After changes to pricing structures for LLM powered code assistants, Mike looks at how a seemingly simple task can burn through tokens when an agent 'looks after it' for you.
