Blog · 2026-09-06 · 7 min
What one observed coding session reveals about the re-reading tax — and how memory ends it.
Direct answer
An observed Claude Code session burned 8.59 million tokens at 97.2% cache hit rate — and roughly 90% of them re-read files the agent had already seen. This is the canonical shape of agent waste, and the reason memory layers exist.
01
8.59M input tokens in a single session. 97.2% served from prompt cache — the bill was already discounted about as far as caching can discount it. Input-to-output ratio around 175:1. And the split that matters: on the order of 90% of tokens spent re-reading repository content, ~10% on actual reasoning toward the goal.
At 2026 Claude-class pricing ($2–5 per million fresh input tokens, ~10% for cache reads), that session cost tens of dollars — nearly all of it paying the agent to re-learn what it had already learned. Multiply by sessions per week per developer and the re-reading tax becomes a budget line.
02
Three compounding mechanics. First, every token loaded is re-billed on every later turn until compaction — a 10K file read at turn 3 costs another 10K at turns 4, 5, 6. Second, agents over-read by design: opening a whole file is safer than opening nothing, so navigation defaults to dumps. Third, nothing persists — the next session starts from zero and pays the orientation bill again, in full.
Caching cannot fix this. The session was already 97.2% cached. Cache discounts price; it does not shrink attention load or orientation repetition. Only memory — durable, curated, retrieved under budget — attacks the quantity side.
03
With a compounding memory graph, the same work looks different: first session pays orientation and records typed nodes (decisions ~16x, bugfixes ~24x, threads ~25x compression). Session two retrieves instead of re-reading. Typical turns assemble 2–4K packed tokens instead of ~50K raw — the 12–25x per-node design estimate measured in the Anvaya Mind repos.
The 8.59M-token session becomes, by construction, a ~1M-token session at worst — and its successors keep shrinking as corroborated knowledge replaces exploration. That is the compounding thesis: session N+1 cheaper and sharper than session N.
04
One task per session — a 3-hour mega-thread costs 3–5x the history tax of three focused sessions. Reference spans, not files. Scope test commands and pipe verbose output through head. Keep AGENTS.md/CLAUDE.md to non-inferable rules. Then add the layer files can't provide: a memory graph that remembers why, not just what.
Questions
No — it is an observed Claude Code session from public reporting, used industry-wide as the canonical waste example. Anvaya's own per-node ratios (12–25x) are design estimates from the Mind repos; open measurement trials with published scripts arrive Q3 2026.
No. The session fit in the window — that was the problem. It paid to attend to 8.59M tokens regardless of ceiling. Windows are headroom, not efficiency.
Cheaper than fresh, still expensive: cache reads cost ~10% of fresh, and 8.59M × 10% across hundreds of turns is real money — before counting the attention cost of noise.
One binary. 11+9 Rust crates. 545 tests. Hand-written HNSW index. Three transport modes. Four providers, Ollama, Anthropic, OpenAI, Siemens. Zero API keys required to start. Mind remembers everything after the first session.