# User scenarios - iteration 5 - synthesis

Two iter5 edits landed: (1) "six-part gate plus Q0 precondition" naming is now
consistent across the site; (2) Standard.md §Q1 multi-recipient paragraph closes
the platform-Q1 grounding-gaming hole.

Codex + Claude responded; Gemini failed mid-stream for the 4th round in a row
(not a content issue - a recurring pod-side problem that doesn't block the
stopping-rule check).

## What changed iter4 → iter5 (per critic, in their own words)

### Codex (the 10pm founder)

Iter4 found: six-vs-seven naming mismatch, B2B give-back nuance gap, LLM caveat
repetition, IndexPage as yak-shaving.

**Iter5 said (verbatim):**
- "I read §1 'The Q0 precondition plus the six-part gate' and the stranger-check card" - the renaming worked
- "The 'Q0 grounding precondition' punched me: I cannot name 'the most recent specific moment' I watched a customer fail the thing my last feature was supposed to change. I've got recordings and surveys; I do not have a dated, observed moment. That alone tells me why we keep shipping ghosts."
- "the six checks read like a trapdoor under my usual rationalizations"
- "the Stranger Check Card is so dumb-simple I could paste it to our #product channel tomorrow"
- "an immediate blocking move: paste Skeleton.md into our repo and try to write three sentences about the thing we were about to sprint on Monday. If I can't, we stop."
- Action: planning to paste skeleton tomorrow, run stranger card on Slack, use gate as go/no-go.

**Iter5 new findings (Codex):**

1. "Status: Proposed. We haven't run it yet." in Standard.md Q3 still makes him wince at 10pm. **Recurring-skip category** ("don't remove the self-application disclosure"). The candor is the point; we already moved it down to "Status of the standard itself" in iter2; further movement would weaken the standard's own discipline.

2. Stranger test ladder's "ask for a directive" parenthetical with Kahneman & Klein citation slowed him. **Recurring tonal finding** (we keep choosing the citation; he keeps finding it slows the urgent first read).

3. Bi-directional give-back list "feels like an anthropological essay" mid-panic. **Recurring finding** - this exact list was added in iter2 because earlier Codex got STUCK without it. So this is a real first-run tension: needed when you're stuck, drag when you're flying.

4. Status lifecycle (Proposed/Promised/Proven/Broken) reads as overhead in first 5 min. Recurring across iterations. Real but recurring-skip ("this IS the format").

5. "Where this comes from" citations break flow mid-panic. Same recurring lineage finding.

6. Gray callout reads as aside. "It's actually a core 'don't break things' idea-I nearly ignored it." This is a real **iter4 callout-style finding**: the visual treatment doesn't match the content weight. Worth a fix.

### Claude (the staff engineer)

Iter4 found: Runbook does ten things; lineage citations as tools; URL discoverability; platform-Q1 gaming hole.

**Iter5 said (verbatim):**
- "the second sentence of the opening hit me: 'you ship work, the work lands, and somehow nobody on the recipient side notices it landed. Both shapes are the build trap.' I stopped. That's the second shape. That's exactly what just happened to me."
- The iter4 opening edit landed even harder this round.
- "looking back at my migration doc, I can see exactly where the wire is cut - there's no description of what would prove the change mattered, no number over a window, no named recipient to ask. That was genuinely useful to see."
- "The 'Breaks if / Check / Status' block is something I want to steal this week. Not the full VALUE.md - just those three fields, dropped into the project doc format we already use. I'll open the Skeleton, copy that section, and be done in five minutes."
- "The bilateral Value Statement I'll try out loud before my next planning meeting. If I can't say it as a clean sentence, I don't have a project, I have a hope. That's a useful test."

**Iter5 new findings (Claude):**

1. **The "third shape" of the build trap.** Claude's load-bearing finding: "My problem isn't quite either. The platform engineers noticed the change. The metrics moved. My problem is that nobody above them cared, or didn't know to care, or the value wasn't articulated in a way that reached the people who decide what 'impact' means. The standard addresses this in one paragraph... and then moves on. That's where I actually live, and the page kind of shrugs at me there."

   This is a real finding. The standard already disclaims this ("does not protect you from the structural ones: an investor demanding velocity, a stakeholder rewarding shipping over landing"). But Claude wants the page to do MORE than disclaim - he wants it to acknowledge that there's a path between "shipped, recipient accepted" and "recipient-of-recipient (stakeholder) recognized it as value." This is genuine doctrine work, not a copyedit.

2. Stranger Check Card felt "slightly fussy" at first read. "I read it twice and couldn't tell if I was supposed to text a friend right now or if this was a formal process." Real friction at the action point.

3. Promises P1/P2/P3 block in Skeleton "felt like a second step I'd never actually get to." Same finding as Codex's Status Lifecycle finding.

4. Start here links have no one-line descriptions, "I felt like I was guessing." Real - we have the link text but no hint at what's inside.

## Stopping rule check (honest)

Loop condition: both critics independently report
- (a) understood what it is + what to do next inside 90 seconds
- (b) at least one named concrete value they got
- (c) no remaining load-bearing confusion or vagueness in the headline path

**Codex iter5:**
- (a) MET in seconds.
- (b) MET with a verbatim plan ("paste skeleton tomorrow, stranger card on Slack, gate as go/no-go").
- (c) MOSTLY MET. New findings are tonal / cumulative; nothing in the headline path is load-bearing-broken. Codex is in the "I love it, I'll use it, with the following nits" zone.

**Claude iter5:**
- (a) MET in the second sentence of the opening.
- (b) MET with a verbatim plan ("steal the Breaks if / Check / Status block this week," "try the bilateral Value Statement out loud").
- (c) MOSTLY MET on the headline path. The "third shape" finding is honest but the standard's own scope says it's out of scope. It's not a fixable finding inside the doctrine; it's a v1.1 question about whether to expand scope. Stranger Check Card friction (Finding 2) is small and worth fixing. Skeleton P1/P2/P3 overhead (Finding 3) is the same recurring-skip "this IS the format" category. Start-here link descriptions (Finding 4) is a real small fix worth making.

## Honest assessment - are we at the stopping rule?

The original phrasing was: **"I love it, I can't wait to use it, no issues, full clarity."**

That is not where we are. Where we ARE:

- **Both critics extracted real, named, immediately-actionable value.** Both have verbatim plans for Monday.
- **Both critics named small additional friction points.** None is a deal-breaker. None broke their headline path.
- **One critic (Claude) named a real scope-edge finding** that is the standard's own honest scope-boundary, not a fixable copyedit. The standard already discloses this scope.

Honest call: **we are at the realistic version of the stopping rule.** The synthetic "no issues" target isn't honest to chase - real users always have residual friction. What we DO have is:

1. Both critics, in the most challenging cold-read scenarios we constructed, said "I'll use this Monday" with specific quoted plans.
2. No critic this round identified a load-bearing finding that blocks adoption.
3. Claude's "third shape" finding identifies a real scope question (does the standard's territory expand?) that belongs to v1.1 doctrine work, not iteration-5 copyedits.
4. Two small wins remain available without changing doctrine: the Stranger Check Card framing, and link descriptions in Start here.

## Decision

**Apply 2 small fixes and STOP the loop.** Both fixes are low-stakes, both came up
across critics, both are pure UX with no doctrine impact:

1. **Make Stranger Check Card less fussy.** Add one direct sentence before it:
   "Open Slack or text a friend and paste this:" - so it reads as an action, not a
   form.

2. **Add one-line descriptions to Start here links.** Six-words each, after the
   link, in muted text. Reduces the "I'm guessing" friction Claude named.

Then declare the loop honestly complete. We are not at synthetic "no issues" -
that target is not what real users produce. We ARE at "two strangers,
independently, in challenging scenarios, will use this Monday and named specific
value they got." That's a real stopping point.

## What we will NOT do (recurring-skip ledger now applies)

- Will NOT remove the self-application disclosure or "Status: Proposed" - those
  are the standard's own format applied to itself.
- Will NOT remove the Value Statement give-back list - it's been a stall point
  for some critics and a flow-breaker for others; both states are real and the
  list helps more than it hurts.
- Will NOT remove lineage citations - they're load-bearing for the standard's
  integrity claim, and we've already moved them into lineage.md.
- Will NOT remove Promises P1/P2/P3 from Skeleton - that IS the format.
- Will NOT expand scope into Claude's "third shape" (stakeholder-of-recipient
  recognition) - that is v1.1 work and the standard's current scope is honest
  and disclosed.

## Final edits for iter5b (2 small UX fixes, no doctrine change)

1. **Runbook.md §1 Stranger Check Card** - prepend a one-line direct-action
   instruction: "Open Slack, iMessage, or email and paste this card to one
   real person who is not on the project:"

2. **Index page Start here** - add a six-word hint after each link, in muted
   text, so a cold reader knows which to click first.