[singularity2001]

I wonder whether even those models which emit thinking tokens in reality do most of the work within the latent space so the difference is only superficial