Bild: The New StackClaude, Gemini, and GPT-5 can handle every SDLC task. Almost none of them should. - The New Stack
• The article argues that while frontier LLMs like Claude, Gemini, and GPT-5 can technically perform every Software Development Life Cycle (SDLC) task, relying on a single general-purpose model is inefficient for enterprise-wide strategies. • Over-reliance on these large models for code generation, testing, and documentation leads to excessive consumption of LLM inference and compute resources. • The author suggests that AI-native software delivery requires a shift toward specialized models, dedicated AI control planes, and a more modular architecture.
thenewstack.io

