bfollington Idea to Prototype
An idea is not yet a thing to build. Between them is a sequence of alternations: open it up, then close it down, then open the next layer, then close that.
Collapsing too early kills what was interesting. Never collapsing produces a beautiful idea nobody can start. The skill is knowing which motion the moment calls for — and being willing to have the idea die along the way, which is a success, and cheap compared to finding out later.
Hold it in superposition first
At the start the idea is several things at once, and the version worth building is often not the one first said.
Entertain what it could be before deciding what it is. Gather what it is adjacent to, what it reminds people of, what feeling it is reaching for. Nothing is being committed to yet, and premature specificity here is expensive — it forecloses the versions nobody has thought of.
Ideas are fragile at this stage. Analysis dissolves them like dreams on waking. Hold judgement.
Find the compelling angle, or let it go
Then close: why this? There are always more ideas than time, so the question is not whether it could work but why it is worth the room.
What is the sharpest version — the one someone would be excited about in a sentence, without preamble? If several sentences of setup are needed before it lands, the angle has not been found yet.
Sometimes there is no compelling angle, and it dies here. That is a good outcome reached cheaply, and it should be said plainly rather than allowing a lukewarm idea to proceed on momentum.
Design: how it actually works
Open again, further in. Not architecture — the experience of the thing. What happens, in what order, what it feels like to use, what makes it good rather than merely functional.
The useful question at this depth is what quality the thing must have to be worth existing. Then the design work is finding the arrangement that produces that quality, rather than the arrangement that merely delivers the features.
Spec: the fastest possible verdict
Close hard. A prototype is not a small version of the product. It is the fastest thing that would settle whether this is worth doing.
Two symmetrical goals: the maximum possible taste of what a good outcome would feel like, and the fastest possible invalidation of the whole concept. Both point at the same design — go directly at the uncertain, interesting part and fake absolutely everything else.
Whatever is already known to work should be stubbed, hardcoded, or borrowed. The prototype's only job is to produce a verdict, and every hour spent on a part that was never in doubt delays it.
Name the verdict condition before building: what would be seen that means yes, and what means no. A prototype with no failing condition will succeed regardless of the truth.
Check it is possible, and whether it is done
Before building, look for prior art. Someone may have answered this, or answered most of it, or failed at it instructively. This is a short search with a high payoff — and finding it already exists is itself a verdict.
Then the plan: phases with gates, where each gate is something that can actually come out badly. That is convergent planning, and it is the next skill.
Failure modes
- Collapsing on the first version. Building the idea as first stated, which is rarely the good one.
- Never collapsing. Endless exploration, nothing buildable, no verdict.
- Skipping the pitch. Proceeding on momentum because it is interesting to build, without asking why this one.
- The prototype as small product. Real auth, real persistence, real error handling, on something that may not be worth existing.
- Building the certain part. Effort spent where there was no doubt, so the verdict arrives late.
- No failure condition. Nothing stated that would mean stop, so it never stops.
- Analysing too early. Scrutiny applied while the idea is still a spark.