Rapid iteration in the modern coding world

I recently wrote about how LLMs have made simple software trivial to create, but additionally, it's made features incredibly simple to test with very little effort.

As an example, just today, I was using my own Mastodon app, Best-o-Masto, and was thinking about what I could do next. One idea I had was to create an alternative feed layout which presented each post in full screen and let you swipe right and left to move between posts. All the post actions would be at the bottom of the screen so you could easily interact and browse your feed one-handed. It sounded like an interesting idea in my head, but I wasn't sure how it would feel in practice.

But of course, we live in a world with coding agents. Over my morning coffee, I simply requested the feature in Claude Code, giving it quite a few details to make sure it understood what I was trying to do, and had it go build out the feature. It took a few additional prompts to lock in an experience that I thought was solid, but I was able to do this while I was preparing for the day and making my wife's lunch. Because I try to be a good developer, I also monitored what sorts of changes it was making to ensure that it wasn't introducing any new bugs or complexities, but that was about it.

After playing around with the feature for a few minutes on my own device and feeling pretty happy with it, I shared a build to the public beta where I can get feedback. It's really the velocity that excites me about this. I was able to go from an idea for something that might work to having it in the hands of users in what is effectively zero time in the world of traditional development.

If the feature works, then great; I can refine it further to make sure it is as good as it can be and then release it to everybody. If it turns out people don't particularly like it, then I can just revert the change and not ship it to production. There's no real sunk cost I'm worried about in the way I would feel if I had spent several days of focused time implementing this.

Not all features are this simple, and obviously, larger ones will take more time to implement. But what I am saying is that there are a bunch of small changes to software that simply don't happen because our velocity is low enough that we need to prioritize things in a way where it doesn't make sense for a lot of features to even get made. I really like that in the current world, these sorts of changes can be made with little risk in time and effort.

If you'd like to try this out, you can join the public TestFlight for Best-o-Masto and try it today.