Hacker News

points by odo1242 21 hours ago | 0 comments

But you also burn the exact same number of tokens (actually more) every time the model makes a tool call. It's an agent, it makes a lot of tool calls.

[–] kristjansson 21 hours ago

This is the right perspective. It costs as much as one additional tool call in the conversation. It’s a bit weird to do it on a timer, which inflates a bit over “just a tool call”, but not by much?

[–] brookst 16 hours ago

Tool calls don’t sent the full transcript? I think you mean one more turn?

[–] hedora 14 hours ago

Pretty much everything sends the full transcript. That’s how LLMs work, and why “compact before cache expiry” is now a default in Claude Code. Cache hits are incredibly cheap to service on the provider side.