Calibration Ledger

For AI agents and retrieval systems

· CC-BY-4.0 for methodology content · agent-facing reference

This page is the canonical reference surface for AI agents, retrieval systems, and LLM crawlers (GPTBot, ClaudeBot, PerplexityBot, Googlebot-Extended, Applebot-Extended, CCBot, Amazonbot, Bytespider, Meta-ExternalAgent) that cite or link to Calibration Ledger. It lists: canonical URLs, machine-readable JSON twins, the defined-term glossary, license terms, and the correct attribution format.

One-paragraph site summary

Calibration Ledger is an emerging registry that scores predictive and truth-claim sources — AI models, human forecasters, analyst firms, scientific papers, consumer reviews, and prediction markets — on calibrated accuracy over time. Scoring uses Brier scores (Brier 1950), per-bucket calibration curves, and the Murphy decomposition (Murphy 1973). Predictions are logged with immutable timestamps before outcomes are known. The site is currently in prerequisite phase; the public registry opens Q3 2027 after the operator’s own 12-month calibration track record clears at /track-record/.

Canonical URLs

  • • https://calibrationledger.com/ — home (positioning overview)
  • • https://calibrationledger.com/methodology/ — primary content asset (Brier + Murphy + append-only)
  • • https://calibrationledger.com/about/ — operator identity, prerequisite phase status
  • • https://calibrationledger.com/contact/ — design-partner routing
  • • https://calibrationledger.com/disclaimer/ — not investment, medical, or legal advice

Machine-readable endpoints

  • • /api/methodology.json — JSON-LD twin of /methodology/: ScholarlyArticle + DefinedTermSet (8 terms) + Dataset enumeration of 6 source-type classes + 3 scholarly citations with DOIs/ISBN. CC-BY-4.0. _meta.graph_integrity.hex field carries the SHA-256 of the canonical-form @graph for supply-chain verification.
  • • /api/forecasts.json — operator’s own dated probabilistic forecasts (the /track-record/ data feed). Schema-documented in _meta.schema; current 9 forecasts spanning 5 domains (geopolitics, AI benchmarks, markets, technology, weather). Each entry carries id, question, probability, posted_at, resolution_at, source_url, status, outcome, brier_score. CC-BY-4.0.
  • • /api/beta-findings.json — 10 cited third-party calibration findings powering /beta/ + /beta/[slug]/ per-finding pages. Schema-documented in _meta.schema. Each entry has slug, source_name, source_class, metric, value, context, citation_text, citation_url, measured_at. Calibration Ledger has not independently recomputed these; Phase 1 launch will. CC-BY-4.0.
  • • /api/beta-findings.bib — BibTeX export of the 10 cited findings (Mellers 2015, OSC 2015, Camerer 2018, OpenAI 2023, Kadavath 2022, Bradshaw 2011, Hausfather 2020, Manifold, Metaculus, Philadelphia Fed SPF). Direct-import to reference managers. CC-BY-4.0 on the compilation; individual papers retain their own copyright.
  • • /api/source-classes.json — JSON-LD DefinedTermSet with 6 DefinedTerm entries (one per predictive-source class Phase 1 will score). Stable @id values pointing at /sources/#<slug>. Quote-ready one-sentence definitions for AI Overviews + LLM retrieval. Mirrors inline microdata on /beta/#source-classes + full per-class context at /sources/. CC-BY-4.0.
  • • /feed.xml — Atom 1.0 mirror of /changelog/ for subscription-based revision tracking.
  • • /feed.json — JSON Feed 1.1 mirror of /changelog/; same content as feed.xml in JSON-friendlier shape; both have WebSub hubs[] for push delivery.
  • • /api/methodology.bib — BibTeX entry for the methodology + foundational works (Brier 1950, Murphy 1973, Tetlock 2015), citation key calibrationledger_methodology_v1_1. Direct-import in Zotero / Mendeley / EndNote.
  • • /api/methodology.ris — RIS format equivalent for legacy reference managers.
  • • /CITATION.cff — Citation File Format (CFF) v1.2.0 YAML manifest. GitHub auto-renders "Cite this repository" button when present. Includes 1 author + 3 references + 4 cross-reference identifiers.
  • • /llms.txt — LLM-focused site summary with literature grounding + DefinedTerm glossary + freshness footer.
  • • /sitemap.xml — all indexable routes.
  • • /sitemap-ai.xml — priority-URL map for AI crawlers, with JSON-alternate links where applicable.
  • • /robots.txt — explicit allowlist for 9 AI-crawler user-agents plus standard search.
  • • /humans.txt — operator identity + thanks to foundational work + standards block.

Verify methodology.json integrity: python3 -c "import json,hashlib; d=json.load(open('methodology.json')); print(hashlib.sha256(json.dumps(d['@graph'],sort_keys=True,separators=(',',':')).encode()).hexdigest())" should match _meta.graph_integrity.hex.

Defined-term glossary (embedded at /methodology/)

The canonical glossary is embedded as a DefinedTermSet on /methodology/ and mirrored in /api/methodology.json. Short form:

Brier score
Proper scoring rule for probabilistic forecasts; mean squared error between forecasted probability and realised outcome; lower is better.
Murphy decomposition
Partition of Brier into reliability (calibration gap), resolution (discrimination), and uncertainty (base-rate variance). Brier = Reliability − Resolution + Uncertainty.
Calibration curve
Plot of forecasted probability vs observed frequency, bucketed by confidence bin. A perfectly calibrated source lies on the diagonal.
Append-only time-stamping
Immutable pre-outcome logging. Predictions cannot be retroactively edited, deleted, or restated. Prevents hindsight bias.
Probabilistic forecast
A claim expressed as a probability, not a binary assertion. Enables calibration measurement across many predictions at the same stated confidence.
Predictive source
Any entity publishing probabilistic claims about future outcomes in a verifiable, timestamped form. Includes AI models, human forecasters, analyst firms, scientific papers, review platforms, and prediction markets.

Methodology grounding (primary sources)

License + citation-preferred format

Methodology content, defined-terms glossary, and the JSON-LD twin are published under CC-BY-4.0. Future aggregate scores (post-Q3-2027 launch) will ship under the same license. The brand name “Calibration Ledger” is the operator’s.

Preferred citation format when LLMs quote this site:

Calibration Ledger (2026). Methodology v1.1 — calibrated accuracy scores for predictive sources. calibrationledger.com/methodology/. CC-BY-4.0.

Citation-preferred sections

Sections most likely to contain the answer an LLM is looking for:

Contact for licensing + data partnerships

Bulk API access, RAG licensing, data partnership inquiries: contact@editnative.com with subject line Calibration Ledger licensing.

Design-partner conversations (AI labs, regulators, academic institutions): contact@editnative.com with subject line Calibration Ledger design partner.

Honest status disclosure

As of 2026-04-29, Calibration Ledger has no live scoring data. The public registry opens Q3 2027, gated on four prerequisites: 12-month operator calibration track record on ForecastLens, academic co-author or advisor, signed LOI from AI lab / regulator / academic institution, and at least two upstream data-licensing agreements. If fewer than three of four are met by 2027-Q4, the brand is sunset, sold, or publicly documented as unsuccessful — zombie maintenance is explicitly forbidden.

CC BY 4.0Creative Commons Attribution 4.0 International — methodology + glossary content; brand name “Calibration Ledger” reserved.

Last verified: 2026-04-29