Watchtower Working Standard The operational layer of the Watchtower family Route first. Archive later.

The Watchtower Framework is too high a bar for modern general journalism as currently produced and distributed. It is not obviously too high a bar for what a citizen should require before treating a story as archived reality.

On the other hand, modern Scientific Publications can withstand the Watchtower For Scientific publications because media Domain 4 polices the megaphone; scientific Domain 4 polices the lab’s right to be wrong in public. 

The best way to use the original 

 Watchtower  to audit the media, is as a red-flag and routing test, not as a claim that only 90+ objects are news. 

For that reason we are releasing the Watchtower Working Standard (WWS)

View the WWS is a megaphone patch. A paper that is already a methods document does not need it.

WATCHTOWER ROUTING RUBRIC

An Operational Adaptation of the Generalized Media Watchtower Framework for Modern Journalism

Version 1.0  ·  1 September 2026  ·  Article Integrity scored separately from Platform Effects

This rubric keeps the original Watchtower purpose — separate engineered packaging from verifiable reality — but changes how the score is used. It is a red-flag and routing test. It is not a claim that only a 90-point object counts as news, and it does not treat a weak package as proof that the underlying event is false.

1. What changed, and what did not

Preserved from the original

•        Four-domain architecture and the original criterion names, so audits remain comparable.

•        Refusal of surface partisan scoring. Party valence is not a criterion.

•        Preference for primary records, named sources, defined terms, and symmetric scrutiny.

•        Suspicion of middleman narrative capture (litigants, NGOs, think tanks, anonymous officials, house experts).

Changed for operational use

•        Two-layer scoring. Article Integrity (Domains 1–3) routes the reader. Platform & Agency (Domain 4) is a consumption warning, not a veto of the article.

•        No whole-piece pass/fail kill switches. Undefined labels and engagement packaging are scored and flagged. They no longer zero an otherwise sourced report.

•        Recalibrated bands. 70–89 becomes Usable First Look. 90–100 is reserved for archive-grade packets. A new 50–69 band catches high-flag stories that still contain a real lead.

•        Hard gates only for band entry. A story cannot enter Usable First Look or Archive Grade unless a short list of operational conditions is met.

•        The event and the package are separated. “Do not use as record” means do not treat this write-up as the dossier. It does not mean the lawsuit, vote, strike, or filing did not happen.

2. Two-layer scoring

Layer A — Article Integrity Score (AIS), 0–100

AIS is the routing number. It is the equally weighted mean of Domains 1, 2, and 3. Score the text the reader can actually inspect: headline, dek, body, captions, links, and update note. If the social card is what most people will see, score the card as a separate unit under Layer B; do not let the card replace the article score.

Layer B — Platform & Agency Score (PAS), 0–100

PAS is reported beside AIS, never averaged into it. A strong article on a rage-optimized card should read as “AIS 78 / PAS 24,” not as a single demolished 40. A weak article on a quiet document page should read as “AIS 41 / PAS 70.”

Layer	What it measures	How you use it
AIS (1–3)	Structure, sourcing, verification	Routes: archive / first look / cross-examine / do not use as record
PAS (4)	Packaging, house narrative, reader challenge path	Warns: read in full / distrust the card / assume capture / assume closed loop

3. Routing bands (AIS only)

AIS	Band	Routing action
90–100	Archive Grade	Consume and keep. The text plus its links are a durable record. Still check primaries, but the package itself is safe to file.
70–89	Usable First Look	Read it. Treat it as a dated first look, not the last word. Immediately open the linked documents and the other side’s stated ground.
50–69	Cross-Examine	The event may be real; the package is not trustworthy as a map. Extract names, dates, and docket numbers only. Rebuild the story from records.
0–49	Do Not Use as Record	Reject the package. Do not quote its frame, labels, or implied causality. If the event matters, start from the primary record as if the article did not exist.

Original Watchtower bands mapped as follows: 90–100 remains Sovereign / Archive Grade. Original 70–89 “Reject & Cross-Examine” is split: 70–89 is now first-look; 50–69 carries the old cross-examine duty. Original 0–69 demolition is narrowed to 0–49 for the package only.

4. Five-minute red-flag scan

Run this before scoring. Flags do not assign points. They decide how suspicious you should be when you score.

ID	Red flag	Trigger
R1	Undefined load-bearing label	Headline or lead uses far-right, extremist, purge, censorship, misinformation, crisis, experts agree, without a definition, metric, or attributed quote.
R2	Anonymous insulation on the decisive fact	The fact that makes the story true or false rests on officials familiar with, sources said, or an unnamed senior official.
R3	Missing adverse ground	A named institution acted. Its stated reason is public or obtainable, and the story omits it or replaces it with motive language.
R4	Scope leap	One firing, one poll, one clinic, one ship becomes a structural era without a time series or a stated limit.
R5	Middleman supplies the moral	A litigant, NGO, think tank, or house expert writes the conclusion; raw records are absent or ornamental.
R6	Card is the product	The distributed object is a headline + photo + compressed lead. Controlling facts live only below the fold or not at all.
R7	No path to a primary record	No docket, PDF, transcript, dataset, on-the-record quote, or official notice the reader can open.

Flag count → posture

•        0 flags: score normally.

•        1–2 flags: score, but cap AIS at 89 unless the flags are confined to packaging and the body is clean.

•        3–4 flags: default band is Cross-Examine or below. AIS may rise only if linked primaries independently carry the facts.

•        5–7 flags: default band is Do Not Use as Record. Extract dates and names only.

5. Domain 1 — Structural Architecture & Semantic Sincerity

Weight inside AIS: one-third. Original Watchtower weight inside the four-domain system was 25%.

1.1 Definition Bounds — 30% of Domain 1

Requirement: Every load-bearing label in the headline, dek, or causal spine has an explicit bound, a metric, or is quarantined as someone’s attributed claim.

Scoring guide: 90–100 labels defined or unused. 70–89 loaded words appear only inside quotes or are immediately bounded. 40–69 undefined labels appear in body but not as the headline’s payload. 0–39 the headline or lead’s meaning depends on an undefined epithet.

Adaptation note: This is no longer a pass/fail on the whole article. An R1 flag is still raised.

1.2 Structural Complexity Index — 35% of Domain 1

Requirement: A reader can index who / what / when / where / according to whom without climbing nested historical or moral frames.

Scoring guide: 90–100 chronological and flat. 70–89 one contextual paragraph that does not smuggle causality. 40–69 several nested frames. 0–39 the event is buried inside a thesis about an era, a movement, or a villain.

1.3 Scope Creep Insulation — 35% of Domain 1

Requirement: Isolated facts stay isolated unless the text supplies a series, a rate, or a hard limit.

Scoring guide: 90–100 explicit boundary (“this filing; not a finding on the merits”). 70–89 no leap, even if context is present. 40–69 suggestive “latest in a streak / part of a broader assault” language. 0–39 the isolated incident is treated as proof of a system.

6. Domain 2 — Information Routing & Middleman Insulation

Weight inside AIS: one-third.

2.1 Sourcing Integrity & Proximity — 40% of Domain 2

Requirement: The facts that decide the story come from named speakers, signed documents, or unredacted public records the reader can reach.

Scoring guide: 90–100 decisive facts are on-record or documentary. 70–89 named sources carry the spine; anonymous color is labeled and non-decisive. 40–69 mixed, with at least one decisive anonymous claim. 0–39 the story collapses if the unnamed official is removed.

First-look concession: “Declined to comment” counts as an on-record institutional act. It does not replace the institution’s last stated ground if that ground is already public.

2.2 Middleman Narrative Insulation — 40% of Domain 2

Requirement: Raw data and third-party interpretation are visibly separated. Lawyers, NGOs, think tanks, and staff experts may be quoted; they may not own the lede’s conclusion.

Scoring guide: 90–100 interpretation is labeled and optional. 70–89 quotes are heavy but the reporter states the docket facts first. 40–69 the moral arrives from a middleman before the record. 0–39 the middleman writes the story.

2.3 Uncertainty & Correction Path — 20% of Domain 2

Original 2.3 scored the outlet’s clawback machinery. That outlet test now lives in PAS. Inside AIS, score only what this article does.

Requirement: The piece marks what is alleged, what is established, and how a reader will know if the story moves.

Scoring guide: 90–100 allegations tagged; update hook present. 70–89 complaint vs. fact is distinguished in plain language. 40–69 some hedging, no update path. 0–39 contested claims written as completed facts.

7. Domain 3 — Metric Verification & Accountability Controls

Weight inside AIS: one-third.

3.1 Primary Dossier Standard — 40% of Domain 3

Requirement: Verification rests on records a reader can inspect, scaled to story type.

Scaled credit (this is the main operational concession):

•        Full credit: controlling statute, notice, transcript, dataset, or filing excerpted or linked and described.

•        Substantial credit: working public link to the complaint, order, release, or dataset.

•        Partial credit: precise identifiers (case number, bill number, date-time-place) that make the record findable.

•        No credit: paraphrase of “documents obtained by” with no handle.

Scoring guide: 90–100 controlling records in the piece. 70–89 findable primary plus accurate paraphrase. 40–69 identifiers only. 0–39 no dossier path.

3.2 Asymmetric Narrative Firewall — 40% of Domain 3

Requirement: The same tests are applied to the powerful institution and to the sympathetic actor.

Minimum first-look test: If Party A acted and Party B sued, both of these are on the page — A’s stated ground and B’s stated ground. Motive language cannot substitute for either.

Scoring guide: 90–100 reader can steel-man both sides from the text. 70–89 both grounds present, some tonal tilt. 40–69 one side is a mind-reader exercise. 0–39 one side is a villain class, the other a victim class.

3.3 Update Trigger — 20% of Domain 3

Original demand was an automated sunset date. That remains the archive-grade rule. First-look rule is weaker and usable.

First-look (enough for 70–89 on this sub-criterion): name the next fact that would change the piece (answer brief, TRO ruling, after-action report, audit).

Archive-grade (required for 90–100 on this sub-criterion): a dated review or expiration note.

Scoring guide: 90–100 dated trigger. 70–89 next-fact trigger. 40–69 “developing” with no hook. 0–39 a still-open matter written as a settled morality play.

8. Layer B — Platform & Agency Score (PAS)

PAS does not change the AIS band. Record it as a warning label.

4.1 Engagement Packaging — 40% of PAS

Score the unit most people will see. A calm 1,200-word piece with a panic card gets a low 4.1. Original 4.1 pass/fail is retired as a whole-article veto.

Scoring guide: 90–100 headline is a factual sentence; no rage bait. 70–89 ordinary news headline. 40–69 loaded card, calmer body. 0–39 the card is a moral advertisement.

4.2 House-Narrative Autonomy — 30% of PAS

Requirement: This story could have been written if it cut against the outlet’s running plot. Undisclosed institutional conflict counts against the score (for example, the outlet is itself litigating the same agency).

Scoring guide: 90–100 local or independent desk, conflict disclosed. 70–89 house line present but facts first. 40–69 story is an episode of an existing series. 0–39 dissenting facts would not have shipped.

4.3 Reader Challenge Path — 30% of PAS

Requirement: A serious factual challenge has a human path — corrections desk, public editor, or an update the outlet will actually attach.

Scoring guide: 90–100 visible correction mechanism tied to the piece. 70–89 a real corrections page exists and is used. 40–69 form-mail only. 0–39 comments and challenges are algorithmically disposed of, or impossible.

PAS warning labels

PAS	Warning printed next to AIS
80–100	Low packaging risk. Read the piece as written.
60–79	Standard modern distribution. Do not let the card substitute for the body.
40–59	House frame or engagement card is doing real work. Open the links first.
0–39	Treat the distributed object as advocacy packaging. AIS must be earned by the body and the records, not the wrap.

9. Hard gates

A high average cannot buy a band the piece did not earn.

Gates for Usable First Look (AIS 70–89)

All four must be true. If any fails, cap AIS at 69.

1.     Named actor or linked primary on every decisive factual claim.

2.     Adverse party’s stated ground is on the page, or the story documents a refusal to state one.

3.     No undefined epithet is doing load-bearing causal work in the headline or lead. Quoted epithets are allowed if attributed.

4.     A public competing account is not omitted when the reporter had it or could have had it from the same day’s record.

Gates for Archive Grade (AIS 90–100)

All First Look gates, plus all four of these. If any fails, cap AIS at 89.

1.     Controlling documents are linked or excerpted and described, not merely alluded to.

2.     A reader can steel-man both sides from the text alone.

3.     An explicit update or sunset condition is stated.

4.     The scored unit is the article (or a document packet), not a social card.

10. How to score in practice

1.     Identify the unit. Article body and social card are different objects if they diverge.

2.     Run the seven red flags. Record R1–R7 as Y/N.

3.     Score each Domain 1–3 sub-criterion 0–100. Apply the weights. Average the three domains. That is raw AIS.

4.     Apply hard gates. Cap if needed. That is final AIS and the routing band.

5.     Score PAS separately. Print it as a warning, not a blended grade.

6.     Write one routing sentence: what to open next (docket, statute, transcript, dataset).

Genre expectations (do not punish a wire for not being a book)

•        Spot news / wire: can reach 70–89 with named actors, both grounds, and a link or case number. 90 is rare and should be rare.

•        Investigation: 90 is available if the dossier is published with the story.

•        Live politics brief: usually 50–79. Score the omissions, do not demand a methods appendix.

•        Opinion / analysis: score as analysis. It cannot enter Archive Grade unless it is a documented brief. Mark the genre on the scoresheet.

•        Social card: score PAS 4.1 against the card. Never let a card enter Archive Grade.

11. Worked example

Object: The New York Times, “Fired Journalists From Stars and Stripes Sue Pentagon,” Minho Kim, 27 August 2026.

Unit scored: published article. Social card scored only inside PAS 4.1.

Red flags

R1 Y (censorship / retaliation / independence used as unbounded frame). R2 N on the filing itself. R3 Y (Pentagon’s public insubordination / CBS-interview ground thin or absent). R4 Y (“streak of public pushbacks,” policies that “severely limit”). R5 Y (complaint language supplies the moral). R6 Y if the card is the distributed object. R7 N (complaint exists on the public docket).  Flag count: 5. Posture: default Do Not Use as Record unless primaries independently carry the facts. Here the filing is real, so extract the docket and rebuild; do not keep the frame.

AIS (indicative)

Sub-criterion	Score	Note
1.1 Definition Bounds	35	Independence, retaliation, censorship unbounded
1.2 Complexity	55	Readable, but nested war / history / spokesman frames
1.3 Scope creep	30	Three firings become a censorship era
2.1 Sourcing	58	Filing + decline-to-comment; adverse ground thin
2.2 Middleman insulation	32	Complaint owns the conclusion
2.3 Uncertainty path	40	Some “accused / claiming”; no update hook
3.1 Dossier	62	Public complaint; controlling directives not shown
3.2 Symmetric scrutiny	28	Pentagon motived; plaintiffs untested
3.3 Update trigger	20	Live case, no next-fact stamp

Domain 1 ≈ 40.  Domain 2 ≈ 45.  Domain 3 ≈ 39.  Raw AIS ≈ 41.

Hard gates: fails First Look gates 2 and 3 (adverse ground; load-bearing labels). Final AIS capped at 69 in any case; raw score already sits in Do Not Use as Record / high Cross-Examine.

Final call: AIS 41 — Do Not Use as Record. PAS ≈ 28 — treat the wrap as advocacy packaging. Routing sentence: open the complaint, the termination notices, the 5 July CBS interview, and the DoD instruction that governs Stars and Stripes. The lawsuit happened. This package is not the record of why.

Contrast: a same-day AP write-up that names the three plaintiffs and states the insubordination / CBS ground would likely clear more of Domain 2 and 3.2, land nearer AIS 55–65 (Cross-Examine), and still miss First Look if labels stay unbounded and no update hook appears. That is the intended behavior: two stories about one filing can be ranked without pretending either is archive-grade.

12. One-page scoresheet

Copy this block for each audit.

Outlet / headline / date / URL:

Unit scored (article / card / both):

Genre (wire / investigation / brief / opinion):

Red flags R1–R7 (Y/N) and count:

D1 (1.1 / 1.2 / 1.3) → domain score:

D2 (2.1 / 2.2 / 2.3) → domain score:

D3 (3.1 / 3.2 / 3.3) → domain score:

Raw AIS / gates applied / Final AIS / band:

PAS (4.1 / 4.2 / 4.3) / warning label:

Routing sentence (what to open next):

13. Relation to the original Watchtower

This document is an adaptation, not a replacement. Use the original Generalized Media Watchtower Framework when the question is: “Does this object meet Sovereign Standard as CotoBuzz defined it?” Use this Routing Rubric when the question is: “What should a reader do with this piece of modern journalism today?”

The original remains the stricter archive test. This rubric exists so that a 31 and a 74 are no longer both filed under demolition, and so that a platform sin cannot erase a clean wire.

Core Watchtower commitments are unchanged: no partisan scoring, no credit for emotional sincerity, no treating institutional consensus as evidence, and no archiving a closed narrative loop as if it were a record.

End of Version 1.0