Depot

Depot

@depot.dev

Software delivery built for machine scale.

94 followers11 following831 posts

Depot reposted

Matthieu Napoli
@mnapoli.bsky.social

The dust from yesterday is settling, I want to do a shoutout to @depot.dev Building Bref container images with patched PHP versions took minutes, sometimes seconds. Thanks to them, apps on Bref could be patched in 31 minutes after the vulnerability was published ❤️

Depot
@depot.dev

light mode. for those of you who enjoy looking directly at the sun.

Depot
@depot.dev

the whole reason we reimplemented lvm: its volume-group lock is held ~100ms per op. we launch microvms up to 200 times a second in parallel, and that single lock, not the data path, is what breaks the latency target. depot.dev/blog/why-i-r...

depot.devWhy 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, ...
Depot
@depot.dev

delegation does not transfer ownership. an agent can write the change, but you still own how the system behaves in production. so iris keeps the parts that build context: planning decisions, close review, a limit on parallel work. depot.dev/blog/delegat...

depot.devHow I delegate the work, not my judgmentAgents can investigate, plan, and implement faster than I can, but reading a final diff is not the same as understanding why the system works the way it does. I own how the system behaves in productio...
Depot
@depot.dev

a container can build in seconds and still crawl on deploy, because every node pulls and decompresses gigabytes it never reads to boot. soci fetches only the bytes needed to start. the index gets written at build time. depot.dev/blog/why-soc...

depot.devWhy SOCI belongs in the buildMost 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 in...
Depot
@depot.dev

when we rebuilt depot registry's garbage collector, tlc checked 14m+ states in 21 minutes and verified in the model that a committed manifest never loses its blob. wito walks through the whole gc model and how to run one yourself. depot.dev/blog/tla-ver...

depot.devImproving 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, an...
Depot
@depot.dev

we reimplemented lvm for our microvm hypervisors, with worse guarantees on purpose. its ~100ms volume-group lock was fine weekly but brutal at 200 launches/sec. keeping device-mapper, new control plane, ~100x faster allocation. depot.dev/blog/why-i-r...

depot.devWhy 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, ...
Depot
@depot.dev

noisy test failures burn hours: people reread the same logs, retry jobs to see if a failure disappears, and slowly stop trusting failures that point at something worth fixing. one approach: give each failure a stable identity you can compare over time. depot.dev/blog/failure...

depot.devFailure fingerprints: Comparing test failures over timeTo 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...
Depot
@depot.dev

an agent can hand you a fix that works and passes its tests but could still be the wrong call for your system. you only catch that if you've kept enough context to challenge it. iris on where she stays involved to keep that context. depot.dev/blog/delegat...

depot.devHow I delegate the work, not my judgmentAgents can investigate, plan, and implement faster than I can, but reading a final diff is not the same as understanding why the system works the way it does. I own how the system behaves in productio...
Depot
@depot.dev

when the agent's code breaks at 2am, you're the one on call, not the agent. here's a practical ai-assisted workflow for staying responsible for what ships: linter guardrails, reference tests, and reviewable prs. depot.dev/blog/staying...

depot.devStaying in control of your codebase in the AI eraAI can produce code faster than you can review it, but you're still the one on call when it breaks. You need an AI-assisted workflow you trust: linter guardrails, reference tests the agent can copy, t...
Depot
@depot.dev

ci logs are a good starting point, but they're noisy and don't give you data you can filter or join across runs. structured test results let you ask which tests fail most, which are slow, and what's getting worse org-wide. now ga in depot. depot.dev/blog/now-ava...

depot.devNow available: Test results in your CI jobsReport JUnit from your Depot CI or Depot GitHub Actions jobs to see failures in context. A single command runs your suite and uploads the report. Split suites by historical timing and they'll automati...
Depot
@depot.dev

deleting from a content-addressable store is trickier than writing to one. you can't re-check s3 and delete in one step, so we record the version gc inspected and delete only that, staying safe even against a concurrent re-push. depot.dev/blog/tla-ver...

depot.devImproving 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, an...