A short section acknowledging the economic facts about LLMs.
- Current subscription models are subsidized, costs will likely go up
- Understanding how tokens cost per token: $ / million tokens (input, output, cache).
- Use the smallest model for the task. It takes time to develop experience in this issue.
- In the long term, hopefully self-hosted, open-weights model will become practical as a coding tool. This isn't quite the reality yet.
- Explain what a token is
A short section acknowledging the economic facts about LLMs.