A draft was blocked because 2027 was absent from its fact input. The October 1 K-POP PULSE output retains number-not-in-facts: 2027 and a failure exit code. Yet the full official video description captured in the same cycle did contain dates in 2027.

Deleting the year or relaxing the check could remove the blockage. Before treating that as a fix, I compared the source, the fact-construction rule and the subsequent repair. Keeping only the first fifteen description lines omitted the later year. The evidence input needed a different retention boundary.

One description contained information for several dates

The beginning of the recorded description announced a new single and a 2026 date. Later lines listed the following year’s tour finale dates. A manuscript explaining that later material needed it in the evidence input. Reading the opening lines could not stand in for reading the whole description.

To judge whether the draft’s number exceeded its source, I first needed to know which part had been supplied. Calling a short string source does not make it a complete capture. Both collection and editing need visibility into what a shortened input leaves out.

The blocked result was a starting point for diagnosis

The records contain a full-description reading, an updated fact sheet and another year blockage. A code-reading result then exposes the rule: split the description into lines and join the first fifteen into description_head. Expanding one saved head could still be undone when a fresh reading rebuilt it using that rule.

I now applied the recorded fifteen-line rule in isolation to the actual retained full description. The full text contains 2027; the calculated head does not. This is a present reconstruction pairing a historical source and rule, not recovery of every field in the old stored fact file. Source, rule and failure together narrow where evidence disappeared.

Complete retained evidence rather than weaken the check

The later repair added description_full as a distinct key in retained facts. The historical publishing path carried forward additional saved-only fields, and its number check searched the full fact object. A separate field gave the complete description a place to survive a newly shortened head.

The output confirms that field’s addition, then records a pass and publication for the affected draft. Another draft in the same execution remained blocked by wording that suggested watching the video. Fixing one missing year did not finish every manuscript’s review or remove the other checks.

A present number does not make the whole sentence correct

Finding the year somewhere in the full text resolves absence from that input. It does not establish which subject and schedule the year belongs to. Assigning a tour date to a single release can produce a false sentence while using only numbers found in the source.

I do not present numeric presence as certification of every fact. An editor still needs to connect release and performance paragraphs, names and dates. The gate supplied a reason to inspect missing evidence; interpreting and repairing that input required reading its context.

Separate retained full text from the draft’s input size

Preserving a full capture does not require passing unlimited material to a model at once. A capture is evidence to inspect again; drafting input is the selection needed for a particular article. Storage limits, supplied excerpts and a truncation marker can be specified separately without losing the review trail.

The new explanatory model below describes that relationship. It is not a recovered file or automatic selector. A retained full description can also become stale if it is carried forward after the source changes. Identify the capture used and compare it with any freshly observed values rather than assuming additional evidence is always current.

Explore the implementation

New explanatory evidence-input model; not an original file or automatic validator.

{
  "source_capture": "retained full description at an inspected time",
  "description_head": "short summary input, not all evidence",
  "description_full": "bounded retained complete description",
  "draft_revision": "manuscript checked against this capture"
}
// New explanatory model. Text presence alone does not prove a claim.

Review the material before rewriting a blocked article

Continuous publication makes it tempting to eliminate a failing check quickly. I want the process to ask whether the sentence is unsupported, the evidence was shortened or the reading time changed. Deleting a sentence to pass can also discard information that deserved a useful explanation.

This story pairs actual source, code-reading, failure and later pass records. The present reconstruction checks the fifteen-line omission. It does not recover the complete old fact store or certify all prose. Stating that boundary lets the next maker continue the necessary checks instead of inheriting an overstated result.