notes
things i've learned while building.
-
why does inference need a KV cache?draft
-
what actually happens during quantization?draft
-
why PDF compression is harder than it looksdraft
-
building a tiny inference engine in Javadraft