Emergent test data

The Fitting Room brief was pasted word for word (1,689 characters) into the prompt box on Emergent's home screen on 27 October 2026. The clock stops when the app is live and the checks pass: 25 of 30 on the full scale, applied here as the 26 checks that could run.

The agent asked five questions: how to send emails, which address to use for the admin, whether the $24 single class could be bought, whether the timetable should show dates or weekly slots, and whether to seed the instructor and admin logins. I answered each from the brief, using info+admin@ as the admin address and the shared test password for the three seeded logins.

The agent then built the app and reported it finished on the preview address 11 minutes after I pressed send, with its own test run. I ran my checks on that preview. The first build used 26.79 credits.

It was not published: Publish opens a plan chooser on the free plan, so the address is a preview address, which opens for anyone.

• One prompt of mine outside the brief was added later (the Members tab); it is counted in the cost, not in this time.

The 30 written checks, run by script and by hand against the preview address after the build. The four payment checks (C24 to C27) need Stripe and were not run. The 26 that ran: sign-up, log-out, log-in and password reset; the instructor and admin screens; the timetable with times, places left and the seeded facts; booking, My bookings, cancelling and the credit coming back; the full-class refusal and the no-credit message; one member never seeing another's bookings; the instructor's own classes, members and attendance; the admin adding, editing and deleting a class, seeing every booking, the CSV and the count per class; the Ask the studio form and the admin seeing it; and the booking and admin emails.

All 26 passed. A seventh member was refused on the full Power Reformer class with 'This class is full' and kept her credit. A member with no credits was told 'You need a class credit. Buy a 5-class pass to book.' The CSV had the columns booking, class, day, time, duration, instructor, member name, member email, status, attended and booked at.

Nothing was invented: the three classes, their days, times, lengths, capacities and instructors, the $95 pass, the $24 single class, the address, phone and email all match the brief. The $24 class is information only, as I told the agent. Monday and Wednesday Reformer Basics appear as two rows.

• I could not pay with Stripe, so I added credits through the Members tab I asked for. Seven test members were made by me.

Sign-up, log-in and password reset for a member, log-in for both instructors and the admin, then the two attacks from the brief, run with the app's own requests from a second member's session, plus a sign-up that asked for the admin role.

Attack 1, a member reading another member's booking by its number, returned 404 Not Found; cancelling another member's booking returned 404; the member's own list ignored a member id I added and returned only their own. Attack 2, a member calling the admin delete-class action, returned 403 Not allowed, as did the admin bookings, CSV, members list and the add-credit action.

A sign-up request with role admin created a member. The admin and instructor logins exist only because the agent seeded them. The password reset email arrived within a minute and the new password worked. Emergent's support confirmed that roles live in the app's code, not in a setting.

• Emergent's own security scan was not run: it is not part of the free plan flow I saw.

The six data checks: the form and bookings write real records, they survive a reload, a redeploy and a design change, the admin can view, filter, edit and delete them, and they export as a file.

Proved (four of the five I could run): the Ask the studio form and the bookings wrote records; they were still there after reloads and after the colour change; and the admin's CSV of bookings downloaded with the right columns. The admin also viewed, edited and deleted classes in the app, and the delete message says active bookings are cancelled and credits returned.

Failed: the admin cannot filter the bookings list. Not run: the redeploy check, because I could not publish. Support says the whole database can be dumped from the Republish screen on any plan, which I did not open.

• The data panel of the builder was not opened.

Publishing was tried on the free plan. Speed was Lighthouse mobile on the preview home page, three runs; layout was checked for sideways scrolling at 360, 768 and 1440 pixels.

Pressing Publish opens a Choose your plan screen: Standard Trial at $1 for seven days with 100 credits that renews at $20 a month, Standard, and Pro at $200 with free deployment. Nothing was clicked. The app stayed on its preview address.

The three Lighthouse runs scored 39 each. The main content took about 44 seconds on a simulated phone, the page weighs 8.7 MB over 100 requests, layout shift was 0.02 and blocking time about 650 ms. There was no sideways scroll at 360, 768 or 1440 pixels.

• No domain connected, no real phone, no 100 seeded classes, no layout fault count beyond sideways scroll.

One email to support@emergent.sh from info@rankdude.ai on 27 October 2026 at 18:56 UTC with three questions: how to stop admin sign-up, how to export data and code, and how to cancel and delete the account.

The reply came at 19:02 UTC, 6 minutes later, from the Emergent Support Team. It answered all three: roles are set in the app's code and the agent should be told that public sign-up must always create a member; database records export on any plan from Republish, Database and the MongoDB viewer, and code goes out through GitHub, GitLab or Bitbucket on Standard and above; and cancelling is Profile icon, Account Settings, Plans and Invoices, Manage Plan, Cancel Subscription, with the plan active to the end of the period. Deleting the account starts a 45-day grace period.

It noted my account had no subscription, so nothing could be cancelled. I did not test the cancel.

• The cancel and the final invoice were not tested, because no plan was bought.

The pricing page, the plan chooser opened by Publish and the credit meter were read on 27 October 2026, with a screenshot at each step. Emergent charges in credits.

The Free plan cost nothing. The pricing page lists it as 10 free monthly credits; the home screen showed 10 of 10 before my first prompt, then 23.21 of 50 after the build, so I started with 50. The first build used 26.79, the Members tab 10.74 and the colour change the last 12.47. After that the chat said I had used all my credits and offered Upgrade Plan.

Standard is $20 a month for 100 credits and Pro is $200 for 750. The plan chooser also offers a $1 seven-day trial with 100 credits that renews at $20 a month. The chooser says extra credits can be bought on Standard, and I saw no spending cap. At the rate I measured, a build and twelve changes would need about 27 plus twelve times about 11, roughly 160 credits, more than Standard's 100. I did not read the running cost at 1,000 bookings.

• Yearly prices were not read. The credits-per-change estimate comes from two changes.