the model that got worse on purpose

A lab I won’t name shipped a smaller, slower version of their flagship model last week and called it an upgrade. It is, for almost everyone. I spent the weekend figuring out why that’s the more interesting story than the frontier model it’s not … more, eventually →

The full piece breaks down why capability and usefulness diverge, why the frontier race is a distraction for most real applications, and what happens when the industry realizes the bottleneck isn’t model size — it’s the unglamorous layer nobody wanted to invest in.