@culurciello.bsky.social
Intelligence architect, father and husband, also: AI, computing, deep learning, machine learning, hardware, SW, HW
This means the communication between GPUs and servers and racks is barely utilized. And the fun fact is that all hyper-scalars over-optimize these communication links spending millions, when one can do away with much cheaper and simplest systems!
October 29, 2025 at 5:05 PM
This means the communication between GPUs and servers and racks is barely utilized. And the fun fact is that all hyper-scalars over-optimize these communication links spending millions, when one can do away with much cheaper and simplest systems!
"memory is always the next bottleneck"
August 14, 2025 at 2:58 PM
"memory is always the next bottleneck"
But at some point human consumption will flatten because there will be no wages to pay for things. So... ???
July 28, 2025 at 12:30 PM
But at some point human consumption will flatten because there will be no wages to pay for things. So... ???