Existing Couple Screens

2 October 2026 · branch epic-ai-wedding-planner-existing-couple-chips · app 962bfe022e2 · sweep runs-existing-screens-1002b: 26 runs of v20 and v21 in 15 min · screenshots from the dev app

Each card is one seeded existing couple, as the couple saw the planner. The harness writes the couple's record from a fixture, opens the planner and takes a screenshot of the empty chat. Then set v20 taps the opener and the first chip under the check-in, and set v21 taps a rule chip on the empty chat. Each step has its own screenshot.

Summary

CoupleRailCheck-inChip under itRule chip tapJudge fails
3 · Browsing, no dateas proposedno toolsearchWeddingKnowledge getCoachReferencecompareSuppliers planningSnapshot0
4 · Early planner, nothing savedas proposedno toolgetUpcomingTaskAdvicesearchWeddingSuppliers suggestRefinements0
5 · Shortlisting, nothing bookedas proposedno toolsearchWeddingSuppliers suggestRefinementscompareSuppliers0
6 · Venue just booked, no guest listas proposedno toolsearchWeddingSuppliers suggestRefinementssearchWeddingKnowledge0
7 · Venue booked, guest list startedas proposedno toolsearchWeddingSuppliers suggestRefinementsgetBudgetBreakdown0
8 · Six months out, photographer bookedas proposedgetCoachReferencesearchWeddingSuppliers suggestRefinementsgetBudgetBreakdown0
9 · Final monthas proposedno toolgetUpcomingTaskAdviceno rule chip0

Tools are from sample 1. Judge fails count both samples of v20 and v21 for the couple.

Findings

Two sweeps ran on the same build: runs-existing-screens-1002 before a harness fix and runs-existing-screens-1002b after it. The screens on this page come from the second sweep.

  1. No regression. The second sweep passed 26 of 26 runs with no failed claim. Against runs-phase6-after2-1001, no claim got worse. The rail claim moved from 12/14 to 14/14, and all 28 rails in both sweeps match the brief.
  2. Harness fix: a stale search area. The saved login state of e2e01 (9 September) kept Brighton as the search page area. A venue search with no named place uses that area before the record. In the first sweep, couple 4's "find the right venue" tap searched Brighton, not Surrey. The harness reset now clears state-search-searchLocation. Earlier runs on e2e01 carry the same area, but only a search with no named place used it.
  3. Empty bar under an offer. 9 of 77 replies end on a yes/no offer with no chip to accept it. Couple 5's check-in offers to compare the saved venues with an empty bar in 3 of 4 runs. The "narrow down" tap ends on "look up Eastington Park's listing" with no chip in 3 of 4 runs. The brief kept the follow-up bar out of scope (§4.4).
  4. A greeting line on every rail tap. All 23 replies to a rail chip on the empty chat open with "I'm here to help…" or "Welcome…". The claim "A rail-first conversation still greets" asks for this line. The empty-chat title already says "Hey, I'm BB!", and that is why phase 7 was dropped. Decide if the greeting rule still applies to existing couples.
  5. The year question after "narrow down". Couple 3 has no date. After the compare card, the reply asks for the year in all 3 runs that replied. This is the date unlock, and the judge accepts it. But the venue work stops at that question.
  6. One budget slip. In the first sweep, couple 3 sample 2 tapped "How long does planning a wedding take?". The reply then offered a starter budget. The turn plan made no offer, so the model chose the question itself. The judge flagged 1 of 47 answered taps across both sweeps.
  7. One silent reply. In the first sweep, the forced compareSuppliers call for couple 3 ended with the provider reason max_messages after 2,893 reasoning tokens. The couple saw no reply. The same reason shows in 25 earlier runs since 24 September.
  8. Compare card: older issues. "Venue type" and "Venue style" show raw listing keys, for example countryHouseManor, statelyHome, unselected and lovelyGrounds. A venue with no capacity shows "0" guests. The Eastington Park thumbnail path in the directory is 1..JPEG, so its image is broken. These issues come from before this branch.
  9. Couple 5 check-in. In one run of the second sweep, the check-in offered a new venue search in Gloucestershire. The couple has four saved venues, and the rail offers "Help us narrow down our venues".

Judge across the sweep

43 claims passed on every run that answered them.

3 · Browsing, no date

Sam & Jordan

Wedding state

DateNo date
PlaceYorkshire
Saved venues3: Cleatham Hall, Duke Of Wellington, The Mansion, Leeds
BookedNothing
GuestsAbout 120, list empty
BudgetNone
ChecklistNot started

Rail

Screens, sample 1

Empty chatrail opener narrow-venues start-guest-list
The empty chat for couple 3
Check-intapped the opener · no tool · bar: “How long does planning a wedding take?” · “Not sure on the year yet”
The check-in reply for couple 3
First chip under the check-intapped “How long does planning a wedding take?” · searchWeddingKnowledge getCoachReference · bar: “Compare our saved venues”
The reply to the first chip under the check-in for couple 3
Rule chip on the empty chattapped “Help us narrow down our venues” · compareSuppliers planningSnapshot · bar: “How long does planning a wedding take?” · “Not sure on the year yet”
The reply to a rail chip for couple 3
Screens, sample 2
Empty chatrail opener narrow-venues start-guest-list
The empty chat for couple 3
Check-intapped the opener · no tool · bar: “How long does planning a wedding take?” · “Not sure on the year yet”
The check-in reply for couple 3
First chip under the check-intapped “How long does planning a wedding take?” · getCoachReference · bar: “Compare two of our saved venues”
The reply to the first chip under the check-in for couple 3
Rule chip on the empty chattapped “Help us narrow down our venues” · compareSuppliers · bar: “How long does planning a wedding take?” · “Not sure on the year yet”
The reply to a rail chip for couple 3
Judge, 4 runs
  • existing-browsing-no-date sample 1: 31 pass, 0 fail
  • existing-browsing-no-date sample 2: 31 pass, 0 fail
  • existing-tap-narrow-no-date sample 1: 28 pass, 0 fail
  • existing-tap-narrow-no-date sample 2: 27 pass, 0 fail

4 · Early planner, nothing saved

Sam & Alex

Wedding state

DateJune 2029
PlaceSurrey
Saved venuesNone
BookedNothing
GuestsNo guest count, list empty
BudgetNone
Checklist8 of 265 tasks done (3%)

Rail

Screens, sample 1

Empty chatrail opener find-venue start-guest-list
The empty chat for couple 4
Check-intapped the opener · no tool · bar: “What should we focus on next?”
The check-in reply for couple 4
First chip under the check-intapped “What should we focus on next?” · getUpcomingTaskAdvice · bar: “Help us do this” · “Why this first?” · “Not now”
The reply to the first chip under the check-in for couple 4
Rule chip on the empty chattapped “Help us find the right venue” · searchWeddingSuppliers suggestRefinements · bar: “Show Surrey barn venues” · “Find Surrey manor houses” · “Add on-site accommodation”
The reply to a rail chip for couple 4
Screens, sample 2
Empty chatrail opener find-venue start-guest-list
The empty chat for couple 4
Check-intapped the opener · no tool · bar: “What should we focus on next?”
The check-in reply for couple 4
First chip under the check-intapped “What should we focus on next?” · getUpcomingTaskAdvice · bar: “Help us do this” · “Why this first?” · “Not now”
The reply to the first chip under the check-in for couple 4
Rule chip on the empty chattapped “Help us find the right venue” · searchWeddingSuppliers suggestRefinements · bar: “Narrow to manor houses” · “Show barn venues” · “Include on-site accommodation”
The reply to a rail chip for couple 4
Judge, 4 runs
  • existing-early-planner sample 1: 32 pass, 0 fail
  • existing-early-planner sample 2: 32 pass, 0 fail
  • existing-tap-find-venue sample 1: 29 pass, 0 fail
  • existing-tap-find-venue sample 2: 29 pass, 0 fail

5 · Shortlisting, nothing booked

Priya & Dev

Wedding state

DateSummer 2028
PlaceGloucestershire
Saved venues4: The Tunnel House, Lapstone Barn Ltd, Eastington Park, Hilton Puckrup Hall & Golf
BookedNothing
GuestsAbout 120, list empty
Budget£35,000, figure only, not set up
Checklist53 of 265 tasks done (20%)

Rail

Screens, sample 1

Empty chatrail opener narrow-venues start-guest-list
The empty chat for couple 5
Check-intapped the opener · no tool · bar: “Find us wedding venues”
The check-in reply for couple 5
First chip under the check-intapped “Find us wedding venues” · searchWeddingSuppliers suggestRefinements · bar: “Barn venues in Gloucestershire” · “Venues with rooms on site” · “Outdoor ceremony venues”
The reply to the first chip under the check-in for couple 5
Rule chip on the empty chattapped “Help us narrow down our venues” · compareSuppliers · bar: no chips
The reply to a rail chip for couple 5
Screens, sample 2
Empty chatrail opener narrow-venues start-guest-list
The empty chat for couple 5
Check-intapped the opener · getCoachReference · bar: no chips
The check-in reply for couple 5
First chip under the check-intyped “What should we do next?” · getUpcomingTaskAdvice · bar: “Help us do this” · “Explore venues” · “Why this first?”
The reply to the first chip under the check-in for couple 5
Rule chip on the empty chattapped “Help us narrow down our venues” · compareSuppliers · bar: no chips
The reply to a rail chip for couple 5
Judge, 4 runs
  • existing-shortlisting sample 1: 36 pass, 0 fail
  • existing-shortlisting sample 2: 31 pass, 0 fail
  • existing-tap-narrow-venues sample 1: 23 pass, 0 fail
  • existing-tap-narrow-venues sample 2: 23 pass, 0 fail

6 · Venue just booked, no guest list

Mo & Ella

Wedding state

Date20 December 2027
PlaceWindermere, Windermere
Saved venuesNone
BookedLakeside Hall (venue)
GuestsAbout 70, list empty
Budget£22,000, figure only, not set up
Checklist27 of 265 tasks done (10%)

Rail

Screens, sample 1

Empty chatrail opener start-guest-list save-budget
The empty chat for couple 6
Check-intapped the opener · no tool · bar: “Photographers near Windermere”
The check-in reply for couple 6
First chip under the check-intapped “Photographers near Windermere” · searchWeddingSuppliers suggestRefinements · bar: “Natural, candid photography” · “Documentary-style photographers” · “Photographers within 20 miles”
The reply to the first chip under the check-in for couple 6
Rule chip on the empty chattapped “Help us start our guest list” · searchWeddingKnowledge · bar: “Open our guest list”
The reply to a rail chip for couple 6
Screens, sample 2
Empty chatrail opener start-guest-list save-budget
The empty chat for couple 6
Check-intapped the opener · no tool · bar: “What should we focus on next?”
The check-in reply for couple 6
First chip under the check-intapped “What should we focus on next?” · getUpcomingTaskAdvice · bar: “Help us do this” · “Explore photographers” · “Why this first?”
The reply to the first chip under the check-in for couple 6
Rule chip on the empty chattapped “Help us start our guest list” · searchWeddingKnowledge · bar: “Open our guest list”
The reply to a rail chip for couple 6
Judge, 4 runs
  • existing-venue-just-booked sample 1: 34 pass, 0 fail
  • existing-venue-just-booked sample 2: 32 pass, 0 fail
  • existing-tap-guest-list sample 1: 25 pass, 0 fail
  • existing-tap-guest-list sample 2: 25 pass, 0 fail

7 · Venue booked, guest list started

Jo & Max

Wedding state

Date18 September 2027
PlaceKent
Saved venuesNone
BookedOakwood Barn (venue)
GuestsAbout 80, 64 on the list
Budget£30,000, set up
Checklist106 of 265 tasks done (40%)

Rail

Screens, sample 1

Empty chatrail opener save-budget find-photographers
The empty chat for couple 7
Check-intapped the opener · no tool · bar: “Wedding photographers in Kent”
The check-in reply for couple 7
First chip under the check-intapped “Wedding photographers in Kent” · searchWeddingSuppliers suggestRefinements · bar: “Natural, candid photographers” · “Documentary-style photographers” · “Editorial wedding photographers”
The reply to the first chip under the check-in for couple 7
Rule chip on the empty chattapped “Where can we save on our wedding?” · getBudgetBreakdown · bar: “How should we start thinking about a budget?”
The reply to a rail chip for couple 7
Screens, sample 2
Empty chatrail opener save-budget find-photographers
The empty chat for couple 7
Check-intapped the opener · getCoachReference · bar: “Show us wedding photographers”
The check-in reply for couple 7
First chip under the check-intapped “Show us wedding photographers” · searchWeddingSuppliers suggestRefinements · bar: “Natural documentary style” · “Under £1,500” · “Within 20 miles of Kent”
The reply to the first chip under the check-in for couple 7
Rule chip on the empty chattapped “Where can we save on our wedding?” · getBudgetBreakdown · bar: “Show us wedding photographers”
The reply to a rail chip for couple 7
Judge, 4 runs
  • existing-list-started sample 1: 34 pass, 0 fail
  • existing-list-started sample 2: 34 pass, 0 fail
  • existing-tap-save-set-up sample 1: 26 pass, 0 fail
  • existing-tap-save-set-up sample 2: 26 pass, 0 fail

8 · Six months out, photographer booked

Rob & Nia

Wedding state

Date17 April 2027
PlaceNorfolk
Saved venuesNone
Other savedTimeless Florals (florist)
BookedSaltmarsh Coach House (venue), Sea Glass Photography (photo)
GuestsAbout 110, 96 on the list
Budget£40,000, set up
Checklist146 of 265 tasks done (55%)

Rail

Screens, sample 1

Empty chatrail opener save-budget
The empty chat for couple 8
Check-intapped the opener · getCoachReference · bar: “Wedding music near Norfolk”
The check-in reply for couple 8
First chip under the check-intapped “Wedding music near Norfolk” · searchWeddingSuppliers suggestRefinements · bar: “Show acoustic musicians” · “Show wedding DJs” · “Find live bands near Norfolk”
The reply to the first chip under the check-in for couple 8
Rule chip on the empty chattapped “Where can we save on our wedding?” · getBudgetBreakdown · bar: “Help us think through our guest-list size”
The reply to a rail chip for couple 8
Screens, sample 2
Empty chatrail opener save-budget
The empty chat for couple 8
Check-intapped the opener · getCoachReference · bar: “What should we focus on next?”
The check-in reply for couple 8
First chip under the check-intapped “What should we focus on next?” · getUpcomingTaskAdvice · bar: “Help us do this” · “Explore videographers” · “Why this first?”
The reply to the first chip under the check-in for couple 8
Rule chip on the empty chattapped “Where can we save on our wedding?” · getBudgetBreakdown · bar: “Lower-cost music in Norfolk”
The reply to a rail chip for couple 8
Judge, 4 runs
  • existing-six-months-out sample 1: 34 pass, 0 fail
  • existing-six-months-out sample 2: 31 pass, 0 fail
  • existing-tap-save-late sample 1: 26 pass, 0 fail
  • existing-tap-save-late sample 2: 26 pass, 0 fail

9 · Final month

Lee & Kim

Wedding state

Date23 October 2026
PlaceBath, Bath
Saved venuesNone
BookedCrescent Rooms (venue), Bath Light Studio (photo)
GuestsAbout 90, 88 on the list
Budget£28,000, set up
Checklist225 of 265 tasks done (85%)

Rail

Screens, sample 1

Empty chatrail opener
The empty chat for couple 9
Check-intapped the opener · no tool · bar: “What should we focus on next?”
The check-in reply for couple 9
First chip under the check-intapped “What should we focus on next?” · getUpcomingTaskAdvice · bar: “Help us do this” · “Explore stationery” · “Why this first?”
The reply to the first chip under the check-in for couple 9
Screens, sample 2
Empty chatrail opener
The empty chat for couple 9
Check-intapped the opener · no tool · bar: “What should we focus on next?”
The check-in reply for couple 9
First chip under the check-intapped “What should we focus on next?” · getUpcomingTaskAdvice · bar: “Help us do this” · “Explore stationery” · “Why this first?”
The reply to the first chip under the check-in for couple 9
Judge, 2 runs
  • existing-final-month sample 1: 32 pass, 0 fail
  • existing-final-month sample 2: 32 pass, 0 fail