Prototype the hard part first

Play this lesson

What could sink Shard Garden?

Four friends on four continents in one garden: will planting feel instant when one player is 300 ms away? If that fails, nothing else about Shard Garden matters.

Construx's cone of uncertainty poster says a well-run project "attacks areas of highest variability first": the parts the team is least sure about.

A prototype answers one question

A prototype is quick, throwaway code built to answer one question fast.

Remy's first one: two players, gray boxes for plants, one server in Europe and a tester in Australia. No art, no menus, no saves. The question: "Does planting feel instant at 300 ms?"

Quick check

Which is a good question for a prototype?

  1. Can two players 300 ms apart plant together without it feeling slow?
  2. Is the game good?
  3. Will the final art look nice?
Show the answer

Can two players 300 ms apart plant together without it feeling slow?

A prototype answers one specific question the team could get wrong. "Is the game good?" is too big to answer with gray boxes.

Easy parts feel like progress

Starting with the menu, the logo and the settings screen feels productive, and none of them can fail. The team could spend months on them, then learn the core idea does not work.

Hard part first moves the bad news to week two, when changing course is still cheap.

Prove it, then rebuild it

Malte Ubl notes that a prototype showing that something works can be a strong argument for a design. But prototype code is quick and messy on purpose.

Once it has answered its question, the studio rebuilds the idea properly, with tests, using what it learned. The prototype itself is not shipped.

Quick check

A prototype that works should become the shipped game's code, to save time.

Show the answer

False

A prototype skips tests, structure and error handling to answer a question fast. Shipping it means shipping all those shortcuts.

Pick the order

Kai's to-do list has three items. Which should Remy prototype first?

  1. Four players in one garden over a slow network
  2. The settings screen
  3. The store page trailer
Show the answer

Four players in one garden over a slow network

Playing together across continents is the part most likely to fail, and the whole game depends on it. Settings and trailers cannot sink the project.

It works. Now show it.

The prototype says yes: with the right netcode, planting feels instant even at 300 ms. Now the publisher wants to see what the real game will look like.

Bolt says: Next: vertical slices, a small piece of the game at full quality.

Fact-checked 12 October 2026