For a while, the number on our dashboard just sat at 999.

We kept refreshing it. One of us would reload the page, see 999, and a minute later someone else would do the same thing. We were all looking at the same number from different places, waiting for one more person. A few of us had a beer bottle ready. Everyone was waiting to shout. It was a strange, curious little moment, and we were all completely in it.

Somewhere in the middle of all that refreshing, a funny thought came up. Did we ever actually think about 1000 users? We didn't. Nobody wrote it on a whiteboard or put it in a plan. We'd been too busy building, fixing whatever broke that week, trying to make Ari teach a little better than it did the week before. And then, without really noticing, we were one person away.

When it finally turned over to 1000, it was exciting. By then we'd also passed 2000 learning sessions. But after the excitement settled, we kept coming back to a quieter question. What does 1000 users actually mean?

At first the answer felt simple and kind of amazing. Somewhere, someone bothered to come to our site and learn with Ari. Then someone else did. And it kept happening until it had happened a thousand times. We remember when our numbers were mostly family and friends, and we knew almost every name. That's long gone now. The people signing up today are people we've never met, who found us on their own and decided to give it a try. That still surprises us. We're grateful in a way that's hard to put into words.

But the longer we sat with it, the less 1000 felt like the thing to cheer about.

We're not a startup trying to grow fast so somebody buys us. We're not a social media company counting millions of users. For companies like that, the user count is the whole game. For us it isn't. Nobody comes to Ari because they want to be a user. They come because they want to pass the GED, or a certification that could get them a better job. Signing up is just the beginning, and honestly, it's the easy part.

So we've started saying it plainly to each other. Our goal is your pass.

If someone signs up and never passes, we didn't do our job, no matter how good the dashboard looks. Seen that way, 1000 users is less of a trophy and more of a reminder. A thousand people are trusting us with their time, and we owe them something that actually works.

And there are so many more people out there. Around 350,000 people take the GED in the US, and that's just one test. There are so many other certifications on top of it. Every one of those people has a test date coming and a score they need to get. That's who we think about now.

It's also why the small wins mean so much to us. Before, the math on the screen wasn't always easy to read. When we finally got it working properly with a better library, we were genuinely happy, because people could see the questions clearly. Then we were able to add graphs, and everything got so much clearer. The whole team was happy about that one. It's not a flashy feature, and nobody signs up because of it. But if you can't read the question, you can't learn how to answer it, and you can't pass. It still isn't perfect. We're working on better ways to show shapes and the questions Ari makes. But compared to where we started, it's much cleaner, and we feel proud of it.

One of our volunteers commented on our workshop and said how easy Ari was to use. That was really rewarding to hear, because it's exactly what we're trying to do, little by little: give more accurate instruction, so someone can finally learn something they weren't able to do before.

Not all of it is fun. We went back and forth with AWS customer support for a long time trying to get better model access. But it was worth the effort, because a better model means better teaching, and better teaching means more people ready on test day.

We see other learning apps getting better all the time, too. It would be easy to feel discouraged by that, and we've decided not to be. This isn't a competition for us. If someone finds a better way to help people learn, that's good news. We want to stay open-minded, learn from the new techniques, adopt what works, and bring it to the people who use Ari.

That's part of why we're writing a paper about how Ari teaches. We're starting with arXiv, and we want to keep adding case studies as we go. We genuinely believe this is working. We just want to show it properly, in a way other people can look at and check for themselves.

But there's something we haven't heard yet. Nobody has written to us to say, "I passed!"

We understand why. Ari is still new. Most of what people use today went live in late July and August, and getting ready for a test like the GED takes months, not weeks. People are still studying. So it's going to take a while. But we believe we'll hear it, and we think about that day a lot.

That's the moment we're really waiting for. Not 1000 users, and not 2000 sessions, but many, many people telling us they passed. That's when the beer bottles come out for real.

Until then, we'll keep working on Ari, keep making it better, and keep trying to reach more people, so that when you finally sit down for your test, you're ready.