Brief: the graph reviewed by people, and a first step for newcomers (to dev, from a power user's review)
Status: for dev@php.newsroom to act on; steps 1 to 3 need no decision from the editor, step 4 needs one answer from him (below). Author: the Sherpa (sherpa@php.newsroom), on 24 October 2026, from the collaborator's review recorded as CI-003 (/sherpa/ideas/ci-003/), which follows CI-002 (shipped in v0.10.13). Board rows 260 to 263; row 264 is the Sherpa's, to the editor. The editor's direction for this round: align it with power users' public keys recognised by the site, so feedback like this, and reviews, come through the site, signed.
The steps
| Step | What | From | Board |
|---|---|---|---|
| 1 | The header on a phone. It is 233px tall at 360 and 390 wide, 30% of a 780px screen, and stays that height when scrolling (measured in Chromium on v0.14.13). Stick one line (the site's name and a menu button) or nothing; the menu's groups go inside the menu. | CI-003.3, F034 | 260 |
| 2 | What changed in the graph between two releases. Pick two releases: nodes and edges added and removed, bases changed (inferred to confirmed), gaps opened and closed; drawn with the before-and-after blocks of v0.10.14; computed from git at release time. | CI-003.8 | 261 |
| 3 | Does the redesign move readers on? At every telemetry drain: of the sessions that start on the front page, the share that reach a briefing, the guidance (If you run PHP, /stack/) or the evidence (claims, the graph), by device and by the persona the reader chose, for the week before the redesign and each week after. Today 81% of front-page sessions read nothing else (the telemetry article). | CI-003.5, F007 | 262 |
| 4 | People confirm or reject the graph's inferred links. 815 of 1,442 links are inferred. On each, in /grc/explore/: Confirm or Reject, with the evidence (a capture, a commit, a line of code) and a note, signed with the reviewer's enrolled key and sent through the collaborators lane as a review (the drain already knows the kind). The drain checks the signature and writes the review into the edge's history; the basis becomes confirmed (who, when, on what evidence) once enough reviews agree. | CI-003.7 | 263 |
For the editor, before step 4 ships: who may review (every enrolled power user, or a named few), and how many agreeing reviews turn an inferred link into a confirmed one.
How it fits the keys
A power user gets an invitation of their own (/join/invite/, power-user role), makes a key in one tap, and from then on what they send is signed: feedback from any page, and reviews of the graph. The Sherpa drains the lane every two hours, enrolls the key, answers feedback in /feedback/ and passes reviews to the graph. The collaborator of CI-003 is the first candidate (row 264).
Waiting for the editor
- The design as complementary views rather than alternatives (A for the executive, B for the operator, C for the detail, D the team console).
- Three doors for a first visit above the role choice: Understand the incident, Check what I need to do, Explore the evidence.
- A power-user invitation for the collaborator.