Y
Hacker News
new
|
ask
|
show
|
jobs
by
whimsicalism
188 days ago
I imagine you have to start decoding many speculative completions in parallel to have true low latency.