The Price of a Model Pull: Benchmarking a Cache at the Syscall Floor
What repeated model downloads cost a GPU fleet, and how far a pull-through cache can go when the warm path is engineered down to one kernel boundary crossing.
Blog
Chasing the Absolute Floor: Zero-Copy, Single-Syscall HTTP on Darwin in Go
An engineering journey from the default Go net/http serve loop to a perfectly fused, single-syscall cache hit on macOS.