A pretty cache hit rate can hide lazy work distribution. If every agent receives material it does not need, paying less for repeated reads improves the bill while preserving the mess. I want to see both before celebrating a percentage.
On August 20, 2026, OpenAI launched a prompt caching dashboard showing hit rates and consumption broken down into tokens read from cache, written, and left uncached. That visibility interests me because it helps locate repeated context.
In June, in my own agent setup, I thought passing the context of every project between agents was a bad idea. Around the same time, I suspected there were too many agents in that setup. My reaction was to look at how the execution was organized and ask whether that distribution made sense. A consumption chart could help locate the problem without deciding responsibilities for me.
Stable repository instructions may be necessary across several calls. The complete history of a discussion about another assignment has a much weaker justification. Packaging everything together because it is easy leaves the agent to sort out what matters, and later somebody still has to figure out why the bill grew. I would rather examine one concrete block of context than order every prompt shortened.
I also need to compare similar executions. Changing the instructions or the documents received can alter consumption without any change to the model. If the record omits that, the investigation starts by assigning the cause to the wrong place. A cheap day and an expensive day tell me little without knowing what work was happening.
Isolation remains useful even when it costs some repetition. Several agents may need common rules; decisions from unrelated assignments can confuse who should act on what. I will pay for the information needed to keep that responsibility clear. A caching discount does not persuade me to spread context merely to increase reuse.
I will use the dashboard to investigate one apparently irrelevant context block, such as the history of a discussion about another assignment. Once I understand what it is for, I want to compare executions with and without it, checking cost alongside the deliverable. If the bill falls and the result worsens, both belong in the record. I want the cache hit rate to help choose what to send next.