Frameworks are not enough: The future stack for AI development
AI makes code generation dramatically faster, but faster code does not automatically mean better software. Frameworks, developer experience and performance still matter, but the AI era brings new questions: How quickly can we detect errors? How much quality does the stack enforce automatically? And how easily can humans and AI agents understand and debug the same codebase?
- AI-assisted engineering
- developer experience
Delivered at FrontKon 2026, Slides
Available formats
Frameworks are not enough: The future stack for AI development
25 minutes
AI makes code generation dramatically faster, but faster code does not automatically mean better software. Frameworks, developer experience and performance still matter, but the AI era brings new questions: How quickly can we detect errors? How much quality does the stack enforce automatically? And how easily can humans and AI agents understand and debug the same codebase? Instead of ranking frameworks, this talk presents the future development stack as a system of feedback loops. It focuses on frontend and related systems while also covering types, schemas, tests, linting, CI, logging, monitoring, observability and quality gates. These tools determine whether a successful prototype can safely become production software. Based on hands-on work with teams adopting AI-assisted development, this is not a talk about replacing developers or writing better prompts. It is a practical and opinionated look at a stack that can survive the AI era: strict enough to prevent chaos, observable enough to explain failures and repairable enough to keep moving.