The DEEP Build Kit · Card 14
Stretch and constraint cards
Printable cards for purposeful depth: S13 content handoff and S14 two-boundary tests are Level 3 core work; other stretches follow passing checks. Constraints help Level 3 test its own spec.
For: Stretch cards: anyone whose checks all pass. Constraint cards: Level 3 ■ Black draws two for the blank-page spec
Card 14 · Cut-up cards · stretch at the save point (1:15) or in Sprint 2 (2:15–2:35); constraints at the spec (0:40)
Print one set per table, on card if you can, and cut along the lines. Put the constraint cards in the Level 3 ■ Black zone.
How to use them
- Stretch cards: take one when every success check passes, at the save point or in Sprint 2. Write the new success check first. One change per prompt, ending "Keep everything else unchanged. Explain what you changed and how I'd check it." Re-run all your checks after each change.
- Teacher workflows: choose S8, S12, S13 or S14 when they suit the task. Learning-design cards apply when your build is for learning; a preparation checklist needs no retrieval quiz.
- Constraint cards (Level 3 ■ Black): at 0:40, draw two at random. Write each into your spec as a requirement, with its check. Draw again if a card is already in your spec, does not serve your user's task, or clashes with your target rung (for example, "Nothing saved" on a Remember build).
Level 3 core depth: S13 and S14 belong to your main learning plan today, before optional publishing. A colleague changes one named content-data item without changing interaction logic; you test two relevant boundaries and rerun the three original checks after edits or repairs. Explain two coding choices and the difference between content, state and interface. If a core check blocks you, save, repair it and record any unfinished handoff honestly.
Choose purposeful depth: skip any stretch whose behaviour already exists. A personal workflow takes S12 first; do not add retrieval, wrong-answer feedback or Remember against its goal or no-persistence red line. If no check failed in Sprint 2, add a boundary expectation or clarify a requirement, test it and repair only a real failure. Do not invent bugs or add unrelated features.
Stretch cards
| Card | For | Stretch | Do this | Done when |
|---|---|---|---|---|
| S1 | Level 1 ● Blue | A second tool | Paste the same spec into a second tool (card 09) | You have run your three checks in both tools, and written on your station card which passed more |
| S2 | Level 2 ▲ Orange | Climb one rung | Respond → Remember (card 11 has the prompt). A Remember build may store progress in this browser only, never what students type, with a Clear button. Test it where students will open it: browsers differ for a file opened from the computer, and a private window forgets it. | A new check for the Remember rung passes, Clear starts it clean, and your first three checks still pass |
| S3 | Level 3 ■ Black | Publish | Publish on GitHub Pages (card 17) | Three of your checks pass on the live link, on a phone |
| S4 | Level 3 ■ Black | The keyboard and contrast audit | Run the keyboard and contrast audit (card 08) | Every line is logged, and each failure is fixed or written into the spec |
| S5 | Any level | Retrieval at the end (card 08, item 3) | Add one question answered from memory at the end, with the earlier screens hidden (card 06 has the prompt) | The question works, and shows the right answer and why |
| S6 | Any level | Predict first (item 2) | Add a step where the user predicts, and commits, before any answer or result is shown (card 06) | The prediction appears beside the result |
| S7 | Any level | Feedback that explains (item 4) | Make every wrong answer say why it is wrong and what to do next. Name one misconception | You have given every wrong answer once and read each explanation |
| S8 | Any level | Nothing decorative (item 6) | Remove anything that does not serve the goal (card 06) | Nothing moves, sounds or scores unless your spec says why |
| S9 | Any level | Reuse it (item 11) | Decide when your students will meet it again, spaced out and mixed with other topics | Two dates written on your canvas |
| S10 | Any level | A check of learning (item 12) | Write the one question you will ask a week later, beyond "did you enjoy it?" (card 07) | The question is written, with its answer |
| S11 | Level 2 ▲ and Level 3 ■ | Make an agent: the job description | Write the job description and the test for a spec-checker agent that works for you, never for students (card 19). The test is a copy of your spec with one problem planted in it. Build the agent after the session | The job description is written, and your test copy has one planted problem you expect it to find |
| S12 | Any level | Peer handoff and retest | Give a partner the final file or permitted link and your three checks, without a demonstration. Ask them to run it, name one unclear requirement or useful improvement, and record whether you accept it and why. If a check fails, make one fix and ask the partner to retest it | Your partner opens the final artifact and runs all three checks independently; your station records their contribution and, for any repair, their successful retest |
| S13 | Level 3 ■ Black: core | A colleague can change the content | Keep your checked questions or steps in one clearly named data section. Ask a colleague to change one item, answer or explanation using made-up content, without changing the interaction code | The changed content appears correctly, your colleague changed no interaction code, and the original normal, invalid-input and keyboard checks still pass |
| S14 | Level 3 ■ Black: core | Catch two missed boundaries | Write checks for two cases your demo missed: an empty list, repeated Check, last item, Reset, or a valid answer in another form. Predict each result, then have a peer run both. Log the exact actions and results; repair only a real failure | Both new checks have recorded expected and observed results, any repair is independently retested, and all three original checks still pass |
Constraint cards (Level 3 ■ Black: draw two)
| Card | Constraint | Write it into your spec as | Check it by | From card 08 |
|---|---|---|---|---|
| C1 | Keyboard only | Everything works with the keyboard alone, in a sensible order, with a visible focus outline | Putting the mouse away and finishing the whole task with Tab, Shift+Tab, Enter, Space and the arrow keys | Item 10 (WCAG 2.2) |
| C2 | Phone width | Usable at 320 CSS pixels wide with no sideways scrolling | Narrow the browser window to its smallest, or send the file to your phone. | Item 10 (WCAG 2.2) |
| C3 | Worked example first | Before the first question or task, it shows one worked example or a short explanation | Opening it fresh: a worked example or short explanation comes before anything you answer | Item 5 |
| C4 | Not colour alone | Right and wrong shown in words, not colour alone | Ignoring the colours: the words alone say right or wrong | Item 10 (WCAG 2.2) |
| C5 | Three controls | No more than three controls (buttons, sliders or boxes) on a screen, not counting answer options | Counting the controls on every screen | Item 8 (a rule of thumb, not a tested number) |
| C6 | Every wrong answer explains | Every wrong answer says why it is wrong and what to do next | Giving every wrong answer once and reading the feedback | Item 4 |
| C7 | Contrast | All text contrast at least 4.5:1 against its background | Checking each text and background colour pair in a contrast checker | Item 10 (WCAG 2.2) |
| C8 | One idea per screen | One idea per screen | Saying what each screen is for in one short sentence | Item 7 |
| C9 | Nothing saved | Nothing is saved: a reload starts clean | Answering two questions, then reloading: it starts from the beginning | Safe building, 3 |
| C10 | Click, don't drag | Anything the user drags can also be done by clicking | Doing every drag by clicking instead | Item 10 (WCAG 2.2) |
DEEP Education Network · Build with AI · Two rules: no student names or data anywhere; nothing student-facing calls AI.