---
type: "post"
title: "Is the future of software a Craftsman house?"
description: "What does commercial software look like when anyone can vibe code a replacement?"
author: "Trey Causey"
date: 2026-10-20
slug: "is-the-future-of-software-a-craftsman-house"
url: "https://www.treycausey.com/blog/posts/is-the-future-of-software-a-craftsman-house/"
---

I've been thinking a lot about the deflationary aspects of AI, especially when it comes to software development. In the limit[^1], building new software to sell doesn't make sense right now. Why?

a) It's trivial to have an agent build a custom version for ~free and

b) Even if agents struggle with your particular use case now, they won't in ~six months (and even if they struggle in six months, enough progress will have been made that it will still feel close enough to discourage building it).

I personally use more and more bespoke software each day and cancel subscriptions as I do.

As someone who makes his living in software, that can feel pretty discouraging! And, to anyone thinking about starting a startup, it can feel downright prohibitive. What's the point of building an app if anyone can make their own version of it for free? This is true for small / consumer apps for the most part (think anything that is basically CRUD under the hood). Design, aesthetics, etc. all matter but I think we also have a lot of data that they matter a lot less to the modal consumer than some of us would prefer.

For enterprises, of course, the needs are often much more around custom integrations, regulatory compliance, legacy systems, etc. These are often seen as moats for large software providers. That too, though, is ripe ground for agents, and they're particularly adept at working through the kinds of toilsome tasks needed at enterprises.

But I don't think that means every enterprise is going to be out there vibe coding their own CRM or ERP any time soon (not to mention the liability assurances that are associated with using vendors). So what does big, complex software look like?

One idea I'm starting to kick around that I think may be a more common use case in the near-term is something like the Craftsman houses that we love here in the PNW or kit cars. You purchase something that contains all of the materials, instructions, and then you assemble it and customize it to your specific needs.

I can imagine soon selling software that is ~80% complete because there's no need to reinvent the wheel every day. The remaining 20% could be a detailed interview that you provide to software agents, either bring-your-own-key or as part of an upsell from the vendor, followed by agents building the remaining bits and integrating where needed. Those parts are bespoke, specific to your use cases, and use the domain expertise you have.

Think of it as the big software update you inevitably have to do when you buy a new piece of electronics and power it on for the first time.

[^1]: Note that this is in the limit! I don't expect the world to stop making software right now, this is more of a thought exercise.