Engineering


Iris Scholten
26.09.16
How we load tested Depot Metal before scaling up
We built a production-shaped load test to find control-plane limits before moving more GitHub Actions traffic to Depot Metal. Each run made our scaling changes measurable.


Luke Morris
26.09.10
GitHub Actions without GitHub
Depot CI runs GitHub Actions workflows without GitHub’s control plane, keeping CI online during outages and making workflows portable to new forges like Origin.


Peter Hasko-Nagy
26.08.26
Introducing the Fibonacci Benchmark™ for CI runners
Taking the Sandbox TTI benchmarks one step further for fun.


Héja Péter (Vau)
26.08.13
Why I reimplemented LVM (with worse guarantees)
Our microVMs launch in under a second, and assembling their block devices in parallel put LVM's roughly 100ms
volume group-wide lock directly in the critical path. We kept the kernel's device-mapper, wrote our own control
plane and gave up crash safety our ephemeral workloads never needed, making allocation roughly 100x faster.


Chris Goller
26.08.04
Why SOCI belongs in the build
Most of a large container's startup goes to pulling and decompressing files it never reads to boot. SOCI makes the image seekable so the runtime loads only the bytes it needs, and Depot writes that index during the build, so the image can lazy-load the instant it's pushed.


Wito Delnat
26.07.20
Improving system safety with Temporal Logic of Actions (TLA+)
Formal verification never seemed worth it: writing a faithful TLA+ spec took too long. But now agents can write the spec from your code. We tried it on the rebuilt Depot Registry garbage collector, and the model checker found something our tests and reviews didn't: a race that could delete live data.


Andrea Anderson

Pedro Guerra
26.07.15
How we use PostHog to keep our AI assistant on track
Keeping our AI assistant working well is a human responsibility at Depot. PostHog's AI engineering features turned our never-ending reviews into 30 minutes a week.


Iris Scholten
26.07.02
Failure fingerprints: Comparing test failures over time
To make recurring test failures easier to reason about we needed to include historical context. Here's how we derive a stable identity for each test and a fingerprint for each failure, so you can tell a brand-new problem from the same flaky test you've seen a dozen times.


Luke Morris
26.06.16
Observing Lambda durable functions
Durable Lambda functions break most of the tracing assumptions you carry over from regular code. They replay from the top every time they wake, so spans get emitted twice, parent spans lose their children, and queue wait time hides in the gaps. Here's how we restored end-to-end visibility into Depot CI's durable functions.


Peter Hasko-Nagy
26.05.06
How we got microVMs booting in under a second
Depot CI's VM scheduler is just-in-time: no pre-warming, no warm pool of standby VMs. We stacked optimizations to get microVM cold boots from 7 to 9 seconds down to under a second, and this post walks through each one.
Older posts
Page 4 / 4
