Speed is a feature, and we publish the number we have not beaten

The rule we wrote down before building anything: under two seconds from pressing send to seeing your app change.

We have not met it. This is the record, because a company that quietly moves its own goalposts will quietly move other things.

Why two seconds

The loop between "I want this" and "I see this" is the entire value of a tool like this. At two seconds it feels like moving something with your hands and you keep going. At ten you switch to another tab and the session is over.

That is not a preference about polish. It changes what people build, because a fast loop gets used forty times in an evening and a slow one gets used four times and abandoned.

The measurements, in order

• First honest measurement: eleven to fourteen seconds. Genuinely bad.
• After removing round trips — five model calls down to one — under four seconds.
• Then, re-measured on a mature app dozens of changes deep: five to thirteen seconds. The earlier number had quietly stopped applying as apps grew.
• Cause: the builder was reading your app before editing it, because the one file holding the whole app had grown past a size limit and was the one file left out of what it was shown.
• After showing it the whole thing: about three to five seconds, and cheaper per change than before.

That last detail is the one worth dwelling on. Removing a step made it faster AND cheaper, which is what deleting work looks like rather than trading one cost for another.

Where the remaining time goes

Almost all of it is waiting for the model to answer. There is nothing structural left to remove: a change is a single call now. Getting to two seconds would mean paying roughly four times as much per change for a faster tier, and it is not certain that would even reach two seconds.

So the decision is deferred rather than forgotten, and in the meantime the honest figure goes on the website: three to five seconds.

Why tell you this

Because you will find out anyway, the first time you use it on a mature app. A company that publishes "under two seconds" and delivers four has taught you to discount everything else it says. A company that publishes four and delivers four has not.

And there is a second reason: the number is a commitment device. It is much harder to let performance slide when the current figure is written on your own front page.