For years I was obsessed with building custom tools and software for pretty much any use-case. Ecom analytics, social media stats and tracking, stock forecasting, etc.
It was never really viable and to be honest, also not really smart to do it, as there were somewhat great solutions for all these applications, if one puts in the time setting it up and getting to know many different tools and UIs, which seemed more of a problem in a larger company as I had to explain each new tool to more and more people.
With new models and agentic coding, this is changing.
And changing quickly.
Over the last 6-8 weeks I built different internal software for my company to help improve efficiency in our processes and even build functionality not really available in SaaS software in the market.
The thing with the new models like Opus 4.8 or Fable now is that it takes less time writing the prompt and having Claude Code one-shot the output, with maybe 1-2 hours of really refining it down to personal preference and use case, than setting up any existing tool and workflow. Especially when the use case is not super clear or well defined and would need some customization from a out-of-the-box SaaS service.
So we are not talking about costs here mostly, at least not for most software and most companies.
It feels to me like we are getting to an intersection point, where for the first time, it really is cheaper again to just vibe-code software, rather than buying it.
This tool stack of internal software also feels the same in usage, look and outputs. So switching from a finance tool to a shipment tracking tool is much more seamless and coherent.
At scale, things will look very different for sure, but I can see a time where everyone can at least hyper-customize their front ends and functionality, adding their own preferences while all are sharing the same internal data.
Software for one, not many.