I am not nearly as impressed or interested in apps that are built in a single prompt as I am in systems and projects that are able to evolve over time with stability.

While the speed and ease of outputting code has increased dramatically (which I for one am super grateful for), the importance of solid architecture has remained the same. It’s fine if a pet project falls over, can’t handle scale, has no observability outside of it working or not, etc. But for a business, or product built for others - in my opinion, that isn’t acceptable.

While building Sanctum, I was not intently monitoring every line of code that the agent put into the product. Instead I was very purposefully ordering and layering features into the application that forced an intended architecture to emerge.

Creating core feature primitives before their full expression built a solid foundation which enabled me to layer features on top of it as the need for them became more clear.

The system didn’t support tags until it supported notes that could be tagged. Notes need text, the text comes from an audio file, the audio file comes from a recording. And all of this, all of this came from the ethos that I want us to own our files, so therefore everything should be plain text, portable and accessible from the filesystem.

The feature isn’t the first thing, in my opinion, the ethos is.
This informs how I’ll go about even beginning to think about the infrastructure. When data privacy and ownership are part of the core ethos, the overarching architecture must fit within that.

The architecture of Obsidian is drastically different than that of Notion. Obsidian was offline first from the beginning, meanwhile it took Notion years to support that.

There’s a reason why a finance building doesn’t look like a Church or a Temple. They both provide shelter but the perspective, purpose, the intentionality…the ethos is different.

So the next feature you are building, the product idea, the seed of an app that exists for you, I invite you to get clear about the core principles that are driving it, and then to begin to build through them.
Instead of relying on really large feature or product specs, or massive prompts that contain everything, ask what is the minimal viable feature, what groundwork should be laid to support where you are headed or where you may head.

That’s all for now, thank you for your time and attention, may you have a beautiful day

PS - if you are keen to try Sanctum, it’s LIVE on the iOS app store :) https://apps.apple.com/us/app/sanctum-the-voice-journal/id6796428879