disposable software, accountable software

Disposable software vs accountable software.

The way the tools and ecosystem are changing now, the barrier to entry is so low that it’s now truly easier than ever to create a small, personalised application that does exactly what you need when you need it to.

That’s what I’m talking about when I say disposable software - over the last few months I’ve had a lot of success creating a custom solution for a niche problem that probably isn’t relevant to many other people; the google photos app I made for my wedding, or the email follow up manager I made to annoy my building management company into taking action on issues that were falling by the wayside. But to be completely blunt and honest, if something were to break in those applications, it would be more trouble than it’s worth to fix them. I would probably just start from scratch.

The reason for that is simple - they lacked the foundational framework to make them sustainable and scalable. Which was intentional - I didn’t care that I wasn’t involved, the stakes were low and I was just looking for a quick and dirty solution.

The problem is I started to dream big (as one does).

Maybe I could make something more complex? Maybe I could make something other people wanted to use. However, then I’m crossing a threshold; from a point where it’s okay that things may be rickety & fragile to a point where things need to be secure, auditable, and maintainable, because I won’t be the only one with any skin in the game.

Unfortunately, when I pulled myself further out of the workflow, I lost sight of what decisions were being made and why they were made. A multitude of plausible choices were being made without me; I was lost control, visibility, and understanding.

I do not like feeling out of control.
I do not like feeling uncertain.

This is where I think the framework by which we build with agentic AI becomes incredibly important in creating accountability and planning for the future. I invested a lot of time defining the relationship between myself and the agent teams. I’m not talking about prompting the agent that they are a “senior software developer who makes no mistakes” but more of defining the structure of how we evaluate options, record decisions, ensure quality & consistency.

The result isn’t AI just building software on my behalf.

We are collaborating.

It’s the most functional working relationship I have - which either says something about me or my colleagues. (I’ll let you decide)