We want your work to travel.

We judge our work twice. Once when you use it, and once when you see what it made. The first should need no explaining. The second should need no fixing.

The difficulty is ours to carry

Anything hard can be handed to the person using it. Put every option on screen, call it powerful, and the work of understanding it quietly becomes theirs. We think that is backwards. Every choice we can make correctly on your behalf is a choice you never have to learn, and the ones left over should be the ones you actually have an opinion about.

The test we hold ourselves to is that nobody should need to be taught. No tour, no onboarding video, no settings page you have to understand before you start. If you cannot tell what to do on the first screen, that is our bug and not your confusion.

Finished, not nearly finished

It is easy to build something that gets most of the way and hands you the rest. The last stretch is where the real work is, and it is also where it is tempting to stop, because the difference between almost right and right is invisible in a demo and obvious the moment you try to use the thing.

So we measure ourselves by what you have to redo. If you open what we gave you and start nudging, retiming, recropping, rewriting, then we did not finish the job, we just moved it. Ready to use means ready to use, and the amount of work left over after we are done is the number we care about.

Small, and close to the work

There are not many of us, which we treat as an advantage while it lasts. The people who build it answer the email, so when something comes out wrong you are talking straight to whoever can fix it. Telling us it came out wrong is the most useful thing you can do for us.

Toronto

We are here because it is a good place to do this. A deep bench in machine learning, a lot of people making things on the internet, and enough winter to get some work done. Say hello: [email protected].