Slide 06 — Expand Cached Token explanation (what it is, how it forms)

[KardelRuveyda]

Feedback

Slide 06 mentions Cached Token but doesn't really explain 'what it is, how it gets created, why it costs less'.

Problem: If the listener is hearing 'cached token' for the first time, the rest of the formula stays in the air.

Information to add

• Definition: Instead of re-sending the same prompt context to the model, the previously processed part is reused from cache.
• How it forms: Prevents re-generating tokens — especially useful for long, repeating system / context prompts.
• Why it's cheaper: Lower inference cost on the provider side → reflected back to us as a lower credit rate.
• Practical effect: Credit usage drops noticeably on follow-up questions in the same file.

Suggestion

• Add a small diagram (Full token vs Cached token).
• One-line formula: Cached token = no regeneration = lower credit rate.

File

• src/components/sections/AiCreditsSection.tsx (taxi-meter / formula section).