The people below the fold

REM finally found the owner's older contacts. The People page said there were 803 of them, but offered no way to reach them from that first screen. The disclosure sat after 590 mapped People rows. A small invented notebook had made the new directory look easy to find; the private real map exposed the mistake as soon as it filled the screen with actual scale.

The two lists have different meanings. A mapped person has a page REM can investigate. An “other contact” is an address seen in mail without enough evidence to call it a person memory. Combining them into one table would blur that distinction. Hiding the second list at the end of the first made the distinction practically useless.

We moved the collapsed other-contacts section above the mapped roster. On the real 390-pixel phone view, its top moved from y=1314 to y=515; on a 576-person invented fixture it moved from y=1297 to y=441. The disclosure still leads to its own search and “Prepare in terminal” command, while the mapped roster's filters and sorting remain below. The browser check opens it with the keyboard and sees no horizontal overflow in phone or desktop layouts.

This is a reachability fix, not a claim that 803 addresses have been understood. An all-person model run and a review of what its pages actually say remain the next test of REM's promise.

Checking the published a41 wheel exposed an earlier break in the same promise: it read 15,551 mail headers before two later connection failures left the People map with only the owner. A reachable directory cannot help when its entries disappear during init. The a42 candidate therefore includes both the directory position and preservation of completed mail years. The failure analysis records what remains: durable checkpoints and a complete real investigation.