Is there something we are all missing? Using Claude feels like magic sometimes, but can't everyone see the limitation now that we are 4 years and 100s of billio

Is there something we are all missing? Using Claude feels like magic sometimes, but can't everyone see the limitation now that we are 4 years and 100s of billions down the road? Are they still really hoping that they are gonna tweak a model and feed it an even bigger dataset and it will be AGI?