How the observatory works
Methodology
The Wake records the distance between first notice and later knowledge. Its purpose is not to prove that Rough at Sea was early, nor to declare that lightly covered stories were suppressed. It preserves what was known, what remained uncertain, and what happened next.
Observation before hindsight
Point-zero packets are frozen historical observations. Later knowledge is appended rather than used to rewrite the beginning.
Evidence is not attention
A claim can be well supported and little noticed, or widely repeated and weakly supported. Those dimensions remain separate.
Repetition is not corroboration
Articles derived from the same paper, press release, wire story, interview, or announcement share an evidentiary lineage.
Missing beats fabricated
An unknown date, count, or metric remains empty. Precise-looking guesses do more damage than an honest gap.
The pipeline
From Soundings to an active record
- Daily capture
Morning Soundings chooses what matters today
Editorial importance comes first. Longitudinal potential is only a tiebreaker between otherwise comparable subjects. Every signal and deep sounding receives a structured sidecar packet, even when it is unlikely to become an active Wake record.
- Point zero
The initial state is preserved
The packet records a narrow claim, supported facts, uncertainty, source roles, evidentiary lineages, open questions, tracking terms, and concrete future evidence. Contemporary packets are created during publication; the earliest six were explicitly reconstructed and remain labeled as such.
- Weekly admission
A small collection is chosen deliberately
The weekly process looks for stable subjects with consequential unresolved questions, observable future evidence, and durable sources. It ordinarily admits no more than one to three new records and may admit none. Deferred candidates remain visible in run logs.
- Longitudinal review
Records are reviewed by cadence and event
A record is revisited when its date is due, a named milestone occurs, a new Soundings signal continues it, or credible new evidence appears. Not every record is researched every week. A performed review with no material change can still be valuable data.
- Append-only change
New observations do not erase old ones
Developments, disagreements, corrections, resolutions, and informed no-change reviews become dated observations. The current state mirrors the latest observation, while the full timeline preserves how the assessment evolved.
Three independent readings
Lifecycle, evidence, and attention
Lifecycle
What is happening to the subject or claim.
- Emerging
- Meaningful future evidence is expected.
- Corroborated
- Important parts gained independent support.
- Disputed
- Material evidence or interpretation remains contested.
- Implemented
- A promised program, product, policy, or design was deployed.
- Stalled
- Expected progress stopped without clear resolution.
- Contradicted
- Later evidence substantially weakened the original claim.
- Resolved
- The central open question has a defensible answer.
- Dormant
- No useful update is expected soon.
- Retracted
- The source or claim was formally withdrawn.
Evidence
What kind and breadth of support currently exist.
- Primary supported
- Durable primary evidence supports the bounded claim.
- Mixed
- Several kinds of evidence exist, with meaningful qualifications.
- First-party only
- The claim still rests on an interested source.
- Preliminary
- Early evidence exists but replication, review, or scale is missing.
- Disputed
- Qualified sources materially disagree.
- Synthesis
- The record joins several results rather than following one bounded claim.
Attention
How much reproducibly measured public coverage the subject receives.
- Unmeasured
- No defensible quantitative attention method has been applied.
- Low to high
- Reserved for a future reproducible corpus and fixed method.
- Surging or declining
- Reserved for measured change between comparable snapshots.
Ordinary search-result estimates are not used. They are unstable, personalized, incompletely deduplicated, and often count one syndicated lineage as many reports.
Evidentiary ancestry
Why source lineage matters
Ten articles can appear to corroborate a claim while all ten paraphrase one press release. The Wake assigns those sources the same lineage. Independent support requires a genuinely different route to the fact: another experiment, another dataset, direct observation, independent documentation, or reporting based on separate evidence.
Lineage does not make a source bad. A primary paper and the university article explaining it can both be useful. It prevents publication count from impersonating evidentiary diversity.
Selection limits
What The Wake does not claim
It is not exhaustive. The collection reflects what Morning Soundings found and what the weekly process could responsibly follow.
Low coverage is not proof of suppression. Sparse attention may have many causes and is not assigned a motive without evidence.
Admission is not endorsement. A record may be followed precisely because an important claim is uncertain or disputed.
Status is not a quality score. A disputed record can be more informative than a corroborated one.
A review is not a demand for novelty. “No material change” is acceptable when a real search and source check occurred.
Automation does not remove judgment. The schemas constrain what must be recorded, but evidence and significance still require accountable editorial decisions.
Corrections
The record must be allowed to embarrass us
When an earlier observation contains a factual error, it is not silently rewritten. A correction points to the affected observation, explains what changed, cites the supporting evidence, and appends a new timeline entry. Typographical repairs that do not alter meaning may be fixed normally; substantive history remains visible.
Current pilot
10 records · 22 observations · 2 run
The first run is a bootstrap admission ledger, not a retrospective claim that all seven subjects changed after publication. Attention remains unmeasured. The public interface and both downloads are generated from the same records that the weekly task updates.