This time.
For sure.

Operation Beep Beep: a design review of a capture plan with one very fast problem.

Road Runner (Velocitus unconvincibus)

The inventor asked for this review to approve a larger machine. Let the picture carry the confidence: a tower, a counterweight, a suspended load and, bottom right, the only participant who has not agreed to anything. The review has a narrower job than the pitch: decide what is worth funding next. This is a fictional design exercise for the Rikiki example gallery: recommendations are scenario proposals, not measured results or canonical specifications. No rockets, explosives or cliff edges are proposed.

The goal is a caught bird, not a fired machine.

Mechanism success

The load drops, on time, exactly where it was aimed.

Mission success

The right subject is under the load when it does.

Every previous report measured the left column. This review measures the right one.

The inventor's past reports all conclude that the machine worked. They are correct: it fired. The mission, though, was never "fire the machine". Stating the goal as an outcome on the target changes which evidence counts for the rest of this review.

What we know, and what we are only assuming.

Observed

The bird uses the desert road daily. It inspects unfamiliar objects from a distance. It leaves when approached.

Assumed

It will take the bait. It will stop on the marked spot. It will wait for the cue.
Within this fictional premise, three things have been seen repeatedly and three have only been hoped for. Notice that every assumption concerns the bird's behaviour, and every observation contradicts one of them. That asymmetry is the finding of this review; the rest explains it.

Every arrow in the plan depends on the bird.

Three components are built. The link between them never was.

Read the chain left to right. Bait, trigger and catch are hardware; each has been bench tested and each works. The second box is not hardware at all: it is the bird choosing to stand on the mark. Both arrows touching it are dashed because nobody has ever observed either transition. The whole plan rests on a component that ACME does not manufacture.

As designed, the trigger fires on weight, not on birds.

Bait the mark Weight on the plate Timer releases Load falls Step two never asks whose weight is on the plate. This is the inventor's own sequence, taken from the drawings. Look at step two: the trigger condition is "weight on the plate". It does not ask what the weight is. That detail looks harmless on paper and decides the outcome of the next three slides.

Beat one: nothing takes the bait.

The seed is on the plate. The bird looks at it from the road, then keeps running.

The mark stays empty.

The first beat of the rehearsal reconstruction is an absence. Pause on it. An empty mark is evidence: it is the first place the plan could have been tested cheaply, before any load was ever lifted.

Beat two: the designer walks into the scene.

Nothing happened, so the operator went to look. Follow the markers: from the control table on the left, across the scene, onto the one plate that responds to weight. The drawing never placed the operator inside the scene because the plan assumed nothing would go wrong.

Wait. Now! Oh.

The target inspects the machine from a safe distance.

The designer steps onto his own trigger plate.

Perfectly, two seconds later, on its own operator.

Put the three beats together and click through them. Each one is individually reasonable, and together they catch the wrong subject. Keep the tone dry: the comedy is in the cause and effect, and the room will laugh without being asked to.

The trigger detects position. The mission needs identity.

Component specification

Release when weight is on the plate for two seconds.

Mission requirement

Release only when the target is on the plate.

No refinement of the left column meets the right one. It is the wrong requirement, not a weak part.

This is the causal flaw, stated without the gag. The trigger met its specification every time. The specification itself encoded "something is here" when the mission required "the bird is here". A bigger load, a faster timer or a stronger rope would all improve the left column and leave the failure exactly where it is.

A retry is unacceptable while the operator can be the payload.

Operator in the drop zone

Investigating puts the coyote under the load.

Remove the falling finale

Rehearse with nothing above anyone.

Before discussing alternatives, one condition is non-negotiable. As long as the design has a suspended load and a trigger any visitor can press, the person most likely to be under it is the one who built it. The rehearsal we propose has no falling part at all.

Escalation answers the wrong question.

Catalogue orderRehearse withThe uncertainty it resolves
Oversized anvilAn inert propDoes the cue fire on the right subject?
Rocket assistanceA marked routeDoes the bird follow the route at all?
Emergency spare cliffAn observer outsideWhat actually happened, and when?
The pending catalogue order is shown struck through on purpose. Each item makes the left column of the previous slide stronger. Each replacement costs less and answers one of the three assumptions we listed at the start. That is the whole argument for Revision B.

Revision B tests one thing: does the bird follow the route?

Compare this with the dependency chain earlier. Three nodes instead of four, one label instead of two assumptions, and nothing in it can hurt anyone. The prop is a painted box on the route: if the bird walks up to it, we have learned more than the last six machines taught us.

No bird. No second act.

Rehearsal gate, revision B. Conditions agreed before the day, not results.

These three conditions are agreed now so they cannot be negotiated on the day. The stop condition is the important one: if the bird does not reach the mark, the rehearsal ends and nobody improvises a second act. No results are claimed here; the rehearsal has not happened.

The next review receives an observation log, not a new drawing.

TimeExpectedObservedAction
exampleBird reaches the propBird stops on the road, 30 paces shortStop. Log it.
·Bird reaches the prop
·Bird passes the observer
The first row is an illustrative example of how to fill the log. No rehearsal has been run. This is the only deliverable we ask for: a log with the time, what the plan expected, what the observer actually saw and what was done about it. The italic row shows how an entry should read; it is invented for illustration. The empty rows are what the rehearsal fills in.

Approve revision B.

Less machinery.
More observation.

Fund the rehearsal

A marked route, an inert prop, an observer and a log.

Cancel the anvil

Reopen it only if the log shows the bird on the mark.

The review closes on a spending decision in two halves. Fund the rehearsal, which is cheap and safe. Cancel the escalation, with an explicit condition for reopening it: the log must show the bird actually reaching the mark. Until then, a larger anvil is a larger answer to the wrong question.