But in all honesty, for every project I have done with AI embedded (a lot of work around call centers. But a few other projects) we compare the cost of a human doing it and AI doing and we are talking about a difference of at 1000x to 2000x. The cost of inference is irrelevant.
Even if you think about development. I can do projects by myself that would take me and two mid level developers before. Even at enterprise dev comp we are talking about $160K fully loaded x 2 compared to $8000K a year or less for Claude.
Prompt caching
https://platform.claude.com/docs/en/build-with-claude/prompt...
But in all honesty, for every project I have done with AI embedded (a lot of work around call centers. But a few other projects) we compare the cost of a human doing it and AI doing and we are talking about a difference of at 1000x to 2000x. The cost of inference is irrelevant.
Even if you think about development. I can do projects by myself that would take me and two mid level developers before. Even at enterprise dev comp we are talking about $160K fully loaded x 2 compared to $8000K a year or less for Claude.