Rethinking dev dependencies

Dev dependencies shouldn't be shipped to prod. Or should they?

Currently we're separating in our head the part of the software that ships to prod from the one that runs in our dev environments.

I think we would produce much better software if we merged the two.

The software is also how you ship it. They're the same thing.

In a world of AI agents this is even more true. An agent learning means that it "deploys changes to itself".

I keep being obsessed about all-in-one designs. I think that's the future of software. But maybe I shouldn't call it software.

Maybe it's more than that. Sometimes I am not even sure how to call what I build. Because yes I am building a customer platform, but I also build the software engineer who builds the customer platform.

How do you call this thing? Because the software engineer IS NOT meant to be user-facing. It's literally just for me.

So it's not a dev tool. In fact I don't want dev to use it because it will mean I will have to maintain it for them.

I want this tool to only be used to build bluewind. And the previous statement doesn't make sense, you see why?

Because this software engineer IS part of Bluewind. Look into odoo.sh. It's the same idea: they're starting to build a product to ship the product.

The product and how you develop it are the same thing.

In other terms, the Software development lifecycle needs to be fully integrated into your product.

Heck, the statement above can even be misinterpreted. So I am going to rephrase it:

Your SDLC IS your AI agent. They're literally the same thing.

When you build a true AI agent with you are effectively architecting your own SDLC.

And why am I talking about AI agents? You know, I don't care about the name, call it however you want, as long as it mixes gen AI and traditional software to interact with the outside world.

I am talking about AI agents because it's the future of software.

Which is why the future of software is about unifying the SDLC.

Side note: the SDLC also needs to be unified with your business. Heck, your business IS your AI agent. That sounds crazy right?

Well, how do you plan on building useful features for business if your SDLC and your business are not integrated?

You might say: "Yeah but it's just an integration, it's not that the SDLC and the business should be the same"

I respect this statement, but I want to ask: why?

I think AI agents are also the future of businesses. Because AI agents WILL BE businesses.

I am building and engineering Bluewind. Literally. The Software I produce IS THE SDLC and the business.