgment, feedback, and iteration — not just raw generation. For businesses, this means rethinking who can build software, how fast teams must adapt, and why flexibility may matter more than long-term planning in an AI-driven world.

Source: https://venturebeat.com/infrastructure/why-ai-feels-generic-replit-ceo-on-slop-toys-and-the-missing-ingredient-of