Advanced generation settings
Most of the time you should not touch any of this. You describe what you want, Zippy reads the subject, level and topic out of your description and your scope, and picks the rest. Then it shows you the plan — including what it chose and why — before it writes a word.
The settings behind the gear are there for the times you already know better than the default. Anything you set explicitly is treated as binding: Zippy will not quietly override it.
Pedagogy
The single most consequential setting, and the one most worth understanding.
A generated resource is only as good as the shape it takes. A lesson that teaches a procedure — long division, semicolons, balancing an equation — needs a different structure from one that builds a concept through investigation. Give both the same shape and one of them will be wrong.
Zippy chooses from seven structures. Auto is the default, and it is the right answer more often than not.
| Structure | Best for | The sequence it follows |
|---|---|---|
| Auto | Everything, unless you know otherwise | Zippy picks from the subject, level, topic and your description, then names its choice on the plan card |
| Backward design | Unit plans, exam preparation, anything that has to align to a syllabus | Outcomes → evidence → learning experiences |
| 5E | Science and STEM, concept formation, phenomena students can investigate | Engage → Explore → Explain → Elaborate → Evaluate |
| Gagné's nine events | A general-purpose full lesson when no subject-specific shape fits | Attention → objective → recall → input → guidance → practice → feedback → assess → transfer |
| Explicit / gradual release | Procedures and techniques: writing craft, grammar, maths methods, exam skills | I do → We do → You do together → You do alone → Review |
| Workshop | Writing, reading, art — anything where students need working time | Mini lesson → guided practice → independent work → share |
| Discussion / seminar | Literature, humanities, ethics, anything argued rather than solved | Stimulus → norms → question → structured discussion → synthesis |
| Problem-based | Maths, science, design, case studies — transfer and application | Problem → assumptions → strategies → attempt → feedback → generalise |
What Auto actually picks
If you leave it alone, these are the defaults by subject. They come from the same research base as the rest of Zippy's generation — Understanding by Design, BSCS 5E, Rosenshine's Principles of Instruction, and the gradual release model.
| What you are teaching | Auto chooses | Because |
|---|---|---|
| Primary English writing | Workshop + explicit instruction | Students need a model, then criteria, then time to write and revise |
| English reading and literature | Discussion, or Gagné | Comprehension is built through evidence-based talk, not a worked example |
| Grammar and language skills | Explicit instruction | Examples and non-examples, common errors, short practice, immediate feedback |
| Mathematics | Explicit instruction + worked examples | Prior knowledge, worked example, guided practice, then independent work |
| Science | 5E, with explicit concept consolidation | Students investigate the phenomenon before the formal explanation |
| Humanities and social studies | Inquiry + discussion + source analysis | A big question and source evidence, then claim–evidence–reasoning |
| Exam revision | Backward design + retrieval practice | Skill gaps first, then timed practice on the traps that actually appear |
When to override
Three situations are worth the click:
- Your class has already investigated the phenomenon. 5E asks students to explore before they are told. If they explored it last lesson, switch to explicit instruction so the lesson consolidates rather than repeating the discovery.
- You are short on time. Workshop and problem-based both need working time to be worth anything. For a twenty-minute slot, explicit instruction gets further.
- Low prior knowledge. Problem-based and inquiry structures assume students have something to reason from. When they do not, the lesson stalls. Explicit instruction with worked examples is the kinder choice.
One structure is not "better" than another — the failure mode is using the same one for everything, which is exactly what Auto is built to avoid.
Number of questions
Binding when you set it. Zippy will produce exactly that many, rather than adding one more to feel complete or dropping one to fit.
Left on Auto, the count follows the structure: a story-led lesson closes with one reflective question, a practice set carries eight to twelve, an exit ticket is two or three. A lesson that is mostly questions is a worksheet — if that is what you want, generate a worksheet.
Question types
Which kinds of question to use: MCQ, short answer, cloze, or open-ended. On Auto, Zippy mixes them by what the topic can actually assess — a spelling rule wants a cloze, an inference wants an open-ended answer with a mark scheme.
Multiple-choice questions get distractors tied to real misconceptions rather than obviously wrong answers. That is what makes a quiz diagnostic instead of decorative.
Difficulty
Easy, medium, hard, or mixed. Mixed is the default and usually right: a set that is uniformly hard tells you who is struggling but not why, and a set that is uniformly easy tells you nothing at all.
Difficulty is calibrated against the papers in your scope, not in the abstract — so "hard" for P3 Science means hard for P3 Science.
Duration
For lesson plans and presentations. Zippy assumes 45–60 minutes for a classroom lesson and 15–25 for an online micro-lesson. Setting it explicitly changes how much gets planned, not just the number printed at the top — a 20-minute plan drops the extension task rather than compressing everything.
Inline AI feedback
On by default. Every question gets an evaluation prompt written for that specific question, so a student who answers gets a response that addresses what they actually wrote.
Turn it off for printed material, or when you want to mark the work yourself.
What is not a setting
Two things people look for here and will not find:
- Grounding. Zippy always grounds generation in the real material in your library that matches your scope. You can see what it found — and correct it — from the line under the composer, before you generate. See Generating a lesson.
- Images. Generated media is always opt-in per image. A draft comes back with image placeholders carrying the prompt; you click Generate on the ones you want. Nothing is generated and charged for without you asking.