Very well said. I think that the DeepSeek is now changing that landscape a bit. Running smaller DeepSeek models locally with ollama can now provide quality inference, and provide a solution to these concerns.
Sorry to bug you, @Saif may I get your help with related topic here: How to properly debug AI Personas? Thank you!