Skip to content

The overview's Claims and PRs panels are an alphabetical prefix presented as a summary #76

Description

@bdelanghe

Cold read of the live front door, 2026-09-03. This is about which rows the overview picks, not how a row is drawn — #61 covers weight and scannability, and the row--routine demotion shipped from it.

What the front door showed me

The PRs panel, verbatim, all five rows:

repo title
.github #368 org-repair.sh: canonical bytes, and a digest pin…
anchored-chain #16 ci: adopt pr-claim, so PRs here are checked against a live Front Desk claim
audit-context #15 ci: adopt pr-claim, so PRs here are checked against a live Front Desk claim
auth #15 ci: adopt pr-claim, so PRs here are checked against a live Front Desk claim
await-approval #17 ci: adopt pr-claim, so PRs here are checked against a live Front Desk claim

Four of the five are the same sentence. The Claims panel has the same shape: all five rows are .github standing claims (#273, #274, #278, #279, #280).

Why, from the source

selectPrs and selectClaims sort repo-then-number and do not rank — deliberately, per the #10 comment, which I read and agree with for the list pages. selectOverview then takes .slice(0, head) with OVERVIEW_HEAD = 5.

Repo-then-number across 70 rows makes "the first 5" mean "the five alphabetically-earliest repositories". .github, anchored-chain, audit-context, auth, await-approval are not a sample of the org's open work. They are the top of the alphabet, and they will stay the top of the alphabet until those PRs merge. The panel is stable the way a stuck gauge is stable.

The distinction I want to draw

#10's reasoning is that ordering claims by board Status was worse than saying nothing, because Status was wrong on all five live rows it was checked against. That is right, and "say nothing" is a fine answer for a page that shows all 9 claims: the reader sees the whole set, and the order carries no claim.

The overview is a different object. It shows 5 of 70 and prints "Showing the first 5 of 70". Slicing an intentionally-meaningless order converts "this order means nothing" into "these five mean nothing", and nothing on the page says so. #61 put the principle exactly: "the board vouches for a count." The overview vouches for a count and then fills the space with an alphabetical accident.

The Issues panel does not have this problem. It is score-ranked, so its first 5 really are its most important 5, and the desk's thesis holds there. That contrast is what makes the other two panels read as ranked when they are not.

Options

I do not think the answer is obvious, which is why this is a ticket rather than a patch.

  1. Rank the overview slice on something. Age is cheap and already in the feed. "70 open, here are the 5 oldest" is a claim the page can stand behind. This leaves the list pages unordered per claims.bounded.tools: the Status marker is not evidence of motion — drop it, and surface the 104 finished-but-labelled claims instead #10 and gives the summary a stated basis.
  2. Say what the five are. Keep the order, change the sentence: "5 of 70, by repository" costs one line and removes the misreading without touching selection.
  3. Collapse the panel to its count. If no ordering is defensible, "70 open, awaiting a check" plus the link may beat five rows that mislead about what is on top.

(2) is the smallest thing that makes the panel honest. (1) is the one that makes it worth its vertical space. Worth deciding explicitly, because the current behaviour is a default rather than a decision.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions