Where we are
Session a1 produced an objective naming an ability. Tonight you decide what would count as proof, before you have written a single slide - which is uncomfortable and is the point.
★ What you walk out with tonight. The assessment for your objective, a content list derived from it rather than from your expertise, and a written not-this list that stops the course growing in week three.
Starting from what you know produces a syllabus, not a course
6 min liveThis is not a criticism of experts. It is what expertise does: once you know a subject, everything in it looks connected and necessary, and you genuinely cannot see which parts a beginner can skip.
LiveThe question, and what it does to a draft4 min▶
"What would prove they can do it?" Ask it of your objective before anything else, and then ask it of every piece of content you were planning to include.
For Daybreak's objective - classify a proposed schema change as safe or breaking - the proof is obvious the moment you ask: ten real schema changes, sorted, with the ambiguous two argued out. That is the assessment, and it took one question.
Now run the content list past it:
| Planned content | Does it serve the proof? |
|---|---|
| The history of data contracts | No. Cut |
| Why the last three incidents happened | Partly - one example, not three |
| The six change types and which break | Yes. This is the core |
| Our contract file format | Yes, briefly - they have to write one |
| The full schema registry architecture | No. Interesting, not required |
| Deprecation notice periods | Yes - it is in the objective |
Six planned items, three survive intact. That is a normal ratio and it is where the hour comes from.
10 min history and why it matters · 15 min the six change types · 10 min our registry architecture · 10 min the contract file format · 10 min the last three incidents · 5 min questions
Nothing here is wrong and every minute is defensible. Nobody classified a single schema change, so on Monday nobody can. The quiz at the end tests whether they remember what the six change types are called.
5 min one incident, as the problem · 10 min the six change types with worked examples · 25 min classify ten real changes, argue about two · 15 min draft a deprecation notice and swap for review · 5 min the job aid
Forty of the sixty minutes are them doing the thing the objective names. The assessment is not at the end - it is the middle, which means you find out who cannot do it while you are still in the room.
The not-this list, applied to a curriculum
5 min liveCourses do not grow because designers are undisciplined. They grow because everything suggested is genuinely relevant, and there is no written record of what was already decided against.
Self-studyThree tests for whether content earns its minutes4 min▶
- Does it serve the assessment? If somebody could pass the proof without this, it is optional - which does not mean worthless, it means it belongs in the handout rather than in the room.
- Would they look it up anyway? Reference material should be a job aid, not a memorised thing. Notice periods, the file format, the exact syntax - hand them over, do not teach them.
- Is it here because it is hard to explain, or because it is hard to do? Hard-to-explain things need your time in the room. Hard-to-do things need their hands, and the two require completely different amounts of your talking.
The move that makes this survivable politically: put the cut material in a handout and say so out loud. "That is in the pack, it is not in the hour" is a much easier sentence than "we are not covering that", and it is usually true.
Write the assessment before you write anything else
★ 22 min · everyone buildsRule for the next twenty minutes: you may not write a single piece of content. Objective, proof, practice, and a not-this list. Content is session a3 onward.
Objective. Your 4-of-4 sentence from a1. If it did not pass the linter, fix it now, because everything below inherits its weaknesses.
Proof. What would you have to watch them do to believe it? Write the task, not a quiz. For Daybreak: ten real changes, sorted, two argued out.
Practice. What is the smaller version of the proof that they do first, with support? Usually the same task with three items instead of ten and a worked example in front of it.
Content. Only now. List what they must be told in order to attempt the practice. It is always shorter than you expected and that is the point of the ordering.
Not this. Everything you nearly included, with one line each on why not. Send this to the sponsor with the design, not after somebody asks.
Write your proof task in one sentence. "Given X, they will do Y, and I will look at Z." If you cannot write it, the objective is still too vague - go back rather than forward.
Check the proof against the objective's condition. They should be the same thing. If your objective says "given a proposed schema change" and your proof is a multiple-choice quiz about terminology, one of the two is lying.
Design the practice as a smaller proof. Same shape, fewer items, more support. Practice that is a different shape from the assessment is the single most common design error and it always surprises people in the room.
Now list the content, and time it honestly. Most people find they have 10 to 15 minutes of genuine telling. If you have 40, run each item past the three tests.
Do the arithmetic. Content minutes plus practice minutes plus proof minutes plus slack. If it exceeds your slot, cut content first - never practice. Cutting practice to fit content is how you end up back at the forwards version.
Write four not-this lines. Each with a reason. Then send them to whoever asked for the training, before they ask.
Say the sentence out loud. "That is in the pack, it is not in the hour." Practise it, because you will need it and it works much better than a debate about scope.
If you take one thing from tonight. Write the proof before the content. It is fifteen minutes of discomfort and it removes about a third of a course, permanently, without anybody having to argue about taste.
Before session a3
25 min- Write the proof task for your own objective. One sentence: given X, they do Y, I look at Z.
- Take an existing course and run the question over its contents. "Does this serve the proof?" Count what survives. A third is normal.
- Check one course's practice against its assessment. Same shape? Different shape is the common error.
- Write four not-this lines and send them to a stakeholder unprompted. Note how differently the week-three conversation goes.
- Bring your content list to a3, because a3 is about how little of it anybody can actually hold at once.
Three questions
Clear all three and this session gets stamped in your passport.
1 · Why does starting from what the expert knows reliably produce too much content?
It is a property of expertise rather than a character flaw. The forwards sequence contains no step capable of removing anything, so nothing is removed. Backwards design supplies that step: every item has to justify itself against the proof.
2 · Your objective says "given a proposed schema change, classify it" and your assessment is a multiple-choice quiz on terminology. What is wrong?
The condition in the objective tells you what the assessment must look like: they get a proposed change and they classify it. Multiple choice is fine as a format when the objective is about recognising something - here it tests a different ability from the one you promised.
3 · You are over time. What gets cut first?
Practice is where the ability is built and the assessment is how you find out in the room. Content is the part that can move to a handout with a sentence: "that is in the pack, it is not in the hour." Cutting the doing to preserve the telling undoes the entire session.
What this session teaches, and where it came from
Every fact on this page, and its verification tier, is recorded in the course's source map. Daybreak is a teaching scenario.
Designer a2 cheat sheet · pin this
You have an objective, a proof and a short content list. Session a3 is about the constraint that makes all three survivable: nobody can hold very much at once. Working memory, worked examples and when to take the scaffolding away - plus the reason most training that feels like a motivation problem is actually a load problem.