Before the Torch
On 2026-10-30 this training passed 7 of 7 live-site checks and played to the end on 8 of 8 screen shapes with no layout blockers. 31 of its 31 steps can be finished with the keyboard alone.
1. The live site
Opened at its public address in Chrome and in WebKit (the engine inside Safari) set up as an iPhone. The last check compares the bytes the internet serves with the file that was tested, so this page describes exactly what you get.
| Result | Check |
|---|---|
| pass | answers 200 over https 200 |
| pass | the engine BOOTS: running, not merely served |
| pass | no page errors |
| pass | the deck is real 3 scenes, 31 steps, 80 clips |
| pass | loads nothing from another host |
| pass | the edge serves the bytes that were tested 18178141 bytes |
| pass | iPhone (WebKit, iOS Safari UA, 390x844): boots and the deck is real 3 scenes, 31 steps |
Build fingerprint (SHA-256 of the page the site serves): 83ac9dde2b92e53b317fdb8dc54a01bfa9afc76ef9ff0aa1e1095f3b677a1190, 18,178,141 bytes.
2. Phones, tablets and a laptop
The whole training was played start to finish at each screen shape, answering every step the way a trainee does, and the layout was measured before every screen: a control that cannot be reached or is covered, a page that scrolls sideways, a card too small to read. Those are blockers. Warnings are smaller things, like a target under 40 px.
| Screen | Result | Layout blockers | Warnings | Screens measured |
|---|---|---|---|---|
| iPhone SE, upright (375 x 548) | played to the end | 0 | 315 | 75 |
| iPhone 13/14/15, upright (390 x 664) | played to the end | 0 | 307 | 75 |
| Android phone, upright (360 x 680) | played to the end | 0 | 311 | 75 |
| iPhone held sideways (844 x 340) | played to the end | 0 | 301 | 75 |
| Small phone held sideways (667 x 320) | played to the end | 0 | 292 | 75 |
| iPad, upright (820 x 1106) | played to the end | 0 | 656 | 75 |
| iPad, sideways (1180 x 746) | played to the end | 0 | 578 | 75 |
| Laptop (1366 x 768) | played to the end | 0 | 576 | 75 |
3. Accessibility
Keyboard only. The training was played with no mouse: Tab, Space, Enter and the arrow keys. 31 of 31 steps were finished that way.
Focus. 29 Tab stops were visited. Every one showed a visible focus indicator. After 1 of 49 key presses the focus went back to the top of the page.
Narration. 30 of 30 spoken lines were on screen as text while they played.
Zoom and reflow. 320 CSS px (SC 1.4.10): no sideways scrolling, main button on screen; 200% zoom (SC 1.4.4): no sideways scrolling, main button on screen. With the WCAG text-spacing override, no text was clipped.
Automated rules (axe-core 4.10.3, WCAG 2.x A and AA), run on the first card, on the first step of every kind and on the finish screen. Rules that failed at least once:
- No rule failed.
The full conformance report for the player is the accessibility conformance report (VPAT).
4. What this training says about itself
Pictures. One generated flat picture of a house side yard (not a 360 image), shown as three still scenes.
Told to the trainee: “This picture is generated, not photographed: nobody in it is a real technician”
Declared limits
- No step asks the trainee to judge the work in the picture; it is a place to stand, not evidence.
- Nothing is named from a label: the two grey boxes are never called disconnects, and what is in the black cylinder cannot be read.
- Clearance, torque and vacuum figures belong to one heat-pump family; the instructions for the model actually installed outrank them.
- It deliberately prints no date for the refrigerant transition, because that rule has moved repeatedly; check the rule in force on the day.
- The employer must supply their own lockout procedure, the refrigerant in use and any local code amendments.
- It has not been accessibility-audited.
5. What was not tested
- Real handsets. The screen shapes are Chrome set to each size, with a phone's touch and toolbar behaviour simulated. The iPhone check runs WebKit on Windows, not an iPhone.
- A screen reader (NVDA, JAWS, VoiceOver) played through the whole training by a person.
- Firefox, and Safari on a Mac.
- A slow or dropped network while the training plays.
- A VR headset. This page covers the browser player only.
- Whether the teaching is right for your site. That is a subject-matter review, not a test.
Machine-readable record of this page: hvac-install.json. Every number on this page comes from that record, and every value in it comes from a test run on 2026-10-30. Nothing is typed in by hand.