Very slow startup and falsified benchmarks！

[wu-clan]

Here's a video comparing startup exit with OrbStack on an Apple M5.

2026-06-19 15-14-06.mp4.zip

[linear-code]

ABXD-95

[wu-clan]

This data is from mole.

---

---

There are already multiple containers in orb and arcbox has no containers, in this case the storage footprint of arcbox is several times that of orb, your benchmark is seriously fraudulent!

[AprilNEA]

Thanks — both points are fair.

Disk footprint: You're right, this was a real bug. The data image was physically pre-allocating up to 64 GiB on first start, so a fresh install showed large usage even with zero containers. Fixed in arcboxlabs/arcbox#334 — the image is now fully sparse (only consumes disk for data actually written, like OrbStack's). Ships in the next release.

(Small note: the published benchmarks cover boot/memory/CPU/I-O/network, not disk footprint — but the idle disk usage you saw was genuinely worse than it should be, and that's a fair catch.)

Slow startup: We're actively optimizing this. The kernel boots fast; the latency is in the userspace readiness path after boot, where there are some structural issues we're reworking. The M5 is very new hardware for us, so your video is a useful reference.

Appreciate the report.

[wu-clan]

My title may be a bit extreme, but it reflects my expectations for the project.