I run coding agents all day, so I added up a week of my own API usage: 19,733 Claude calls, about $1,064 at API prices. The biggest line was not output and not cache reads. It was cache writes: $469, 44% of the bill. A cache write is what you pay the first time a prefix is seen. In a long conversation almost everything should be a read. A write that large means the agent kept paying for conversations it had already paid for. Leak 1: a note that “preserves the cache” The agent runs with a contex…

I use Claude Code and Codex every day, and I kept typing the same things: “do all of it, not just one example”, “did you actually run it before saying it’s done?“. So I counted. Across 6,752 of my own agent transcripts (4,990 messages I wrote myself), the most common corrections were: What I corrected Times Did only part of what I asked (“do all 8, not one and ‘the rest the same way’“) 174 Said “done” without running it 63 Went back to a direction we had already dropped in an earlier session 31…