October 25, 2026
Why Tuist should build the harness on top of compute
Compute is commoditizing fast. The opportunity is in the experience on top of it, and I think that's where Tuist should be.
I've been using Amp a lot for coding, and it has changed my mind about something. I was never keen on building things remotely. I don't know why, but doing work on the go always felt weird to me, probably because of my past mental models. I still believed that some features and some work cannot happen remotely, and that at some point you need to sit down and run a system. The more I use it, though, the more I realize how cool this actually is. Mobile platforms can enable experiences that weren't possible when we were so coupled to having an editor to write code. If agents write the code, and review becomes less important because another agent does it (or maybe it won't even be necessary in the future), then you can start thinking about other inputs, like voice, or deep integrations into the system, like Live Activities on iOS.
That experience built a strong appetite in me to bring this product surface into Tuist, and a big part of the reason is what I see happening in the compute market. A lot of the companies that invested in compute are trying to sell it either as CI runners or as boxes you can build your own product on top of, and that market commoditizes extremely quickly. Linux is especially commoditized (macOS is a different story). The switching cost is very low and the contract is simple: give me a sandbox, destroy the sandbox, and sign some legal agreement with the provider. Some of these companies have noticed this too, and they're trying to bring new products so they don't depend on being the backend of someone else's product, like forever providing the environments for Claude sessions that another company runs for GitHub, because those companies hold the control. You want something unique and innovative that expands your market.
The problem is that they have no clue how to do it. In all honesty, the most they've managed is "oh, you can open a terminal in the browser", with some fancy proprietary name for it and a bunch of blog posts saying this is the future. I think that's a missed opportunity. Investing in a great harness, a great experience on top of the compute, is what can make the difference, and this is where I think Tuist should be positioned.
We started from the angle of "we want to make things fast", and I think the next step is the harness. We can begin with the obvious things, based on the data we already have, like proactively fixing flaky tests, or having a conversation about the architecture, making some changes, and validating them. Then we can go beyond that and build product features on the go. It's a coincidence that we started with mobile, and we're becoming more generic, but we're still strong there. We like mobile, we come from mobile, and we have an app. We've also sold distribution on mobile, something Amp hasn't figured out yet. They could, of course, these people are super smart, but if we invest in this I think we could easily become the go-to tool for developers.
The model is simple. You bring your own tokens, whether through a ChatGPT subscription, a Claude subscription, or an inference provider, and we bring the compute. On top of that we add a layer for how people will work in the future and how they will build features. I would try to make it native, because I've noticed that a lot of these experiences are very web-based and you feel it as you navigate around the app. There's a weby feeling that we want to get rid of.
We're finalizing the first use of our compute, which is AI runners, and that was the obvious one. Everyone knows what CI/CD is and everyone needs a place to run their CI workloads. But we're also going to explore new workloads that could run on top of that compute, which would be great for the business too. And if you haven't tried Amp yet, I'd recommend it. It's very cool, I'm building pretty much anything with it these days, and it has built an appetite in me too.