@sequoia is leading our Series C at a $1B valuation.
Since January, we’ve gone from running 5M CI jobs a week to over 90M. Agents are writing an order of magnitude more code, and all of it still needs to be built and tested before it ships.
To restore a cache, Blacksmith runners pull from many storage servers at once, and under load those bursts kept stalling.
We retuned the network path, and the median restore for caches over 256 MiB fell from 5.5 s to 2.8 s.
The slowest 1% went from 35 s to 17 s.
30% less job time on some workloads, just from running in the right region.
→ We record what every job pulls: code, images, data
→ We measure how fast each region reaches it
→ We place each org nearest its traffic
Turborepo caching is now on by default for free on Blacksmith.
400+ teams are already using it. One monorepo that runs over a thousand Turborepo tasks a day now sees cached runs finish 97% faster.
Large downloads from GitHub are now 22% faster on Blacksmith runners (p90: 54 → 65 MB/s).
We shipped a network optimization and measured it on production traffic, before and after. More coming soon.