How we decide what to build next
August 9, 2026 · 1 min read
We get more good ideas than we can build. Here is the rough process we use to choose between them without it turning into a popularity contest.
Every request we receive goes into one list. No categories, no scores, no voting — just a list, in the order it arrived.
Once a fortnight we read the whole thing top to bottom. That sounds inefficient and it is, but reading it end to end is the only way we notice that six differently-worded requests are actually the same underlying problem. Those clusters are what we build.
Then we ask two questions. Does this help someone finish a job they are already trying to do? And can we ship it in under two weeks? If the answer to both is yes, it usually goes next. Big ideas are not off the table, but they have to survive a few rounds of that reading before we commit.
The honest part: some good ideas never get built. We try to say so plainly rather than leaving them sitting in a queue forever.