The last percent#
SLOP2805 · Week 12 · Showcase briefing
Eight minutes. Three paths, live.
Then two challenges on a claim you chose to keep.
Where we were, and where we are going#
Twelve weeks ago you watched a bar sit at 99% for eleven seconds and then fail. You could say something was wrong with it and not much more.
Today you show an interface that does not do that — and you take questions about the parts of it you could not make honest.
There is nothing new to build. The written kit is due tomorrow.
Why a defence and not a presentation#
Everything else in this course is marked on artefacts you control.
A defence tests the one thing an artefact cannot show: whether you know why.
A well-built interface whose maker cannot say which of its numbers are invented is a weaker outcome than a rougher one whose maker can.
The eight minutes#
| Time | What you do |
|---|---|
| ~4 min | drive the deployed site through a success, a stall, and one failure, using the exposed controls |
| ~1 min | name one kept with reason row, its reason, and its cost |
| ~3 min | two challenges from the room |
Live, on the deployed URL. Not localhost. Not a recording.
Three answers that score#
“It holds.” — that string comes from the acknowledgement event, here.
“It doesn’t hold.” — I kept it because of X, and it costs Y.
“I don’t know.” — and here is how I would find out.
The third is a full-credit answer. It is the sentence week 11 spent an afternoon teaching.
One answer that does not#
“It felt better.”
Not because preference is illegitimate — week 6 was an entire studio on preference — but because
“I preferred it and I have no evidence”
is available, true, and takes the same breath.
The room has been trained to call it since week 1.
Worked example — a defence that earns the marks#
Claim: the verification state is held on screen for 400 ms when it lasts ~80 ms.
Challenge: “So the interface is lying about a duration.”
Answer: “Yes, by up to 320 ms, and the register says so. I kept it because at 80 ms the state is unreadable — it reads as a flicker, and two testers in week 11 asked what had flashed. The cost is that anyone comparing it with the elapsed counter sees them disagree. If I had another week I would hold it only when the stage is shorter than 200 ms, rather than always.”
Names the deception, the evidence, the cost, and the thing not done.
Your turn — choose the uncomfortable row#
Pick the kept with reason row you genuinely find hard to defend.
A safe choice produces a safe two minutes and a low Defence mark. The room is not hostile and nobody is trying to catch you out — the criterion rewards the difficulty of the question you volunteered for.
The mistake to expect#
Rehearsing the happy path.
A demo that only shows a success has not shown the course. The three paths are in the spec line, and a stall you cannot trigger on demand is a stall you cannot claim to handle.
Second: demoing from localhost because the deploy is slow. The deployed URL is
what the criterion names, and base-path bugs only appear there.
What this week cannot tell you#
Whether you were right.
A defence establishes that you can say why, from evidence, and that you know which decisions are unsupported. It does not establish they were good.
What you take away is smaller and more portable: you can look at any waiting interface — including one you are about to build at work — and say what it claims, what could support each claim, and which parts were invented to fill a silence.
What leaves the room#
- a defence, worth 10 marks of Assignment 3
- your week 1 claims table, read one last time beside your register
Tomorrow, 17:00: the Honest Waiting Kit. Assembly, not writing — the state reference came from weeks 2 and 8, the register from week 10, the evaluation record from week 11.