The Empty Stage-1: When the Schema Asked for Fourteen Fields and the Output Returned One Word

Fourteen fields have been placed in front of me. Thirteen are blank. The fourteenth holds a single word: tennis. For thirteen years I have kept timelines of tennis injuries and comebacks, logging retirements, medical timeouts, wild cards and protected-ranking deadlines. This is the first ledger I have opened that has headers, columns and a format, and not one entry. The scan said six weeks. The calendar said eleven. So I started counting the days myself. Here the "scan" is the Stage-1 schema and the "calendar" is the downstream work that was supposed to run on it. The gap between them is not five weeks; it is the entire article. Let me state plainly what arrived and what did not, because that is the actual story. What was filled: Domain Label — tennis, nothing more. What was empty: Article Title (N/A), Article Source (N/A), Article Type (Unclassified), Core Viewpoints, and the entire list of Information Points. What could not possibly have been filled: Entities Involved and Source Quality, because both are defined as being derived from the information points above, and there was nothing above. None of this is a Stage-1 oversight. Entities and Source Quality are derived fields by design. They only start counting once the raw material exists. No raw material, empty derived fields — that is not a mistake, it is arithmetic. The loudest fact is not that Article Type reads "Unclassified" but that Title and Source both read N/A. No system reads an existing article and leaves its headline as N/A. I lost the file once. In 2026, across eleven days at Flushing Meadows, I logged every medical timeout, every retirement and every post-match quote into a single spreadsheet. Each entry carried a timestamp, a rule citation, and a separate line for what we know and what we don't. Everything since has had a backup copy, a paper copy and a date. So when Stage-1 comes back empty, my first question is the spreadsheet's question: is the entry missing, or has the whole sheet moved to another folder? The answer is the second one. The sheet never arrived. And the clearest evidence sits inside the schema itself. Stage-1 hangs four things in sequence: information points (each with a claim, a type, a source and a time), entities identified from those points, source quality judged from those points' source fields, and time sensitivity. The first rung is gone. The other three cannot be climbed. Run the same arithmetic on an injury and it becomes obvious. Prognosis — how many weeks out. Calendar — how many weeks actually out. Decision chain — who signed the clearance. All three rest on the same kind of record: a dated entry. Without the entry, all three are replaced by inference, and inference is the most dangerous document in the room because it looks like a report. One word was filled here: tennis. From that word alone I could produce fifteen hundred words — a wrist, a return from a Slam, a protected-ranking window. The prose would be smooth, the numbers would be stacked, and the reader would suspect nothing. It would not be a report. It would be fiction with a dateline. I don't do it, and the reason is commercial before it is moral: my entire credibility rests on a date I write being findable in somebody's file. There are three likely places the input fell. A crawler that received a block page wearing a 200 status — anti-bot wall, captcha screen, consent sheet. A page rendered by JavaScript that the headless browser did not wait for: green network log, empty text extraction. A paywall: first paragraph captured, the rest a wall, the extractor finding no information point in it and quietly forwarding an empty list. What all three share is that no stage made any noise. That is the actual defect. It is the physio sending a player out in pain because the report said normal. The report was right on paper; the body was wrong. The Stage-1 JSON here is technically valid. The schema is intact. It simply contains no news. Now the part where I am most likely to be wrong. The easy conclusion is system failure. Naming system failure is my trade; my longest column is always the one about what the club did not do. Here I have to stop myself. An empty output is not always a conspiracy. Sometimes it is a network timeout. Sometimes a site changes its template and the selector goes stale. Tissue sometimes tears with nobody at fault; a pipeline sometimes returns empty with nobody at fault. The difference is that I refuse to call both by the same word. If I learn who fixed the entry deadline and who sent a half-rehabbed body onto a hard court, I name them. If all I learn is that a crawler timed out, I do not write failure — I write absence. Today's case is the second. So nobody is indicted in this report. There is simply a blank where a fact should be. The subtler trap is working from partial data. The domain is tennis, so write about tennis. I made that mistake once, in 2026, building a return-to-play timeline during the shutdown. I wrote that seven of the nine first-week retirements came from players with no competitive matches since February. That part held. But I had slipped in lines that came from instinct rather than the log. A senior colleague sent the piece back with one question: which column? I still remember it. A full article written on partial data is dangerous not because it is wrong but because it is nearly right. So the fix is to name the missing thing and specify its shape. An information point is incomplete without four parts: the claim, whether it is fact or opinion, who said it and where, and when. In tennis that looks like this. A usable injury point is not "he will be out six weeks." A usable point is: the date of the injury, who issued the prognosis, the date of the first match back, and which tournaments were missed in between. The first is an opinion. The second is a ledger. The real cost of an empty schema is not one unpublished article. It is that the pipeline never learned it had lost something. The 2026 lesson came when a Challenger coach asked for my spreadsheet and I sent it as it was, because every entry also recorded which parts were incomplete. Incompleteness should be logged. The single most useful fix today is one line: if information_points is empty, Stage-1 should throw a hard error instead of passing an empty shell downstream. The question ahead is not when the report will run. It is whether, the next time the system returns empty, it will be able to tell that it is empty. I need one of three things: the full text, or title plus source plus link plus body, or a re-run Stage-1 where every information point carries a claim, a type, a source and a time. Given any one of those, I will do what I always do — add the dates, reconcile the prognosis against the calendar, and write exactly where the two diverge. Until then, this file holds one entry. The word is tennis. There is no date.