Welcome to chapter twenty-two, the capstone. Everything in the book is now applied to one small object you can actually finish. The point is not the object. The point is the portfolio: a documented piece of work whose reasoning is visible, whose evidence is honest, whose limitations are stated, and which a stranger could pick up and understand. That is what you take to a conversation about supervised work, and it is worth more than a stack of certificates.
One sentence sums up this chapter: a finished, humble, well-documented project beats an ambitious unfinished one every time. Over four weeks you will confirm a brief, build a cheap first version, test it honestly, have it reviewed, decide whether better fabrication is justified, and hand it over. Each step uses something from an earlier chapter: need statements, requirements and hazards, the teaching design of the last chapter, and the making methods of chapter fourteen.
Choose one route. A training model, with the running elbow model as the default, partnered with a simulation educator. An assistive technology prototype: a tabletop mock-up of an adaptation discussed with an occupational therapist and the intended user, where any trial follows the team's review process. Or a software or instrumented training tool, using fictional data or a battery-powered sensor giving the learner feedback, with no connection to a person or to live systems. Whichever route you choose, the making methods are open, and most good projects use several. The choice of method comes after the objective, and is recorded with its reason.
If you have no partner yet, complete a clearly labelled independent demonstration project. Use the elbow model with a fictional brief, finish all the technical work, and record every sponsor-dependent criterion as pending. That is an honest and perfectly respectable portfolio. "I completed this independently; items four, seven and nine await educator review" is a far better position than a project whose review was imagined.
Week twenty-seven: confirm the brief.
Here is the default brief. The need: beginning learners need to distinguish a force source from a tendon's transmission role during a simplified elbow movement. The scope: a manually operated tabletop teaching aid that demonstrates bending and is reset by hand. It does not attach to a person, support body weight, model a patient's range of motion, or reproduce every structure. The initial success criteria: the movement is visible, the labels distinguish force generation from transmission, an unfamiliar learner can attempt the questions from chapter twenty-one, and it can be reset without tools. These are criteria to discuss with an educator and get corrected before building, not device specifications.
Before you cut anything, settle these in writing. The need, stated without a solution in it and confirmed by the sponsor. The scope, including an explicit list of what it is not. The relevant anatomy and its level of simplification. The reviewers, named by role. Three to five measurable requirements with acceptance criteria agreed now. The contact category, which for this project is no body contact. Cleaning and storage, agreed with whoever will host it, and who owns it after handover. And what finished means: the specific deliverables and the date.
Do not design one thing. Design two, compare them explicitly, and record the reason for the choice; it takes an hour and is the difference between a decision and a default. Concept A is open card and cord. Concept B is an enclosed fabricated mechanism, machined, moulded or printed. B wins on durability, cleanability and perceived quality, but none of those are yet requirements, and perceived quality is irrelevant to the objective. A wins on the two decisive criteria: the force path is visible, and the risk of teaching the wrong thing is lower. So the choice is Concept A. Write that paragraph down; in three months you will not remember why.
Weeks twenty-eight and twenty-nine: build the first version, cheaply.
Begin with card, even though you have a well-equipped workshop. The temptation is to use the machine tools, the laser cutter, the printer or the electronics bench. Do not, yet. The first question is whether the relationship can be understood, and that is answered faster and more cheaply in card. A quick version lets you change the layout five times before you invest in a polished object that is polished in the wrong shape. You need two stiff card strips about two hundred by thirty-five millimetres, a paper fastener as a pivot, a short non-elastic cord, a small smooth guide loop, and labels. The dimensions are classroom conveniences, not anatomical measurements.
The build sequence. Make the two segments and join them at a pivot about twenty millimetres from one end, loose enough to rotate, with sharp corners rounded. Support the upper-arm strip vertically above the pivot with the forearm horizontal, and label both as schematic. Attach the cord to the forearm about forty millimetres from the pivot, reinforcing the card locally, because that is where it will tear. Fit the guide loop about eighty millimetres above the pivot and pass the cord through, so pulling gently bends the model. Operate it slowly, stop before the strips collide, and add no weights. Then label it, reset it by hand, and mark it version zero point one with the date.
Labelling is the most important step in the build. The free pull end is labelled as the external force input representing muscle action. The cord is the tendon representation: it transmits force and does not generate it. The pivot is a simplified joint that omits cartilage, capsule, ligaments and forearm rotation. The guide loop is a model convenience, not an anatomical tendon route. And the whole object is version zero point one, a tabletop education prototype, schematic and not to scale. If the educator wants to teach actual tendon attachments, the design needs substantial revision and anatomical review. That is a different project.
When something goes wrong, ask what changed. If the strip barely moves, check pivot friction by moving it by hand, then look at whether the cord's line of action passes nearly through the pivot, leaving no moment arm. If the card tears, the load is concentrated, so reinforce and spread it. If the cord slips from the guide, revise the guide and re-test from the beginning, because it is now a new version. If the learner calls the cord a muscle, the source and the transmitter look like one thing. And if it works for you and not for the educator, that is the most valuable failure you will get: you know things the object does not communicate. Change one feature at a time, so you know which change helped.
Run a short bench review. Inspect the pivot, edges, cord and attachment, then repeat the movement and reset at gentle force for a modest count, ten or fifty cycles, recorded explicitly as an exploratory check. Write the result in the form on this slide: how many cycles of how many planned, under what conditions, what you observed including the unexpected, the narrow conclusion it supports, what it does not establish, and the next evidence you need. And never fill in a successful result before the work. It sounds absurd, but it happens constantly, as a template filled in optimistically and never confirmed.
Two reviews, in this order. First, an educator reviews the anatomy, the labels and the stated limitations, and you record their corrections verbatim. If they say the labelling could still be misread, that is the finding; do not argue it away. Second, a learner trial in an appropriate setting: record the learner's prediction before the demonstration, their explanation afterward, every prompt you gave, any confusion in their own words, and whether they named a limitation. Record no personal information you do not need. A likely outcome is that they understand flexion but think the cord supplies its own energy. That is a precise, actionable finding, not a failed project.
Now the question every maker is itching to answer: when is better fabrication justified?
Move beyond card when you have a reason. The card cannot survive the intended handling, shown by a failure you recorded. A clearer or more precise geometry is needed, and you can say which. An agreed classroom requirement calls for a stable assembly. A cleaning requirement has emerged that card cannot meet. Or the learner trial showed something the card version cannot make visible, such as the force source and the cord still looking like one thing. "It would look more professional" is not a reason. A complicated part is not inherently a better teaching aid, and a polished object can teach worse than a crude one by hiding its mechanism.
The elbow model is several elements with different jobs, and each has its own best method. The segments might be laser-cut acrylic or plywood for fast iteration with engraved labels, machined aluminium for durability, printed if a shaped segment is needed, or moulded only if you need many identical kits. The pivot is almost always a turned or bought part: a pin and bushing in brass, stainless steel or acetal, for low, repeatable friction. The muscle might be a sewn fabric sleeve or a cast silicone segment, kept visibly separate from the cord. The tendon is low-stretch braided cord, or stainless cable if stretch becomes a recorded failure. And optional electronics, an angle sensor with a display, only if the objective needs a number.
A plausible version one point zero might be laser-cut acrylic segments with engraved labels, a turned acetal pivot bush on a stainless pin, a sewn red fabric muscle sleeve clearly separate from a white braided cord, and no electronics. Every one of those choices should carry a sentence saying why, and one of them will turn out to be wrong, which is what version one point one is for. And remember: a material change is a new version and changes more than you expect. New weight, new edges, new friction, new failure modes, new cleaning requirements. Repeat every check the change could have affected, and do not carry forward the card version's results.
Week thirty: review, revise, and hand over.
A future volunteer, or you in a year, should be able to pick this up and know what it is for and what has been established. The handover holds the need, the objective and the authorised scope. Drawings and design files for every process used, with revision identifiers. A bill of materials. Assembly and process notes detailed enough to rebuild it, including the firmware version flashed. The version history with reasons, the requirements and test records with their failures, the feedback log with prompts recorded separately, the limitations statement, the current status, and the next decision with who makes it. Then hand it to someone who has not seen the project. Every gap they hit is a documentation defect.
The status statement is one honest paragraph, like the one on this slide. It says what the object is, what it shows, what it omits, and what remains pending, here the infection prevention review for shared classroom use. Update this statement only when the evidence changes, not when your confidence changes.
You must be able to cover the project in five minutes to someone who has not followed it: the need and who said so, the relevant anatomy correctly, the design decision you made and the alternative you rejected with the reason, the evidence including at least one failure, at least three limitations, and the next decision and who makes it. Then ask them: what is the biggest unresolved question here? Your ability to welcome that question, and act on it rather than defend against it, is a substantial part of what readiness for collaborative work actually means.
You have finished the capstone when a sponsor confirms the prototype addresses the agreed need, or that is explicitly recorded as pending. When every agreed test has a recorded outcome, including failures and unresolved requirements. When someone else has reconstructed the design from your documentation alone. When you can give the five-minute explanation and state three limitations unprompted. When the version history explains why each element is made the way it is. And when no prototype is described as approved for patient use unless the responsible organisation has actually authorised it.
The limits. A completed capstone is not a qualification; it is portfolio evidence of your process, which is exactly what it is meant to be. A sponsor's confirmation that it meets the need is not approval to deploy; that is a separate institutional decision. Your bench results characterise your object under your conditions, nothing more. Finishing does not mean you may now work unsupervised; it means you are ready for the next conversation. And a portfolio recording three failures, two pending reviews and one unresolved requirement is more credible than one in which everything passed. Next, next steps, specialisms, and finding a place to contribute.