AI makes for better software

I am convinced now that AI-first software engineering makes for orders of magnitude better code. One reason is that there is no justification anymore to take “architecture” shortcuts. Until now, as engineering leaders we always had to make the trade-off between speed-to-production and certain architectural decisions. One example: when hosting on firebase it’s possible to simply use the client SDK to retrieve data, instead of (properly) putting data retrieval behind an API route. This was vastly faster to production, because you didn’t need to provide the API route, and usually didn’t come with any security tradeoffs. It came with the tradeoff though in that we’d put “data logic” on the client, instead on the backend - making changes harder, requiring frontend deployments.

This doesn’t make sense anymore, when embracing AI-first software development. There is no reason that shortcut should be taken, since time-to-production is virtually the same for both approaches. The more I am working with my AI engineering minions to build large projects, the more I realize that this applies for many other tradeoffs we used to make. So yes, AI-first software development makes for better code, since you can enforce proper architecture from the start, without sacrificing speed.