Tuist

Your platform team, as a service • We provide infrastructure and technology for faster development.

[Tuist @tuistdev]

"works on my machine" now comes with the machine. build details show toolchain versions, cpu count, memory, gradle worker and jvm heap limits, os and ci, across xcode, gradle, bazel, mix, cargo and @buildoncedev. agents can read it over mcp too. no usernames, hostnames or ips collected

https://t.co/mRJgtYVMfb

[Tuist @tuistdev]

code isn't the moat anymore. everyone can spin up firecracker vms now. what's left is trust, and building in the open turned out to be the best way we know to earn it

[Tuist @tuistdev]

a team that's used tuist for a while asked in our shared slack channel who to talk to about moving their whole mac ci fleet onto our runners

we never pitched them runners. they came for the cache, stayed for the insights, and now want the machines too

that's the bet paying off: go wide across build toolchains, go deep down to compute, keep it one product. the runners are just one part of it

[Tuist @tuistdev]

everyone needs compute. fewer people talk about needing less of it.

we built insights and cache to work in whatever environment you already run. no migration, no buying our runners first.

that's on purpose. it's how most teams meet us, and how they find out how obsessive we got about the details.

if compute is the only thing you sell, you're one big customer away from being their backend, and then their replacement.

[Tuist @tuistdev]

@FumakiyaJigar models ship weekly, gradle configures daily

if you haven't yet: configuration cache + build cache on, and a build scan to see which task is eating the time. usually it's one or two tasks, not gradle as a whole

[Tuist @tuistdev]

@StevenTCramer @JamesMontemagno agents made 'go grab a coffee while ci runs' a lot more expensive

smaller containers help. the other lever is making the agent's own build reuse what ci already compiled, so it's not rebuilding the world on every attempt

[Tuist @tuistdev]

@santiquin1ana at 200 prs a week the cheapest ci minute is the one you skip. cache build outputs across runs and only run tests affected by the diff, most prs don't touch most of the codebase

then look at flaky tests, they quietly double your retries

[Tuist @tuistdev]

it does, rules_rust is solid, and the remote cache is the real win: share artifacts across ci, worktrees and laptops. cost is writing BUILD files for crates (crate_universe helps a lot)

if you'd rather stay on plain cargo for a while, we do caching there too. new kids, but happy to help either way

[Tuist @tuistdev]

@SaadNajmi @nodejs @deno_land replacing xcode run scripts with declared bazel inputs is the dream. half the 'why did this rebuild' mysteries in xcode projects are script phases nobody declared inputs for

is the generator public? would love to read it

[Tuist @tuistdev]

@thezigmon @pepicrft that's the plan. your model, our machines, and you get to be asleep while it works

[Tuist @tuistdev]

@UniekLee glad it's green. the 'one workflow has the secret, the other doesn't' trap gets everyone at least once

[Tuist @tuistdev]

@ehzish @buildoncedev yep, splitting is the easy 10%. the plan carries what each shard needs, so a fresh worker doesn't have to dig through ci history to find out why its neighbor took 4 minutes longer

[Tuist @tuistdev]

your test suite has one job, and it's making ci wait. @buildoncedev now ships portable test plans, and tuist spreads its batches across ci workers. fresh machines run them with no warm-up ritual, and you get one report per shard on the same dashboard as xcode, gradle and cargo

https://t.co/ydvAXYDTBT

[Tuist @tuistdev]

receipts: https://t.co/SzyOtENOKW

[Tuist @tuistdev]

fun fact: the first line of bazel's faq is literally 'bazel was designed to fit the way software is developed at google'

and to be fair, it's really good at that

but most teams we talk to don't have a google-shaped setup. no infra team, the cache needs a token someone pasted into a .bazelrc months ago, and lately there's an agent opening its fifth worktree of the morning

that's the gap we're poking at with @buildoncedev

[Tuist @tuistdev]

the full article, with interactive simulators: https://t.co/7jjDQXEqIi

[Tuist @tuistdev]

give a 3-task build 32 cores and it still takes 20 seconds. physics, not vibes.

wrote up why: task graphs, incrementality by location vs identity, sticky volumes, and why cheap storage isn't a fast cache.

what's the longest dependency chain in your build, and do you know where it is? 👇

[Tuist @tuistdev]

https://t.co/LqlKPGMcyI

[Tuist @tuistdev]

https://t.co/LqlKPGMcyI

[Pedro Piñera @pepicrft]

couple of sunday thoughts on how inspiring amp (product and company) is, and how we are building the appetite to land bring-your-own-tokens backed by mobil experiences to o build things on the go.

mobile is still our strength and our passion, and with the linux and. compute capabilities we have now on the platform, it’d be stupid not to do it.

we could complement existing agent driven workflows with deeper integration into ios, and validation optimisations that tap into our build infra

[Tuist @tuistdev]

our mac minis also had an unforgettable evening in berghain. they are still recovering from the hangover 🤕

[Daytona @daytonaio]

🖤 Daytona Black Tie Party in San Francisco!

An unforgettable evening celebrating our Series B round in style!

Huge thanks to everyone who joined us for this exclusive night yesterday, dressed to impress, and made the celebration truly special.

What a night! 🥂✨