ბლოგი

SwifterPM: faster Swift package resolution for generated, Bazel, and Buck2 projects

Our users kept telling us that resolving Swift packages was slow, and that the directories where they get resolved ate gigabytes of disk. That pain has only grown as people run several agents across worktrees at once. So we built SwifterPM. It leaves resolution to the Swift Package Manager and speeds up everything around it: fetching the pinned sources, putting them on disk, and reading their manifests. Starting today it's the default for Tuist generated projects, with an environment variable to opt out, and it ships as a rule for Bazel and Buck2.

The build system you already have

Underneath the surface syntax, every build system is the same graph. After years building remote caching for them, we got curious about what actually makes one tick, and what a build system would look like if you designed it today for coding harnesses.

Inside Atlas: the operations platform we built for ourselves

We are a very small team, and from the beginning we have been obsessed with operating efficiently. Off-the-shelf tools made us bend our work to fit them and siloed our data. So we built Atlas, our own operations platform, leaning heavily on LLMs and agents. Here is a tour of what we have actually built into it: sales, finances, product operations, and the interfaces that tie them together.

Scaling iOS application development with Tuist

Delivery Hero used Tuist to scale its iOS modularization to roughly 115 modules, cut clean builds from 2.5 minutes to 0.5 minutes, and reduce CI build and unit test time by 2.5x.

Monzo slashed CI pipeline time by 70%

Monzo used Tuist project generation, Binary Cache, and selective testing to cut median PR time from 52 minutes to 15 minutes while scaling its iOS codebase to nearly 200 modules.

L10N.md: localization built like we ship software

Existing localization tools take content out of the repository and translate it in environments where changes can't be validated. We tried Crowdin and Weblate, none of them felt right. So we built our own system on top of LLMs, with the repository as the source of truth and a lockfile to keep things incremental.

Hyperconnect optimized its multi-service pipeline with Tuist

Hyperconnect adopted Tuist to optimize a multi-service build pipeline, improving feedback loops in local development and CI with Module Cache and Selective Testing.

Run your test suite across balanced shards

When parallelizing tests on a single machine isn't enough, sharding across multiple CI runners is the next step. Tuist uses historical timing data to create dynamically balanced shards so every runner finishes roughly at the same time.

Turning AI agents into doctors for your project's health

What if AI agents could diagnose your project like a doctor, using Tuist's data as their medical records?

Xcode and Gradle: Two worlds, one infrastructure

We plugged our infrastructure into Gradle in three weeks. Here is what we learned about the differences between Xcode and Gradle, and what it means for anyone building tools that sit between developers and their build systems.