It works.
You typed a paragraph
and something appeared
that does the thing.
Twenty minutes.
The feeling is real
and the working code is real
and both of them are lying to you
about what comes next.
Because a demo only has to survive
the path you walked it down.
Production has to survive
every path at once.
The empty list.
The user who clicks twice.
The record from 2019
that has a field nobody uses anymore
and a value nobody expected.
None of that showed up in the demo
because you did not go there,
and the machine had no reason
to take you.
What got cheap
was the first version.
What did not get cheap
was deciding what happens
when two of these run at the same time.
What did not get cheap
was the migration.
What did not get cheap
was somebody, in eighteen months,
opening this and needing to know
why it is shaped like this.
That work has not moved.
It was always the expensive part.
The danger is not that the demo is bad.
The danger is that a working demo
feels like a finished decision.
You showed it to someone.
They saw it run.
Now the question is no longer
should we build this,
it is when will it be ready,
and nobody in the room
has yet said out loud
what ready means.
So enjoy the twenty minutes.
Then treat what you have
as the cheapest possible sketch
of a thing you have not designed.
Ask what it does when it fails.
Ask who operates it.
Ask what it will cost
to change your mind about it later.
If those answers do not exist,
you do not have a system.
You have a very persuasive
screenshot.
– Serguey Asael Shinder
Top comments (0)