I put a thought into Capture and got one giant Action back. It was supposed to be broken into smaller things I could actually do. Other thoughts were landing in the wrong place, and correcting them wasn't reliably helping.

My agent had broken the latest build. I was annoyed, and my notes from those days are not especially polite. But while I was trying to get the app working properly again, I noticed something else: it had become useful enough in my day-to-day development work that the failure was genuinely disruptive.

I built Capture for myself. It's a place to put thoughts without first deciding how to organize them. I can type or dictate something messy, and the app is meant to separate things to do from ideas I'm still developing, keeping related thoughts together in Threads.

A lot of my use happens while I'm walking or running. On my phone, I use LocalWhisper to turn speech into text. I talk about things I want to build, changes to my apps, something that isn't working, or an idea I haven't quite worked out yet. Those thoughts don't arrive in the order a requirements document would prefer.

Sometimes one capture mentions both Capture and Retake. Sometimes I start with a product problem and end up thinking about how to explain it in an article. I want to finish the thought before deciding which parts belong to development and which parts belong to writing.

That material becomes something I can work with later. I copy the history and bring it into my agent workflow, where we can work through what to build or change. For writing, I find it easier to speak the idea out, let AI help edit it, and read it myself. Capture gives that unfinished thinking somewhere to accumulate before I have to make it coherent for someone else.

In a September note, I described two frictions it was removing: deciding where things should go, and finding the material I wanted to share with an agent. That sounds fairly small when I write it down. But those are things I run into repeatedly while building.

When sorting became unreliable, I had to pay attention to them again. A thought about writing ended up in the Capture Thread when it also belonged in Writing. A paragraph became an Action. Instead of carrying on with the idea, I was trying to get the app to understand where it should have gone.

On September 24, I recorded the observation that started this article: breaking the app had shown me how useful it had become. I had built something to make my own work easier, and the frustration was a fairly direct indication that it had been doing some of that work.

That doesn't tell me how useful it will be to someone else. It certainly doesn't make the bugs a good thing. It does give me a clearer reason to keep working on it, beyond another feature I could add or a better way to explain the product.

Reading back through Bugs & Open Issues also stops me from making this sound like a perfect app that suddenly had one bad day. I'd complained about similar problems before. At one point I wrote that capturing had become easier, but I was still having to move things around manually. The friction hadn't disappeared just because I had built an app to remove it.

To find a version we could trust, I went back through Capture's own history. I looked for the point where everything was still being recorded properly. We used that to find the working version and rolled back to it.

So the app's history became useful in a way I hadn't planned for: it helped us work out where to restart. The record of me using Capture was also a record of when Capture was doing its job.

The latest break made me change how I approach development. I now keep a core baseline that works, then build from it in smaller sprints that go through repeated tests. Aside does most of the testing, and then I do my own pass.

I want those checks to resemble the board I actually use. Can a long capture reach more than one existing Thread? Can it contain tasks as well as an idea? Those are the kinds of tests I've been asking for.

I had already written down the rule for myself that evening: keep a working baseline and be able to get back to it quickly if something breaks. Having reports that tests passed hadn't been enough to make the ordinary experience dependable. I still need to use the thing myself.

There are plenty of directions I could take Capture. My own Threads are full of them. For now, I want to protect the part that made me notice its absence: being able to say what I'm thinking, have it land somewhere sensible, and come back to it when I'm ready to do something with it.

Try it now - https://trycapture.app/