Spend is easy to total. Hard to locate.
The Money Map decomposes spend by step (operation, task class, model), so a rising bill has an address: a retry loop on one task class, rather than a total that went up. It shows cost per outcome first, then each step’s waste split between retries and cache, so the fix is aimed at the line that moved. A metric is real or it does not render.
Locating spend is not the same as ranking it. The map ranks by dollars, because dollars are the part we can measure; two steps burning the same amount can be worth very different things to your business, and nothing on this page knows which is which. Only you do.
See the map, at a buyer’s scale ↓Every dollar our committed evidence cost us, ranked by step. Panel A is the estate’s tree from the previous part, read as a bill. Panel B opens one line, docpipeline-reasoning: the same 115 tasks re-run at thinking budget 128 landed more outcomes for $0.0352, so the rest of the full-budget arm’s bill bought nothing the outcomes needed.
An address, not a total. The biggest line on our own bill is not where the fix was. The fix sat on the 11% line: a budget knob left at its default, and quality rose when it was cut. On your estate the waste hides elsewhere — retries, cold caches, a route nobody re-priced — but the move is the same: locate first, then act.
is_simulated=false in all threecaptured 2026-08-19 → 2026-09-19Source: the committed artifacts provenance/{oss_cost_per_outcome_usd, autoroute_defended_savings_frac, reasoning_effort_savings_frac}.json. The ranking, both shares and the split derive from them at build time; none is typed.
The map itself, at a buyer’s scale.
Our own books are three workflows; yours will not be. This is the Money Map drawn on a stand-in company, simulated and labelled so, every system priced by the same code that priced Fig. 01. Each box is a system with its share of the bill drawn as a bar, and each opens to the operations inside it. Press the preview under the map and it re-prices off the first fix Margin would make. The numbers come from the server; nothing is computed on this page.
- The Meter: measures every call, its cost, tokens, substrate, and whether it worked.
- The Estate: the org chart of your AI workforce.
- The Money Map: where the spend goes, decomposed by step.
- The Parity Gate: proves the cheaper route held quality before it ships.
- The Governor: acts with a human ratifying, and a revert armed; if quality slips it reverts and the loop re-runs.
Watch the loop run on real spend.
The console shows the loop on open-source agents we metered ourselves: measured spend, recorded face-offs and the gate’s verdicts, every number tagged real or sample. No live route has slipped yet, so the revert has not fired.