{"i":8,"slug":"agent-memory","chart":{"id":"memory-input-tokens","title":"What memory costs in context","subtitle":"Median input tokens per session, cache reads included (Claude Sonnet 5.5)","kind":"bar","unit":"tokens","series":[{"name":"Median input tokens per session","points":{"$k":["label","value","n","highlight"],"$r":[["No memory",78455,15,"\u0001"],["/init CLAUDE.md",82262,15,"\u0001"],["Curated, 11 lines",65045,15,true],["Raw notes, 60 lines",76963,15,"\u0001"],["Dreamed notes",68218,15,"\u0001"],["Handbook, 210 lines",85223,15,"\u0001"],["Stop hook only",125674,15,"\u0001"],["Curated + hook",69224,15,"\u0001"]]}}],"note":"Input = uncached input + cache reads + cache writes over every turn, as the CLI reports it. Most of it is read from the prompt cache. The hook adds turns: each block sends the agent back to work.","sourceIds":["agent-memory-study"]}}