levelsio cuts filter responses to 87 milliseconds after Carmack post

levelsio says he pushed the filter responses on his site down to 87 milliseconds, and he credits a single post for it. John Carmack had written that everyone should applaud performance improvements but should gauge them against the speed of light rather than a sloppy competitor, adding that if you are measuring in seconds you may have already anchored yourself incorrectly. levelsio quoted that post and said it was why he went back in.

John Carmack

@ID_AA_Carmack

X

Everyone should applaud performance improvements, but gauge against the “speed of light”, not just a sloppy competitor.

Carmack's post is not a vague call to be faster. It is a list of structural moves. Put locally cached media up a few vblanks after the process launches. Structure the server so one query returns everything the dynamic UI needs, kicked off in parallel with graphics API initialisation. His bar for a launcher is that it should be responsive before your muscle memory can slam the pointer to the place you are about to click.

The other box he shipped

In the same stretch, levelsio added a line under checkout reading "Don't want to join? Tell me why", which opens a modal for the answer. That followed a Paul Graham post saying that founders whose customers will not buy should try asking what they would buy, and that surprisingly few do it. We covered the checkout box earlier.

The business read is that both moves come from the same instinct, which is to stop benchmarking against the thing in front of you. The competitor's load time is not the target, the physical floor is. The customers who already paid are not the ones with the useful answer, the ones who just closed the tab are.

Worth being clear about what is and is not established here. The 87 millisecond figure is levelsio's own number for his own site, posted by him, with no independent measurement attached. Nobody has published what the number was before, so the size of the improvement is unknown. The checkout modal has been live for days and no response rate or sample of answers has been shared yet, so whether a text box aimed at non-buyers actually produces anything usable is still an open question.

What a founder can take from it today is cheap. One text box on the exit path costs an afternoon. One honest look at whether your latency target came from physics or from the slowest tab your competitor serves costs less than that.

Paul Graham

@paulg

X

If customers won't buy your stuff, try asking what they would buy.