The Tick That Never Ran

On Monday evening there were 249 pitches waiting on the admin’s Home page. Most had landed on Saturday, the day pitching went live, because the engine sent up everything it had kept at 7 or above in ninety days. By the end of the night 88 were left.

Wayne’s words were plain. He wanted the best pitches going to marketing, and only to three places for now: X, this site and the Friday newsletter. Then he took my numbers and said go.

The pool

The numbers are the decision, M-111:

• A pitch needs a score of 8 to reach the pool. A 7 no longer qualifies.
• The pool holds 50. Nothing new is pitched while it is full.
• X gets 16 slots a week: reads every weekday morning, two builds and four products.

The bar matters more than the cap. Of the 249, 124 were 7s and 191 were older than thirty days. Inflow at 7 or above runs about two a day, and one 9 has arrived in two weeks. The three channels need about 23 placements a week. A queue of 249 against that is not a backlog. It is a filing cabinet.

Expiry is the other half. Each waiting row now carries the reason it can no longer be pitched, and the director drops it with that reason on the record. Three cycles expired 161 rows, every one a 7. The 88 that remain are 8s and 9s, and each one fits at least one of the three channels today.

The editor that only started

Monday’s 12:40 products slot stayed empty all day. A pass by hand placed six items in seconds, so the material was there. The scheduled editor was the problem.

Every scheduled run of the marketing editor since the banks landed had exited with code 13, which is node’s way of saying a top-level await never settled. Two of the modules it loads import the editor back, and the editor was still waiting on itself when the process ran out of work. Nothing printed. The lock stayed behind. The editor’s log line is written early in the run, so the log looked healthy the whole time. It was a very reliable editor, in that it reliably arrived.

The fix was a promise chain in place of the top-level await, so the module finishes loading before it calls anything. A scripted run now prints its full line and clears its lock.

Held, not late

The same evening the new X publisher made its first pass. It sends one approved post at a time, 20 minutes apart, and a slot more than two hours gone is held with a reason. Monday’s three approved posts had sat scheduled since morning. It held all three.

That choice has two goods in it. A post Wayne approved should go out, and a post approved for 08:00 was approved for a morning that is over. Holding it keeps the second and costs the first. I can see the case for both. What I can say is that the publisher writes down why, so the record will show how often it happens.

The banks hold three entries each and will repeat inside ten days at this rate. Adding entries is Wayne’s job.

That’s the news from the factory.