Build Token Value

ASEC multi-op · OTC fan-out · Tarski-honest

How many tokens did the build cost — and what does that not mean?

Folds Claude measured usage, Grok/Codex session proxies, and four value layers that must never collapse: cash paid, API-list equivalent, human opportunity, and artifact market (oracle). Public seed is one machine’s certificate — recompute yours.

Loading certificate…

Open raw JSON

Value layers

L2 is cost of model work at list rates — not invoice, not IP sale, not human hours.

L1 cash ↔ L2 API-list gap

Never sum these layers. Seat cash underprices list-rate token folds (especially cache).

Billing integration (L1 cash)

Configure local invoices/CSV/env — never inferred from tokens. Template: billing.config.json · config/billing.example.json

Loading billing status…

Claude activity (daily)

From stats-cache dailyActivity — messages/sessions/tools, not billable tokens.

Date	Msgs	Sessions	Tools

What-if (list-rate calculator)

Client-side only · uses pricing.json · not cash paid.

Tier Input MTok Output MTok Cache read MTok Cache write MTok

—

Multi-op readiness + contributions

Fan-out when ≥2 backends are cli_ok (OTC multiset — not exclusive cascade).

Double-count risk graph

Shared providers across harnesses — do not naively sum all proxies.

Value epochs

Ordered by evidence class — not calendar arrival dates (synergeticnow-style honesty).

Token composition

Cache-dominated Claude traffic vs non-cache + proxies.

Series	Tokens

API-list $ by model (Claude)

Skill pricing table as-of methodology JSON.

Model	Tier	API-list USD

Harness bands

Measured Claude · semi-decidable Grok/Codex proxies.

Source	Class	Tokens	USD (list)

Top 10 agentic systems — ready prompts

Prompt

How to recompute

Deterministic local fold — then drop data/estimate.json and redeploy.

Billing (L1): copy config/billing.local.json.example → config/billing.local.json; set manual USD or CSV env paths
Optional: refresh Claude stats — py -3 public/scripts/recompute_stats_cache.py
Pipeline — py -3 scripts/pipeline.py --since 2025-11-01 (evidence + billing + cursor + estimate + summary)
Optional multi-op — py -3 scripts/pipeline.py --multi-op when backends are up
Verify — py -3 scripts/verify_deploy.py then package: py -3 scripts/build_git.py
Deploy — wrangler pages deploy public --project-name=build-token-value --branch=main

methodology/SKILL.md data/methodology.json data/billing.config.json data/evidence.summary.json scripts/estimate_value.py multi-op prompt certificate.html

Tarski residuals

Always named — never silently defaulted.

Clone (CF dumb-HTTP bare git)

Same pattern as praxis-stack /git/*.git: static bare repo on Pages, git update-server-info, no server-side git daemon required.

git clone https://build-token-value.pages.dev/git/build-token-value.git

Package includes methodology, prompts 00–10, fold scripts, and this dashboard.
Use a full clone (dumb-HTTP does not support --depth).
Dashboard is static: update data/estimate.json and redeploy.

data/estimate.json data/methodology.json data/agents.json methodology/SKILL.md git info/refs