LLM Context Profiling: what is eating up your tokens?

"One of the most painful aspects of AI coding agents is the ever-increasing bills for token usage - especially now, when providers are either reducing limits or switching to usage-based billing. Since most of the token costs are typically spent on input, in this session we will delve into the contents of the context window, show how to identify the biggest issues using a custom-built profiler, what these tokens are spent on, and how they could be optimised."