Blog · · 6 min read

Prototype machine: what your first build has to prove, and what it can skip

A prototype machine is a question you build. Learn the four kinds of prototype, which one to build first, what to leave out, and how to tell when it has done its job.

When people say "I want to build a prototype machine", they often mean a smaller, rougher copy of the finished product. That picture leads to slow and expensive first builds. A good prototype is not a rough copy of the product. It is a question, built in hardware, that gives you an answer you could not get on paper.

This article explains the four kinds of prototype, which one comes first, and what you can safely leave out of each.

One prototype, one question

Every prototype should answer one main question. Write the question down before you order a single part.

  • Can this small fan really hold a charcoal fire at a set temperature?
  • Does the drum still spin smoothly when the load sits on one side?
  • Will a person understand the two buttons without a manual?
  • Does everything fit in the housing once the cables are in?

These four questions need four very different builds. The first needs a fan, a sensor, a grill and some tape. The third needs a cardboard box with two buttons drawn on it. If you try to answer all of them with one build, you wait weeks for parts, and when something fails you do not know which decision caused it.

The four kinds of prototype

Engineers use different names, but the types are nearly always these.

1. Proof of concept

It tests the one risky idea in the project, and nothing else. It may be ugly, held together with clamps, and powered from a bench supply. It only has to show that the physics is on your side.

A proof of concept for a jumping toy car is a cylinder, a valve and a weight on a rail. There is no car body, there are no wheels. The question is only: how high does this weight go with this much air?

2. Works-like prototype

It does what the product does, with the real function chain from input to output, but it does not look like the product. The electronics may sit on a development board next to the mechanism. The housing may be a simple printed box.

This is the build where you find out whether the parts work together: whether the motor is strong enough with the real gearbox, whether the sensor is disturbed by the motor, whether the battery lasts as long as the calculation said.

3. Looks-like prototype

It has the shape, size, colour and feel of the product, but little or no function. It can be a printed and painted shell, or even a good rendering for the first round.

Its questions are about people: is it too big for the shelf, can a hand hold it, do buyers understand what it is? Many teams skip this build and learn the answers after the expensive tooling is ordered.

4. Engineering prototype

Here, works-like and looks-like come together. The real electronics are on a custom board, the board fits in the real housing, and the parts are designed for the process that will be used later. Its question is: can this be built again, ten or a hundred times, and does it still pass every test?

This is the first build that deserves the full effort of a finished design. It is also the first one where a mistake is expensive, which is why the three cheap builds come before it.

Which one first?

Start with the biggest risk. Ask yourself: what is the one thing that, if it does not work, ends the project? Build only that.

  • If the risk is physics (enough force, enough heat, enough light), start with a proof of concept.
  • If the risk is the combination of many known parts, start with a works-like prototype from standard modules.
  • If the risk is whether anyone wants it, start with a looks-like model or a rendering and show it to ten people.

A feasibility calculation can often replace part of the first build. If the numbers say the motor needs three times the torque that fits in the housing, you have learned that for free. But a calculation only covers what you thought of. The prototype also shows what you did not think of, and that is its real value.

What you can leave out

The skill in prototyping is leaving things out without leaving out the question. For a first build, you can normally skip:

  • The final housing. An open frame is better, because you can reach everything and change it.
  • The custom circuit board. Development boards and ready-made modules on wires are fine until the function is proven.
  • The final material. A printed plastic part can stand in for a later metal or moulded part, as long as load and heat allow it for a short test.
  • The app, the display, the nice cable routing. A row of numbers on a laptop screen is enough.
  • Cost optimisation. One unit may cost five times the later product. That is normal.

What you must not skip:

  • The risky part at real scale. Half-size tests of heat, force or flow often give misleading answers, because these things do not scale in a straight line.
  • Measurement. A prototype without a measured result is a demo. Decide before the test which number you will write down, and what value counts as a pass.
  • Safety. Mains voltage, heat, pressure, lithium batteries and moving parts are as dangerous in a rough build as in a product. Often more, because the covers are missing. Fuses, guards and a way to switch off quickly belong to the first build.

How to know the prototype has done its job

A prototype is finished when the question is answered, not when it looks good. There are three possible results, and all three are useful.

  1. Yes. The number is inside the target. Write it down and move to the next risk.
  2. No, but close. Change one thing and test again. One thing only, or you will not know what helped.
  3. No, and far away. The concept has to change. This is the cheapest moment in the whole project to learn that.

Keep a short record for each build: the question, the setup with a photo, the measured numbers, and the decision. Three lines and a picture are enough. Later, when someone asks why the fan is this size or the wall is this thick, the answer is in that record.

Where AI helps, and where it does not

An AI design system can speed up the steps around the prototype. It can turn the idea into checked requirements, run the feasibility numbers, draw the parts for printing, lay out a circuit board and write the parts list. It can also do part of the testing in advance, with a strength check or a motion simulation before anything is ordered.

What it cannot do is the test itself. A real fire, a real slope, a real hand on a real button: these still give the answers that count. The useful split is that the system prepares the build fast, so that you spend your time on the test and not on the drawing.

On our platform, each public project shows this path step by step. The DIY BBQ Temperature Controller starts from one clear question, whether a small fan can hold a charcoal fire steady, and builds the concept, the electronics and the housing around it. If you are still at the stage before the first build, the article on writing a one-page spec shows how to turn your idea into numbers you can test.

Watch hardware get built by AI. Follow public projects on makemachine.ai, read every result in full and get an email when there is something new.

Get started See the live projects

More from the blog

Open-hardware projects built with AI