Everyone Chases a Million-Token Context. Most Work Doesn't Need It.

GPT-6.1 Sol now ships a million-token context, and the numbers war is on. I am here to say most of my work does not need it, and chasing it can make results worse.

A bigger context means more to get wrong, more to pay for, and more places for the model to lose the thread you actually care about. Size is not the same as relevance.

For the long-document jobs, sure, give me the window. For the everyday question, a tight, well-scoped prompt beats a firehose of context every time.

Match the context to the job. The biggest context is a flex, not a feature, and treating it like one produces answers that are long, expensive, and oddly less useful.