Made-up example. Cinco Chaminés is an invented company: the homes, prices, availability and bookings are simulated, and the pictures are AI-generated. The weather, sea, UV and events are real, read live from data.promptugal.news. What is this site?

Cinco Chaminésholiday homes in the Algarve (a made-up example)

Round 4

The round report, rendered at build time from team/rounds/4.md.

Aim (conductor, 2026-11-19, about 03:55 UTC). From the debrief's "Open" item 2 (team/DEBRIEF.md) and the round 3 reviews (team/rounds/3/review/BUILD.md and RECHECK.md), inside Dinis's widened brief of 18 November, 17:10 to 17:45 UTC (team/rounds/2.md): make the modes something a house owner can follow and a viewer can play, and make the numbers believable.

Live at the start: v0.1.6 (round 3).

What this round builds

  1. A guided tour. One "Show me" button (on / and /play/) that books a stay, moves time, opens the admin and points at each change, in about 90 seconds; keyboard usable, can be stopped at any step, reduced-motion respected, every step labelled simulation.
  2. A world that responds to the owner. In /world/ and /play/: change a rate or a minimum stay during a run and see conversion and occupancy change; show "what if" next to "what happened". The world's background occupancy lower, or chosen by the seed, so fewer visitors leave for "dates taken" (S10).
  3. A believable P&L. Management, maintenance reserve, IMI, accountant and mortgage or depreciation as labelled assumptions the owner can edit (kept in localStorage); net yield on a house value the owner types; check the Faro tourist-tax history model (S9). Every figure a simulation, not legal or tax advice.
  4. A run score. Occupancy, average rating, profit and injections held, as a run summary in /world/ the viewer can try to beat with a new seed.
  5. One place for the policy story. In /admin/ Today: "what the agents were stopped from doing today", each DENY linked to the text that caused it, the visitor's own held texts included.
  6. The legality section in full. Every row of team/rounds/2/research/LEGAL-PT.md shown with its source, date read and how sure; "not legal or tax advice" at the top and bottom.
  7. Open items from round 3. N1 (the five /d/ designs read the simulation's rate), S2 (going back in time: say plainly that it does not undo, and do not replay hours already played), S4 (the four wording errors).

Rules (unchanged)

Made-up example on every page; the demonstration statement at the end of every booking; IPMA named with read time and terms beside every figure; events with source and link; only dates with real data; state only in the visitor's localStorage, nothing sent, only GETs; no booking-platform brand names; no em dashes, no hype words; name no person; labelled AI-generated pictures; "Claude (simulated)" never presented as a live model.

Status

Conductor's answers to the designer (PROPOSAL.md, questions), 2026-11-19 about 04:20 UTC

  1. The tour may reset the simulation, as proposed (backup to cc.tour.backup.v1, put back at the end or from the put-back bar). If the visitor has a booking of their own, the start dialog says so first: "The tour uses its own simulation for about 90 seconds; yours is kept and put back at the end." Other tabs may show the tour while it runs; the start dialog says that too.
  2. Management fee defaults to 20 percent, labelled ASSUMPTION, with a one-click "I run it myself (0 percent)".
  3. Yes: the researcher writes team/rounds/4/research/PNL-DEFAULTS.md (IMI rates, typical management fee range, accountant fees, maintenance reserve rule of thumb) with sources and date read; the page cites it. Until then each default stays labelled ASSUMPTION.
  4. Keep the score weights and full-marks points as proposed, and show the formula beside the score.
  5. Reset clears everything this site keeps in the browser, assumptions and best scores included, as proposed; the confirm dialog lists what goes.

Conductor's decisions on the proposal review and the P&L research, 2026-11-19 about 04:45 UTC

Report (conductor, 2026-11-19, about 06:30 UTC)

What the round was spent on. One design pass and one revision (team/rounds/4/design/PROPOSAL.md), two proposal reviews (CHANGES BEFORE BUILD, then PASS: BUILD), one research note with sourced P&L defaults (team/rounds/4/research/PNL-DEFAULTS.md), one build, one build review (DO NOT RELEASE YET: the tour's forecast step showed no forecast), one fix pass and one re-check (PASS: RELEASE, team/rounds/4/review/RECHECK.md). The conductor added the signed-commits check (scripts/check-signatures.mjs, in release.sh and CI) and took two source URLs naming booking platforms out of the published research note.

What changed for the visitor. A guided tour ("Show me how it works", about 90 seconds, on / and /play/) that books on dates with a real IPMA forecast, moves time, opens the admin and points at each change, then puts the visitor's own simulation back. /world/ answers the owner: a "Your rates" panel, a "what if you had left the rates" comparison, one site-wide level of bookings from elsewhere, and a run score out of 100 with its formula and best runs. /admin/ Money has 16 editable assumptions, sourced or labelled, gross and net yield on a house value the owner types (simulated, not financial advice): on the seeded history the margin is now about 13 percent, not 75. /admin/ Today shows what the agents were stopped from doing today. Legality shows all 35 rows of the research with sources. The five designs read the simulation's rates; going back in time refuses bookings instead of replaying; each home checks in at its own time.

Reviewer's verdict. PASS: RELEASE. Gates OK; tests 84, 83 pass, 1 skipped.

Open for round 5. From RECHECK.md: a typed rate keeps "set by Helena (synthetic owner)" until the next redraw; same-day bookings after the check-in hour never check in in /world/; check-out is 10:00 for every home and the admin's manual check-in uses a fixed 16:00; the tour's forecast card gives a reason the guest page contradicts when the data is blocked. From BUILD.md: the occupancy part of the score is nearly always full, so the score turns on the margin. Round 5 as planned: five admin designs, a synthetic-user study of the site itself, a full review, then team/HANDOVER.md.

For Dinis. Nothing new to decide. Still open from round 3: read access to AgentiCCells__Site__Factory (now moot, conductor.b@ sent the two files), and the round line to cells.b@ is refused by this session's permission check.

The round's files

31 files under team/rounds/4/, by the role that wrote them. Markdown files are rendered here; screenshots under 300 KB are copied here; other files open on GitHub.

Research (researcher)

Design (designer)

Build (developer)

Review (reviewer)

22 screenshots
Ask the agent (a simulation: pre-written answers, no live model)