How we crush everyone
How land stacks up against the CI platforms, fast runners, test selection tools and merge queues teams use today. We list where others are ahead too, because a comparison that only says nice things about us isn't worth reading.
Last checked 14 November 2026 against each company's own docs and pricing pages. Products change quickly, so if something here is out of date, write to rohan@ataraxy-labs.com and we'll fix it.
The short version
- Most tools speed up the same work. Fast runners make your full suite finish sooner. land runs less: it records which files each test command reads and reruns only the commands a change can affect.
- Most selection is a map you maintain or a model's guess. land records what each test command actually read when it ran, so there is nothing to write and nothing to drift.
- Most CI tests each pull request on its own. land tests the result of every branch landing together, and tells you which branch broke it.
- No one else merges by entity. Every product here uses git's line merge, so two agents editing different entities in one file still conflict there. land merges them.
- land never lands a change on a retry. Other tools retry or quarantine a flaky test so the merge goes through, and a retry that passes can hide a real bug. land runs a failing test command once more on the same code. If it passes the second time, the verdict names it flaky and the change is still refused. We checked this in production with a deliberately flaky test.
- land signs what passed. Each landing comes with a signed statement of the exact tree that passed and the tests that vouched for it, which anyone can check offline.
Side by side
Blue means the product does it well, grey means partly, red means it doesn't offer it.
Scroll sideways on a phone.
| Picks tests by | Change it looks at | Tests the merged result | Merging | Proof main passed | Build cache | Flaky tests | |
|---|---|---|---|---|---|---|---|
| Ours | |||||||
| land | recorded readsskips whole test commands | file | yesevery branch landing together | by entity | signedverifiable offline | dependenciesnot your own build yet | namedrerun on the same code, never landed on a retry |
| CI platforms | |||||||
| GitHub Actions | nonepath filters | n/a | merge queueprivate repos need Enterprise Cloud | git lines | status check | actions/cache | manual re-run |
| GitLab CI | RSpec filesruns them first, then everything | file | merge trainsPremium and up | git lines | pipeline | cache keyword | historyjob retry |
| CircleCI | coverageSmarter Testing | file | no queue | git lines | status check | yesplus Docker layers | detect, rerun |
| Buildkite | none yet | n/a | via GitHub queue | git lines | status check | cache volumes | quarantineEnterprise |
| Harness CI | call graphTest Intelligence | method | no queue | git lines | status check | yesplus Bazel, Gradle | quarantine |
| CloudBees Smart Tests | predictionsimilarity and history | file | no queue | git lines | status check | none native | detects |
| Fast runners | |||||||
| Blacksmith | Vitest betafiles a test imports or reads | file | no queue | git lines | status check | yessame-datacenter, sticky disks | dashboard |
| Depot | none | n/a | no queue | git lines | status check | yesplus shared layer cache | flags flips |
| Namespace | Bazel targets | file | no queue | git lines | status check | cache volumes | none |
| RWX | cached tasksskips unchanged tasks | task inputs | no queue | git lines | status check | content hashed | quarantine, retry |
| Ubicloud, WarpBuild, RunsOn | none | n/a | no queue | git lines | status check | yes | none |
| Test selection and build skipping | |||||||
| Vercel CI with Turborepo | task graphdeclared inputs | file | no queue | git lines | status check | remote cache | none |
| Nx Cloud | task graphplus recorded file reads | file | no queue | git lines | unknown | remote cache | retries |
| Gradle Develocity | predictiontrained on history | file | no queue | git lines | build scan | build cache | detects |
| Datadog Test Impact Analysis | coverageoff on the default branch | file | no queue | git lines | n/a | none | part of suite |
| Preloop | tracingmechanism not published | not published | no queue | git lines | signed records | yes | triage |
| Merge queues | |||||||
| Trunk, Mergify, Graphite, Aviator | jobsby history or globs, or none | file | yesbatching and bisection | git lines | status check | none | Trunk, Mergify detect |
Where others are ahead
- Reusing your own compiled code. Blacksmith, Depot, Namespace, RWX, Harness, Nx and Turborepo reuse it between runs. land reuses your dependencies and rebuilds your own code on each landing.
- Start time after a repository has been idle. Depot, Blacksmith and Buildkite start in seconds. land starts in about a second with a runner up, and takes over a minute after 30 idle minutes. The fix is built and not switched on yet.
- A dashboard of flaky tests. RWX, Trunk, Mergify, Nx, CircleCI and Buildkite have one. land lists flaky tests in each verdict only.
- Skipping inside a test command. CircleCI and Harness can skip individual tests. land skips whole commands, so split a large suite into smaller ones.
- Testing before a commit. Depot, RWX and Blacksmith can check uncommitted changes. land works on committed branches.
land try, which runs the same check in Docker on your own machine before you push, is built and not released yet. - macOS and Windows. Most of these run them. land runs Linux only, because it learns which files each test command reads from the Linux kernel, and macOS and Windows don't allow that kind of tracing in the same way.
- SOC 2 and regions. Blacksmith, WarpBuild, Ubicloud and others are SOC 2 audited. land is not yet, and runs in one region.
When to pick something else
- Your suite already runs in a minute or two and you rarely see merge conflicts. A fast runner is the simpler buy.
- You need macOS or Windows builds, or GPU runners.
- Your security team needs SOC 2 before anything touches your code.
- Your tests are mostly in a language land doesn't select for yet, and your suite is slow mainly because of build time.
land is built for teams whose agents push many changes into the same code at once, where the full suite is slow and line merges keep conflicting.
Sources
Each row comes from the company's own documentation, pricing or changelog.
- GitHub merge queue
- GitHub Actions runner pricing
- GitLab merge trains
- GitLab fail fast testing
- CircleCI Smarter Testing
- Buildkite pricing and Test Engine
- Harness Test Intelligence
- CloudBees Smart Tests
- Blacksmith Vitest selection
- Blacksmith security
- Depot CI for agents
- Depot test results
- Namespace caching
- RWX caching
- RWX Captain
- Ubicloud
- WarpBuild
- RunsOn
- Vercel CI
- Nx affected
- Nx flaky tasks
- Develocity predictive test selection
- Datadog test impact analysis
- Preloop
- Trunk predictive testing
- Mergify batches
- Graphite merge queue
- Aviator batching