AI Prediction Ledger API Documentation

Endpoints

JSON response envelope

On narrow screens, scroll this table sideways. Keyboard users can focus the table area and use the arrow keys.

Prediction record fields

The stable eight-field export does not include statement dates, exact quotations, evidence locators, duplicate-family IDs, historical-score eligibility or full assessment receipts. Consult the ledger for those disclosures. Do not calculate calibration or select numeric scores from this export without the published eligibility policy.

On narrow screens, scroll this table sideways. Keyboard users can focus the table area and use the arrow keys.

Revision object fields

On narrow screens, scroll this table sideways. Keyboard users can focus the table area and use the arrow keys.

Examples

Fetch the JSON dataset:

curl -L https://kingy.ai/wp-json/kingy/v1/ai-predictions

Abbreviated example using current record 001. The live response contains all 100 records and the full revision history; arrays below are shortened for readability. The timestamp is illustrative, not a source-review date.

{
  "schema_version": "1.0",
  "dataset": "ai-prediction-ledger",
  "source_url": "https://kingy.ai/blog/ai-prediction-ledger/",
  "updated_at": "2026-10-05T19:34:46Z",
  "record_count": 100,
  "records": [
    {
      "record_id": "001",
      "forecaster": "Jacob Coxon",
      "claim": "Coxon warned that AI builders believe AI could kill all humans by the end of the decade; this post gives no numerical probability.",
      "status": "pending",
      "score": "Pending",
      "topic": "risk",
      "horizon": "End of the 2020s; qualitative warning attributed to AI builders",
      "source_url": "https://x.com/hilbertspaess/status/2097476203863224394"
    }
  ],
  "revisions": [
    {
      "version": "1.20",
      "date": "2026-10-05",
      "change": "Completed page navigation, high-contrast responsive layout, record permalinks, forecaster filtering and paginated browsing; preserved per-record source/correction history and five historical score limitations. Added a clearly fictional manual-review walkthrough, visible current/historical evidence gaps, source availability report and direct-install SDK package archives. Reconciled the API guide with current claims, timestamp semantics, export limitations and registry authentication status. All 100 canonical records, statuses, scores, historical-score eligibility, prior revisions, methodology 1.0, receipt template 1.0, validator 1.0 and existing download/API-contract bytes unchanged. Monthly automation instructions and cadence unchanged; no new outcome assessment or claim that registry publication succeeded."
    }
  ]
}

Fetch CSV for a spreadsheet or data pipeline:

curl -L https://kingy.ai/wp-json/kingy/v1/ai-predictions.csv -o ai-prediction-ledger.csv

Download the OpenAPI contract:

curl -L https://kingy.ai/wp-json/kingy/v1/ai-predictions/openapi.yaml -o ai-prediction-ledger-openapi.yaml

Download the generated SDKs:

curl -L https://kingy.ai/wp-json/kingy/v1/ai-predictions/sdk/javascript -o ai-prediction-ledger-sdk.js
curl -L https://kingy.ai/wp-json/kingy/v1/ai-predictions/sdk/python -o ai-prediction-ledger-sdk.py

Registry installation (not yet available; publishing authentication required):

npm install kingy-ai-prediction-ledger
python -m pip install kingy-ai-prediction-ledger

Checked October 5, 2026: versions 1.0.0, 1.0.1 and 1.0.2 are not published on npm or PyPI, and this session has no usable publishing credentials. Do not treat the commands above as working registry installation instructions. Direct-install archives below are available now.

Use the JavaScript client:

import AIPredictionLedgerClient from 'kingy-ai-prediction-ledger';
const client = new AIPredictionLedgerClient();
const ledger = await client.getLedger();

Use the Python client:

from kingy_ai_prediction_ledger import AIPredictionLedgerClient
client = AIPredictionLedgerClient()
ledger = client.get_ledger()

Install directly from downloadable packages

SDK version 1.0.2 targets API v1 and schema 1.0. It is independent of the ledger’s content revision. These archives include MIT license, package metadata and the corrected OpenAPI contract 1.0.1.

• Download JavaScript package 1.0.2 (.tgz)
• Download Python package 1.0.2 (.whl)
• Download checksums and installation instructions (JSON)

npm install ./kingy-ai-prediction-ledger-1.0.2.tgz
python -m pip install ./kingy_ai_prediction_ledger-1.0.2-py3-none-any.whl

The clients expose ledger JSON, CSV, revision history and the OpenAPI contract. Standalone source downloads are also available above. No client method writes to the ledger.

Versioning policy

Contract patch 1.0.1: quotes the string status false correctly. Contract 1.0.0 used an unquoted YAML boolean; its schema does not validate that string status. Endpoint namespace v1, schema 1.0 and all record values remain unchanged. Use 1.0.1 for validation. Earlier archives are immutable, not silently replaced. SDK 1.0.2 also corrects the stale Python runtime version marker in SDK 1.0.1; use the current downloads.

• Immutable OpenAPI contract 1.0.1
• Historical OpenAPI contract 1.0.0 (known enum defect)

The /kingy/v1/ namespace is the compatibility promise. Additive fields and new records may appear without changing the namespace. A breaking change to field meaning, response shape, identifier rules or endpoint behavior will ship under /kingy/v2/ with a migration note and an overlap period where practical.

Record IDs are stable within the dataset. A correction changes the current record only through a dated ledger revision; the previous state remains discoverable through the revision history and the monthly review process. Consumers should ignore unknown object fields. The documented v1 status/topic enums remain the contract; a new enum value requires an announced compatible contract update and migration guidance rather than an undocumented assumption.

Errors, caching and data semantics

• Successful responses use HTTP 200. A missing registry returns a structured WordPress REST error rather than an empty dataset.
• Endpoints are public and read-only; no API key or personal data is required.
• Use the revision endpoint for published change summaries; updated_at is the ledger-post modification time, not a claim-verification timestamp. Consumers should cache responsibly and revalidate at least monthly.
• There are 100 tracked records, including scenarios, repeated statements and unresolved assertions. Citation presence does not authenticate an original claim. Source support and outcome resolution are separate; the ledger’s per-record notes and historical-score limitations govern interpretation.

Questions or corrections belong on the ledger page with the original source, date, context and proposed resolution rule.