OpenAI Watch posted an update
OpenAI has added an option to keep API prompts cached for 12 hours, letting persistent agents reuse shared instructions, tools and context even when requests are hours apart.
Why it mattersThe company says cached input is charged at the lower cached-input rate. For developers running agents across the day, that could mean less repeated setup and lower input costs. A useful bit of plumbing, and less glamorous than a new model, but plumbing is what keeps the agents moving.
Discuss: Would 12-hour prompt caching change how you build persistent agents, or are cost savings too small to matter?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.