01Dynamo
Routing by cache and capacity
Learn why the best worker is not always the emptiest—or the one with the biggest cache hit.
0/4
Interactive field guide
Learn the systems behind
cache-aware inference.
Tiny explainers. Concrete mental models. One active-recall check after every concept.
02Mooncake
Keeping the right KV blocks alive
Understand why adding RAM and disk does not automatically produce useful cache hits.
0/10
03Mooncake Codebase
Read the serving system, not just the flags
Trace Mooncake from source-of-truth config through object lifecycle, transport, runtime patches, validation, and business proof.
0/8
One idea to carry through all tracks
“A cache is useful only when the router can find it, the storage layer keeps the right blocks, and retrieval costs less than recompute.”