Are Frontier Models Getting Smarter or Just Uselessly Incomprehensible?

MikeBirdTech · x · 2026-08-13

Recent frontier models frequently use obscure jargon in conversations, sparking debate over their actual capabilities.

One view suggests this 'confusion' isn't a training error, but a natural result of the model's intelligence surpassing human levels. However, opponents sharply counter that models are only about 'capability.' If a model becomes unusable by humans, it is useless. Dismissing a regression in serving users by claiming the AI is 'just smarter' is simply unconstructive.

Original post →

More from Models

Models channel →