# Source Watch: collection and review method

Included in Loop Evidence 1.1. Use Observe sources to collect an authorized public HTTPS source on demand. No hosted subscription or model credits are required.

## A useful engagement

Agree the business question, permitted sources, observation window, review owner and delivery format. A reviewer explains meaningful changes and their implications for that question. A page fingerprint alone is not a research conclusion.

## Recorded evidence

Each observation stores the public URL, local UTC recording time, HTTP response status, content type, byte count, response fingerprint and collection latency. It compares a successful result with the last successful observation. Failed collection remains in history, and the previous successful baseline is preserved. Source contents are not retained or republished by this collector.

The first observation is a baseline. Repeated identical observations establish only that the observed response bytes matched at those times. Different fingerprints can reflect formatting, timestamps or dynamic content. A reviewer must inspect the source and cite actual support for a meaningful difference. Collection gaps are disclosed; no continuous coverage or independent trusted timestamp is claimed.

## Five review criteria

1. Agree the source list, rights, question and observation window.
2. Retain an initial dated observation and fingerprint.
3. Compare a later observation and disclose missing coverage.
4. Review meaningful changes separately from formatting and collection errors.
5. Deliver the findings, limitations and portable evidence for buyer review.

## Boundaries

Public HTTPS only, no credentials, query strings, fragments, custom ports, private/local addresses or redirects. TLS verifies the original hostname against a pinned public address. Responses are capped at 2 MB. The user confirms source permission before each request. Respect access terms; do not use this to bypass login, access controls, blocks or rate limits.

Customers own their observations. Cross-customer benchmarks or secondary use require a separate explicit rights agreement. No exclusive dataset, customer network or long-term historical coverage exists merely because the toolkit is available.

The former Loop Research hosted runner and credit top-ups remain paused. This tool does not restart it or resell API access.