Is this really teachable in four hours?

A plan that adds up to 240 minutes is not the same as a class that fits in four hours. The build checks the first. This page is the check on the second, and it is here so that anybody can redo it rather than take my word for it.

The method#

The question cannot be answered from inside this course, because a course can always claim its own content is deliverable. So it is answered against two courses that have actually been taught: Electronics for Theatre, sixteen classes of four hours, and Show Networking and Control Systems, eight sessions of three hours.

Two measures, both taken from the markdown itself:

Density. Words of taught prose divided by the minutes given to taught blocks. This asks whether a block has more in it than can be delivered in the time. Preparation, the plan, the numbers table, the misconceptions and anything marked Extension: are excluded, because none of those is delivered from the front of a room.

Ratio. What share of the class is at the keyboard or the bench rather than listening. This asks a different question: not whether it can be said, but whether a room of technical direction students will still be with you when it is.

What it found, and what changed#

The density was never the problem. At 15 to 23 words per taught minute the first draft was consistently less dense than either course that is already taught, which is the opposite of the failure everybody expects. There was more time per idea here than in a course that survives contact with a room.

The ratio was the problem, and it has now been fixed twice.

The first fix moved fifteen minutes from talking to doing in every class and turned Class 2's language comparison into reading. That took the bench share from 45 to 52 per cent, which was better and was still under Electronics.

The second fix is the one that actually settled it: the course now has somewhere to do the work. An editor, a runner and a console that live in the class page, sixteen coding exercises whose tests are the brief, and six block challenges for the same subject with the syntax removed. Seventeen of them sit inline in the prose at the point the idea is taught, so a student who has just read what a blocking wait costs writes the non-blocking version in the next paragraph, on any laptop, with nothing installed.

That changed three things at once:

1. Bench time reached 60 per cent, five points under Electronics and comfortably above the level at which a room of technical direction students stays with you.
2. A fifteen minute logic gym opened every class except the capstone. On paper, in pairs, no keyboards, with the rule that you say the answer out loud before you check it. It is the cheapest block in the course and it is the one that trains the thing everything else depends on.
3. Density fell rather than rose, to 12 to 17 words per taught minute, because most of what was added is work rather than exposition. The pages got longer; the talking did not.

Where the extra content went, and why that is not cheating#

Three reference pages now carry material that would not survive being delivered from the front of a room: boards, systems and the model box. None of it is examinable by recall and none of it is on the plan for any class.

That is a deliberate structure, not a way of hiding the overrun. A student asks "which board" and "buy or build" during a production, not during a class, and a page they can open on their phone in a get-in is worth more than four minutes of somebody talking about it in March. The classes point into those pages at the moments the question arises, and the pages point back.

The test of whether that is honest: could the class be taught if the pages did not exist? For all six, yes. Nothing on the plan depends on them. They are depth for the people who want it, which is the correct place for depth.

What is honestly still tight#

Class 2 is the fullest. It carries what a language is, types, references, floating point and notation, and it is the only class where a block running long has nowhere to go. If it overruns, the block to shorten is references and mutation, and the recovery is that the shared-cue bug is a card in the read-the-code deck and an exercise in the lab, so it comes back twice.

Class 4 still has the most exposition at 85 minutes, because the argument about how these models work does not compress well and cannot honestly be turned into a lab. It is the one class where the gym is doing the most work.

Class 5 depends on preparation more than the others. Nothing in it fails gracefully if a student arrives without a working toolchain, because the labs are the class. This is why the preparation tasks are specific and why the readiness check tests them. The in-page lab now softens this: a student whose board will not flash can still do the blocking bug in blocks and in code, and catch up on the hardware afterwards.

The gym is the block most likely to be cut when a class runs late, and it is the wrong one to cut. It is fifteen minutes and it is the only part of the class that trains holding a state in your head and committing to an answer. If something has to go, take five minutes off each lab instead.

Redoing it#

The audit is arithmetic over the markdown, so it can be repeated after any edit:

• taught minutes per class: the Run of the session table, rows whose block is The idea
• keyboard minutes: rows whose block starts Lab, which now includes the gym
• taught words: everything except preparation, the plan, the numbers, the misconceptions and Extension: sections

If a class drifts above about 30 words per taught minute, or below about 50 per cent at the keyboard, something has been added that should either become an extension, move into preparation, move to a reference page, or move to another class. The fix is never to add minutes: the class is four hours whatever the markdown says.

One more check, which is not arithmetic and matters more than either measure: open the class page, scroll it, and count how far you get before something asks you to do rather than read. If that distance is more than about two screens, the balance has drifted back, whatever the numbers say.