Pilot: Prompt caching › Ep 1
Ep 1 · Why your second message is cheaper
Length 1:19Today's goal: You send the exact same thing twice, and the second time it costs a tenth of the price. Why?
What you'll learn
- You never send just your message: your tool wraps it with its own instructions, a description of every tool, your MCP servers and skills, and the whole conversation so far.
- In the demo, two words were the last line of a request more than twenty thousand tokens long (21,227), rebuilt and sent in full on every turn.
- The second one was cheap because the start was identical: the provider kept the work it had already done and only did new work on the end.
- “Identical” means identical. Change anything near the start, and the saving is gone.
Sources
- Burke Holland, “Your AI credits are vanishing - here's the fix” (YouTube): the 6.64 / 0.57 credits and the 21,227-token request come from his demo (about 1 credit = 1 US cent)