AI-pilled and building

We won't be writing code for much longer, whether we like it or not. And a lot of the complaints about agents not being good at it are rooted in fear. They might not be as good as you are today. They will be. This becomes more obvious as codebases grow to the point where nobody in the organization can hold the whole thing in their head. That's not new. That has been everyone's position at companies like Google for years. It just got democratized.

It sounds daunting, and the moment you embrace this model, you expose yourself to critics. Some people have started calling us AI-pilled, with a tone that suggests that's why we sometimes ship bugs. But we ship bugs sometimes because we ship more than we used to. Way, way more. Same odds of introducing a bug, a much larger volume of code, more bugs. That's just math. I believe agents will get better at this, and in the meantime we are building a system to catch these issues earlier instead of slowing down. Even if, at times, the amount we can put out there gives me vertigo.

Many things won't matter as much as we think they will. Some people just need time to accept that. Others might never accept it and will hold on to the idea that their craft is writing code. That's fine. But it won't be most people's craft.

For me, the beauty of all this is in who it unlocks. I have friends building their own software. My in-law is building an interactive web experience to teach Spanish to Moroccan women who emigrated to Spain and don't speak the language. That's what really matters to me.

In the context of Tuist, what matters is that we can compete with the large players. They are drowning in layers of bureaucracy they don't know how to peel. Something I've learned is that few people know how to make the right calls, including letting people go when it's needed. So they let a backward mindset rot the organization and close the door on a whole world of potential. That's not what we want for Tuist.

Some people might feel uncomfortable with us being this way. Some might even churn because of it. I get it. We can't please everyone. But we'd rather embrace new realities, or soon-to-be realities, and navigate them, than let the business stagnate and let our customers down by leaving their requests unimplemented. That's not the future we want. And that's what business is all about.

I spent a week building support for Elixir. Everyone is talking about CI being a bottleneck and how to eliminate it. I could do all that exploration in a week, and now I have something the rest of the team can review and ship next week. All of a sudden, we're giving an entire ecosystem access to technology and infrastructure that helps them make the most of the vCPUs in the environments where their builds run, instead of wasting compute. Isn't that cool? And it's possible because we have access to this new way of working, and we take advantage of it.

I think it's amazing. I'm a builder. I'm AI-pilled. I'm extremely fired up. We'll use it, and we'll learn from it, to bring Tuist to every ecosystem that suffers from slow builds, from Elixir to Rust. We are building a build system because the experience deserves to be elevated, not built around an enterprise's way of working while accepting that developer experience can't get any better.

Yes, it can. In fact, it should. And we are here for that.