Why almost all GPT models are lazy and with every new model they get even lazier for serious tasks?
This laziness is specially a show stopper for use of OpenAI models for coding. Even the new GPT-4.1 used in windsurf is lazy, doesn’t follow instructions, provides half done work in the output)
No matter how smart, when a model is lazy it can not be trusted for real world problems and use cases. Please fix.
Thank you!