dynamicfeed/live-facts-snapshot

Live Facts Snapshot

A daily snapshot of verifiable, post-training-cutoff world-state facts — the kind of ground truth language models cannot know from training data — exported through Dynamic Feed, a live, verifiable data API whose every response is Ed25519-signed. One file per day (data/YYYY-MM-DD.jsonl), one fact per line, and every row carries its own source, source_url and measured_at.

Facts covered per day:

• —Live source: https://dynamicfeed.ai (endpoint POST /v1/batch) — keyless, no signup
• —Source / licence: multiple authoritative sources — each row carries its own `source` and `source_url`; rows from US federal sources are public domain; the compilation is by Dynamic Feed
• —Update cadence: daily

File format

The first line of every file is a `_meta` record ({"_meta": true, ...}) with the export date, the API calls made, and the Ed25519 signature of the live batch response the rows came from. All following lines are fact rows:

Signature verification

Every Dynamic Feed API response is signed with Ed25519 (key id df-ed25519-4cb32e72f333, canonicalization json-sorted-compact):

• —Public key: https://dynamicfeed.ai/.well-known/keys
• —Signing spec: https://dynamicfeed.ai/standard
• —In-browser verifier (paste any live response, tamper-test it): https://dynamicfeed.ai/proof

The _meta line stores the signature object exactly as returned by the live API, plus the SHA-256 of the raw signed response bytes. To verify end-to-end, fetch the live API yourself (snippet below) and check the signature per the spec.

Use the live API (keyless)

import json, urllib.request

req = urllib.request.Request(
    "https://dynamicfeed.ai/v1/batch",
    data=json.dumps({"calls": [
        {"tool": "treasury_yields", "args": {}},
        {"tool": "interest_rates", "args": {}},
        {"tool": "market_hours", "args": {}},
    ]}).encode(),
    headers={"Content-Type": "application/json"},
)
resp = json.loads(urllib.request.urlopen(req).read())
yields = resp["results"][0]["data"]["yields_pct"]
signature = resp["signature"]          # Ed25519, key df-ed25519-4cb32e72f333
print("US 10Y:", yields["10Y"], "% — signed by", signature["key_id"])

Load this dataset

import json

with open("data/2026-06-10.jsonl") as fh:
    meta = json.loads(next(fh))        # first line is the _meta record
    facts = [json.loads(line) for line in fh]
for f in facts[:5]:
    print(f["fact"], "=", f["value"], f"({f['source']}, {f['measured_at']})")

(If you use datasets.load_dataset("json", ...), filter out rows where _meta is true.)

About Dynamic Feed

Dynamic Feed serves 56 live data tools (weather, CVEs, markets, space, hazards, AI model pricing, ...) with a freshness + provenance envelope on every datapoint and an Ed25519 signature on every response. Keyless MCP + REST. Source catalog: https://dynamicfeed.ai/sources