Token Usage

Preamble

Token Usage is the receipt layer for the garden. It asks what building with AI consumes, which project received the work, and where the bill is still missing.

---

The Claim

AI work feels weightless because the cost disappears into subscriptions, rate limits, hidden infrastructure, and half-remembered sessions.

The garden should not get that luxury. If a project was grown through model labor, the page should eventually be able to show the receipt: session count, tokens, cost, attribution, and claim boundary.

The current data is incomplete. That is part of the public claim boundary. The pipeline is being built while the garden is being built, so the honest receipt has to show both the numbers and the places where attribution has not landed yet.

What It Tracks

The project tracks the path from private work to public receipt:

• which coding, writing, or planning session fed which project
• how many tokens were used
• what those tokens cost
• whether a session should be split across multiple projects
• which garden node should show the receipt
• which rows are missing, estimated, or unattributed

Why It Matters

Cost is money, and it is also a proxy for dependency, speed, attention, and machine labor.

A cheap artifact can still be expensive if it replaces judgment. An expensive artifact can be worth it if it creates reusable infrastructure or exposes a useful failure. The ledger leaves that argument open and harder to avoid.

Current Boundary

The public site can show receipts where telemetry exists. It should not invent clean totals while the pipeline is still sparse.

The next version needs stronger session capture, project attribution, allocation across multiple projects, and a public export that distinguishes observed cost from estimated cost.

Branches

• The Metabolic Ledger explains the attribution model.
• Cost Snapshot: Garden Build holds the current sparse public receipt.