Showcase · Read only

A frozen OPFLA planning session. Nothing on this page changes anything.

Operation Florida

Seven people.
Thirty possibilities.
One trip.

A family had one week in Florida and no agreement about what to do with it. This is what OPFLA was given, what it computed, what it noticed, and what the trip became. Every figure below is recomputed from the seven frozen ballots by the same code the live planner runs.

Frozen 2026-08-30 · Preference Engine OPFLA-PE-0.1 · 7 of 7 ballots filed

01The problem

Group trips are usually decided by whoever speaks up most, or by whoever books first. Seven people had never been asked the same question at the same time.

The briefing

You have been assigned to a seven-person operation. Somewhere in Florida there is a version of this trip that makes everyone happy. Nobody has ever asked all seven of us what that actually looks like. We are asking now. Thirty files. Rate each one. Then seven Golden Tickets — the seven things you most want to happen. There are no wrong answers and no budget on this screen.

The hard part is not collecting opinions. It is that four days cannot hold everything seven people want, the disagreements are real, and somebody still has to decide — without being able to claim the numbers made them do it.

02What OPFLA was given

Four fixed things, and one question asked seven times.

Seven travellers

THE ARCHITECT
Systems Design & Mission Planning
Building the plan before anyone else knows we need one
THE SAFEGUARD
Emotional Support & Tactical Nursing
Ensuring no birthday concludes in fewer than four days
THE COMMISSIONER
Competitive Operations & Rules Enforcement
Enforcing regulations he did not personally write
THE WILDCARD
Unconventional Operations & Loophole Discovery
Getting away with things Agent 003 specifically remembers being against the rules
THE INSPECTOR
Quality Control & Operational Oversight
Quietly reviewing the plan and identifying the part you forgot
THE PRODUCER
Logistics, Escalation & Morale Operations
Turning "let's make fun drinks" into a fully provisioned beverage program
THE VETERAN
Tactical Calm & Covert Commentary
Saying very little until he has the funniest line in the room

Seven real people, shown by the role each plays in the family rather than by name. Which seven people they were is not part of what OPFLA did.

Thirty possibilities

Parks, attractions, meals, lodging and trip-style questions — the whole ballot, in the order it was written. Nothing here is ordered by how anyone voted.

Magic Kingdom · EPCOT · Hollywood Studios · Animal Kingdom · Park Hopper · Mickey's Very Merry Christmas Party · Guardians of the Galaxy: Cosmic Rewind · Star Wars: Rise of the Resistance · TRON Lightcycle / Run · Slinky Dog Dash · Avatar Flight of Passage · Disney fireworks · Character meet-and-greets · Resort / pool downtime · Explore Disney resorts · Disney Springs · Character meal · Special Disney dinner · EPCOT food adventure · One big family meal · Universal Studios / Islands of Adventure day · Epic Universe day · Harry Potter experiences · Beach day · Tampa day · Relaxed family day · Stay at Disney's BoardWalk · Premium / preferred BoardWalk view · Maximize park time · Slower vacation pace

Six days, two of them walls

2026-11-16Travel onlyThe flight lands at 9:46 PM. Monday evening is not a usable activity block.
2026-11-17Full
2026-11-18Full
2026-11-19Full
2026-11-20Full
2026-11-21PartialThe return departs at 7:18 PM. The day must absorb Orlando to Tampa travel, vehicle and airport logistics, and an arrival buffer.

TUL → TPA, 14:54–21:46TPA → TUL, 19:18–23:53

What the window can hold

Entered by hand, never computed. OPFLA makes the conflict visible; it does not resolve it, and it has no solver.

Full operating days — 4 available · Tue Nov 17 through Fri Nov 20. Saturday is partial and is not one of these.
Evening slots — 4 available · Nov 17-20. Monday arrives 9:46 PM; Saturday is a partial departure day.
Hotel nights — 5 available · BoardWalk, Nov 16-20, departing Nov 21.
Park days — 0 available · Not yet decided.
Universal days — 0 available · Not yet decided.

03What people were asked

Each traveller rated all thirty on a 1–5 scale, then spent exactly seven Golden Tickets on the things they most wanted. The two are separate questions on purpose: a rating says how much you like something, a ticket says do not let this get cut.

One ballot, as filedagent-1

How this one ballot spent its 30 ratings — mostly middle, which is what an honest ballot looks like:

5 · This matters a lot4 of 30
4 · Really want it3 of 30
3 · Would be nice13 of 30
2 · Not important7 of 30
1 · Don't care3 of 30

And its seven Golden Tickets:

EPCOT food adventure · Epic Universe day · Guardians of the Galaxy: Cosmic Rewind · Slinky Dog Dash · Star Wars: Rise of the Resistance · Tampa day · TRON Lightcycle / Run

Nobody knows whose ballot that is

Including OPFLA. Ballots are separated from their voters when they are read — the labels are assigned at random and reshuffled every time — and the frozen copy this page reads discarded the mapping entirely. There is no lookup table anywhere in the system. The seven people above are real and named; these seven ballots are real and anonymous; nothing connects them. That was a deliberate design decision, made so that nobody would spend the trip knowing who voted against their favourite thing.

04How it was scored

Not by averaging. Averaging a 1–5 scale treats everybody's 4 as the same size, and it is not — some people rate generously and some rate hard. So each ballot is read relative to itself: where a given item sat within that person's own thirty.

Four separate readings come out, and they are deliberately never combined into one number. A single score would look decisive and would hide which kind of evidence produced it.

Relative support — where an item sat inside each person’s own ranking, averaged across the seven.
Declared sentiment — how many rated it high (4–5) and how many rated it low (1–2), as counts of people.
Scarce protection — Golden Tickets, which are capped at seven each and therefore mean something a rating cannot.
Robustness — how far each result moves if any single ballot is removed. Seven is a small electorate and the honest answer is often “a lot”.

All thirty, in observed rank order

Evidence, not a verdict. A position in this table decides nothing on its own — and the rightmost column is why: only a handful of these hold their place when one ballot is dropped.

Rank	Experience	Support	Tickets	If one ballot is removed
1 of 30	One big family meal	0.78	3	Support 0.74–0.82, rank 1–2 across the 7 omission scenarios.
2 of 30	EPCOT	0.72	2	Support 0.69–0.77, rank 1–3 across the 7 omission scenarios.
3 of 30	Hollywood Studios	0.68	2	Support 0.64–0.75, rank 2–5 across the 7 omission scenarios.
4 of 30	Magic Kingdom	0.63	3	Support 0.58–0.70, rank 3–9 across the 7 omission scenarios.
5 of 30	Tampa day	0.61	1	Support 0.57–0.70, rank 2.5–12 across the 7 omission scenarios.
6 of 30	Beach day	0.59	3	Support 0.55–0.65, rank 4–13 across the 7 omission scenarios.
7 of 30	Resort / pool downtime	0.58	3	Support 0.54–0.65, rank 4–15 across the 7 omission scenarios.
8 of 30	Guardians of the Galaxy: Cosmic Rewind	0.58	3	Support 0.52–0.66, rank 4–16 across the 7 omission scenarios.
9.5 of 30	Avatar Flight of Passage	0.57	2	Support 0.52–0.65, rank 6–15 across the 7 omission scenarios.
9.5 of 30	TRON Lightcycle / Run	0.57	1	Support 0.52–0.65, rank 6–14.5 across the 7 omission scenarios.
11 of 30	Star Wars: Rise of the Resistance	0.56	3	Support 0.50–0.64, rank 5–19 across the 7 omission scenarios.
12.5 of 30	EPCOT food adventure	0.56	2	Support 0.52–0.64, rank 4–15 across the 7 omission scenarios.
12.5 of 30	Maximize park time	0.56	2	Support 0.50–0.63, rank 6–17.5 across the 7 omission scenarios.
14 of 30	Special Disney dinner	0.55	1	Support 0.50–0.62, rank 5–18 across the 7 omission scenarios.
15 of 30	Slinky Dog Dash	0.53	1	Support 0.47–0.61, rank 7–21 across the 7 omission scenarios.
16.5 of 30	Epic Universe day	0.53	4	Support 0.48–0.60, rank 8–18.5 across the 7 omission scenarios.
16.5 of 30	Universal Studios / Islands of Adventure day	0.53	2	Support 0.48–0.60, rank 6–18.5 across the 7 omission scenarios.
18 of 30	Character meal	0.49	0	Support 0.44–0.56, rank 13–21 across the 7 omission scenarios.
19.5 of 30	Animal Kingdom	0.47	0	Support 0.43–0.51, rank 16–22.5 across the 7 omission scenarios.
19.5 of 30	Disney fireworks	0.47	1	Support 0.41–0.51, rank 16–22.5 across the 7 omission scenarios.
21 of 30	Mickey's Very Merry Christmas Party	0.45	1	Support 0.39–0.52, rank 15–24 across the 7 omission scenarios.
22 of 30	Stay at Disney's BoardWalk	0.43	1	Support 0.36–0.49, rank 19–25 across the 7 omission scenarios.
23 of 30	Disney Springs	0.41	1	Support 0.34–0.46, rank 21–25 across the 7 omission scenarios.
24 of 30	Park Hopper	0.41	2	Support 0.34–0.47, rank 20–26 across the 7 omission scenarios.
25 of 30	Harry Potter experiences	0.36	2	Support 0.33–0.40, rank 23–27 across the 7 omission scenarios.
26 of 30	Explore Disney resorts	0.35	1	Support 0.26–0.38, rank 24–28 across the 7 omission scenarios.
27 of 30	Relaxed family day	0.33	0	Support 0.24–0.37, rank 25–27.5 across the 7 omission scenarios.
28 of 30	Slower vacation pace	0.27	2	Support 0.18–0.31, rank 27–30 across the 7 omission scenarios.
29 of 30	Character meet-and-greets	0.22	0	Support 0.16–0.24, rank 28–30 across the 7 omission scenarios.
30 of 30	Premium / preferred BoardWalk view	0.20	0	Support 0.15–0.23, rank 29–30 across the 7 omission scenarios.

05What it noticed

The engine answers four questions separately, because they are four different questions. Blending them is exactly how a group decision starts looking more settled than it is.

What does the family broadly want?

Within-respondent relative order, averaged across the cohort, alongside declared sentiment as the share rating an experience 4 or 5. Ordinal evidence about position, not a score and not an enthusiasm total.

One big family meal holds the highest relative support in the cohort at 0.78, at observed rank 1 of 30.
Premium / preferred BoardWalk view holds the lowest at 0.20, at observed rank 30 of 30.
4 of 30 experiences were rated 4 or 5 by at least half the cohort.

One big family meal
Relative support0.78
EPCOT
Relative support0.72
Hollywood Studios
Relative support0.68

What did people protect when choices became scarce?

Golden Tickets are a scarcity-constrained channel: seven per ballot, spent against thirty experiences. They are kept separate from ratings because they answer a different question — not how much someone likes a thing, but what they refused to lose.

25 of 30 experiences were protected on at least one ballot. 5 were protected on none.
Epic Universe day was protected on 4 of 7 ballots, the strongest scarcity signal in the cohort, at observed rank 16.5 of 30.
Epic Universe day sits at observed rank 16.5 of 30 on ordinary ratings and was still protected on 4 of 7 ballots. Rank and scarcity disagree here.

Epic Universe day
4 of 7 ballots protected it
One big family meal
3 of 7 ballots protected it
Magic Kingdom
3 of 7 ballots protected it

Where are we aligned, and where are we split?

Spread of relative positions across the observed cohort. High polarization means both tails are populated at once — it is not the inverse of agreement, and neither is a measure of how much the experience is wanted.

Guardians of the Galaxy: Cosmic Rewind is the widest split in the cohort, at polarization 0.98.
Premium / preferred BoardWalk view is one of the clearest points of alignment, and the alignment is negative: 6 of 7 rated it 1 or 2. Agreement tells us the family puts it in roughly the same place, not that they want it.
Agreement and polarization are separate readings, and neither measures how much an experience is wanted. An experience can be middling in rank because everyone is lukewarm, or because half the cohort is at each end.

Guardians of the Galaxy: Cosmic Rewind
Observed polarization0.98
Beach day
Observed polarization0.49
Resort / pool downtime
Observed polarization0.49

Which conclusions are robust, and which depend on one ballot?

Each experience re-analysed 7 times, once with each ballot removed. The band shows how far the reading travels. It measures dependence on any single agent; it is not a confidence interval, because there is no wider population to be uncertain about.

Only 4 of 30 experiences stay within two rank positions when any single ballot is removed.
One big family meal stays between rank 1 and 2 whichever single ballot is removed — the conclusion survives anybody leaving the room.
Star Wars: Rise of the Resistance travels furthest: rank 5 to 19 across the 7 omission scenarios. That is dependence on one agent, not statistical uncertainty — the cohort is the electorate, not a sample.

One big family meal
Relative support0.78
Support 0.74–0.82, rank 1–2 across the 7 omission scenarios.
Premium / preferred BoardWalk view
Relative support0.20
Support 0.15–0.23, rank 29–30 across the 7 omission scenarios.
EPCOT
Relative support0.72
Support 0.69–0.77, rank 1–3 across the 7 omission scenarios.

One big family meal

Finished first of all 30. 5 of 7 rated it high and nobody rated it low — the only thing on the whole ballot with nothing against it. The most-protected single item was Epic Universe day, at 4 of 7 Golden Tickets — a different item, from a different kind of evidence.

And one thing genuinely split them

Guardians of the Galaxy: Cosmic Rewind — 4 rated it high, 3 rated it low. A spreadsheet average would have reported this as a mild yes. It is not a mild yes; it is a real disagreement, and the trip had to be planned knowing that.

06What changed because of it

This is the part most tools skip. The numbers do not decide anything — a person does, and OPFLA makes them write down why. Every ruling below was typed by the planner, and nothing in the system can create, promote or clear one on its own.

The rulings that argue against themselves

Two things ended up in the trip that the family never voted for. OPFLA records not just that they are in, but that they are in for a structural reason — and warns the next reader not to mistake the decision for evidence of enthusiasm. That is the whole difference between a plan and a leaderboard.

Animal Kingdom
CORE BY PREREQUISITE, NOT BY INDEPENDENT SUPPORT. The park did not independently earn core status — 3 high / 3 low, no Golden Tickets, rank 19.5 of 30. It is core because a core experience requires it: the venue becomes part of the protected plan as the cost of honoring Flight of Passage. Do not later read this core state as stronger preference evidence than the ballots showed.
Stay at Disney's BoardWalk
CORE BY PLANNING NECESSITY, NOT BY BALLOT STRENGTH — the same kind of ruling as Animal Kingdom. The cohort did not independently demand it: rank 22 of 30, 2 high / 3 low, 1 Golden Ticket. It is core because it improves the shape. The deciding fact was deliberately excluded from the ballot (addendum exclusion table: 'Walk to EPCOT and Hollywood Studios — a benefit of staying at BoardWalk, not an independent experience'), so no ballot evidence for it exists or could exist. Do not later read this core state as preference evidence.

One choice pulls another in

Recorded relationships, not inferences. This is why a highly-rated ride can drag a middling park into the trip with it.

Mickey's Very Merry Christmas Party requires Magic Kingdom
Guardians of the Galaxy: Cosmic Rewind requires EPCOT
Star Wars: Rise of the Resistance requires Hollywood Studios
TRON Lightcycle / Run requires Magic Kingdom
Slinky Dog Dash requires Hollywood Studios
Avatar Flight of Passage requires Animal Kingdom
EPCOT food adventure requires EPCOT
Harry Potter experiences requires Universal Studios / Islands of Adventure day
Premium / preferred BoardWalk view requires Stay at Disney's BoardWalk

Three ways something earns protection

The rulings below refer to these by name. Each is a separate test against a different kind of evidence, and clearing one is enough — they are never added together.

Broad support
Rated 4 or 5 on at least 5 of 7 ballots.
Intensity
Protected by a Golden Ticket on at least 3 of 7 ballots.
Robustness
After removing any one ballot, support remains at or above the upper-third threshold (0.667).

Where every experience landed

Core

10 experiences

We intend to make these happen.

Magic Kingdom
Solid in both channels: rank 4 of 30, support 0.63, 3 high / 1 low, and protected by Intensity at 3 of 7 Golden Tickets. Takes the second of the two remaining full days and keeps the four park days inside the WDW envelope.
EPCOT
Most listed the park OR one of the rides within this park highly.
Hollywood Studios
Strongest ratings evidence of the four contenders for the two remaining full days: rank 3 of 30, support 0.68, 4 high / 1 low, the most stable leave-one-out band (rank 2-5), and within 0.025 of clearing Robustness. Also the venue Rise of the Resistance requires. Core on the ratings case, with the prerequisite as a second reason rather than the only one.
Animal Kingdom
CORE BY PREREQUISITE, NOT BY INDEPENDENT SUPPORT. The park did not independently earn core status — 3 high / 3 low, no Golden Tickets, rank 19.5 of 30. It is core because a core experience requires it: the venue becomes part of the protected plan as the cost of honoring Flight of Passage. Do not later read this core state as stronger preference evidence than the ballots showed.
Guardians of the Galaxy: Cosmic Rewind
BACKFILLED 2026-08-25 — the original ruling recorded no rationale. Stated from the evidence rather than reconstructed, and deliberately no richer than the evidence supports: protected by Intensity alone (3 of 7 Golden Tickets), one ballot short of Broad support (4 of 5), Robustness below floor (0.52 against 0.667); rank 8 of 30, 4 high / 3 low. Its recorded prerequisite — Cosmic Rewind requires EPCOT — is satisfied, EPCOT being core. No reasoning beyond that is claimed.
Avatar Flight of Passage
The strongest actual experience in the band: rank 9.5, 3 high, only 2 low, and a respectable support floor even when a ballot disappears. The wide rank band says its exact rank is uncertain, not that the underlying support is weak.
Resort / pool downtime
Could add this as a half day or split time (If some want Galaxy's Edge and others don't could be a resort day option for example)
One big family meal
Just need to figure out where and when. But obviously a meal is near the top of all of our lists!
Stay at Disney's BoardWalk
CORE BY PLANNING NECESSITY, NOT BY BALLOT STRENGTH — the same kind of ruling as Animal Kingdom. The cohort did not independently demand it: rank 22 of 30, 2 high / 3 low, 1 Golden Ticket. It is core because it improves the shape. The deciding fact was deliberately excluded from the ballot (addendum exclusion table: 'Walk to EPCOT and Hollywood Studios — a benefit of staying at BoardWalk, not an independent experience'), so no ballot evidence for it exists or could exist. Do not later read this core state as preference evidence.
Maximize park time
The clearest trip-posture signal in the band: 3 high / 1 low and dramatically stronger than its opposite. This is useful shaping information, not just another attraction.

Maybe

11 experiences

Still under consideration. Not rejected, not protected.

Park Hopper · Star Wars: Rise of the Resistance · TRON Lightcycle / Run · Slinky Dog Dash · Character meal · Special Disney dinner · EPCOT food adventure · Universal Studios / Islands of Adventure day · Epic Universe day · Harry Potter experiences · Beach day

Doesn't fit

5 experiences

Not the same as deciding against. These lost to the shape of the trip, and if a day changes this is where reopening starts.

Mickey's Very Merry Christmas Party
The clearest out of the four: weakest evidence in the park set (rank 21, 1 high / 3 low, 1 Golden Ticket) while adding both time pressure and a separate ticket cost. It would turn Magic Kingdom into dawn-to-close plus a ticketed evening, and compete with the EPCOT family meal for a scarce evening.
Disney fireworks
Weakly supported (rank 19.5, 2 high / 3 low, 1 Golden Ticket) and consumes exactly the resource the shape is trying to protect: humane evenings. Not costly financially, but the preference evidence is too weak to let it become a default late-night obligation.
Revisit if — If an evening turns out to be genuinely free and the family wants it in the moment — this is a rejection of fireworks as a planned obligation, not of fireworks.
Relaxed family day
The evidence is weak, stable, and directionally consistent with the slower-pace ruling. More importantly, it is not the same thing as protecting resort/pool downtime: the trip can deliberately include recovery time without making an entire relaxed day part of the governing shape.
Premium / preferred BoardWalk view
The most decisive rejection in the dataset: rank 30 of 30, 6 of 7 rated it 1 or 2, zero Golden Tickets, agreement 0.80 with polarization 0.00, and a leave-one-out rank band of 29-30 — it never leaves the bottom whoever is removed. No structural need for it either. Pay for location, not for view.
Slower vacation pace
The clearest rejection: rank 28, 0 high, 5 low, stable low ranking, and directly contrary to the much stronger maximize-park-time posture. The two Golden Tickets are important evidence, but they do not outweigh the broader trip-direction signal.

Undecided

4 experiences

Nobody has called these yet. Absence of a ruling is its own state, not a “maybe”.

07What the trip became

Four full days, one park each, and two things deliberately protected so they could not get squeezed out.

Mon
Get there
A late flight and a long drive. Nothing is planned for this night, on purpose.
Tue
EPCOT
A slow start after a late arrival — and the food festival is still on while we are there.
Wed
Magic Kingdom
The biggest day. It has to be Wednesday: on three of our four nights the park closes early for a separate Christmas party, and Wednesday is the only one it does not.
Thu
Hollywood Studios and the pool
Early start, then back for the afternoon. The pool is on the schedule, not squeezed in.
Fri
Animal Kingdom and the dinner
A shorter park day, on purpose — it leaves the longest evening. Animal Kingdom is on the list because Flight of Passage is.
Sat
Breakfast, then home
One last meal together before the airport, with characters if people want them.

This is the trip backbone, not a finished itinerary. At the moment this page was frozen, the day-by-day planning surface held 0 entries — the shape was settled, the hour-by-hour was not. OPFLA reports that rather than filling the gap in.

And what stayed open

These were not voted down. There are four full days and four parks already fill them — so each of these would have to replace a park day. That is a real option, not a closed door, and the family was told so in those words.

Universal and Epic Universe

The most Golden Tickets of anything on the ballot. Also the one that needs a whole day of its own.

A beach day

Three people flagged it. It also split the room more than almost anything else.

A Tampa day

Finished fifth overall. Saturday morning might allow a small version on the way out — nothing decided.

08What a group text could not have done

It asked everyone the same question at the same time. Seven people answered thirty questions privately, before hearing anybody else’s opinion — so nobody’s answer was shaped by whoever spoke first.
It measured intensity separately from preference. Golden Tickets are capped, so “I want this” and “do not let this get cut” stayed different signals. The top-rated item and the most-protected item turned out to be different things.
It reported how fragile each result was. With seven voters most rankings move when one person is removed, and the table above says so instead of presenting a confident order.
It refused to decide. There is no solver here and no composite score. Every ruling was written by a person, with a reason attached — including the two that explicitly record that they are not what the family asked for.
It kept the ballots anonymous, permanently. The family got the findings; nobody got a way to work out who said what.

This is what survived the math

Seven travellers contributed the inputs. OPFLA made the tradeoffs visible. The family still made the decisions.

Colophon

How this was built

Written with Claude Code, in a repository where the rules are numbered. Next.js on Vercel, Postgres on Supabase, and around nineteen thousand lines of tests — fourteen files of which compute nothing at all. They read the source tree and refuse things: that a voter-facing module never imports planning data, that no second scoring function exists anywhere, that this page holds no database connection.

The constraints were written before the features, and most of them are enforced a layer below where they could be forgotten. Ballots are separated from voters in SQL rather than in application code. The scoring engine emits four separate readings and has no function that would combine them. A decision about the trip can only be written by a person — nothing in the system derives one, and the database says so in a comment next to the column.

The full argument for the estimator, the census posture, and what the analysis declines to conclude is written up separately: a methods note. It is for readers who want to know why the numbers are defensible at seven respondents, rather than what they said.

Nothing on this page is stored as a result. Every figure is recomputed when you load it, from one frozen file of seven ballots, by the same functions the private planner runs — so the page cannot quietly drift from the system it describes. Refresh it and you get the identical story, which is the only kind of story worth freezing.