@xaroth.bsky.social
That's it - I'm moving to the brain and becoming the head of HHS!
May 1, 2025 at 4:20 PM
That's it - I'm moving to the brain and becoming the head of HHS!
The best predictor of future behavior is past behavior.
May 1, 2025 at 4:17 PM
The best predictor of future behavior is past behavior.
I don't know why I read it as "de-luxifying", but I was confused as to what it meant to "lux" game pieces in the first place. 😅
April 10, 2025 at 8:38 PM
I don't know why I read it as "de-luxifying", but I was confused as to what it meant to "lux" game pieces in the first place. 😅
Performance on write is a big one, but so is performance on read! Smaller data chunks means better cacheability and fewer network bytes.
Other benefits include:
* Unambiguous data modeling
* Semantic field meanings
* Simpler auditing
* Easier to parallelize for exceptionally large systems
Other benefits include:
* Unambiguous data modeling
* Semantic field meanings
* Simpler auditing
* Easier to parallelize for exceptionally large systems
February 11, 2025 at 11:08 PM
Performance on write is a big one, but so is performance on read! Smaller data chunks means better cacheability and fewer network bytes.
Other benefits include:
* Unambiguous data modeling
* Semantic field meanings
* Simpler auditing
* Easier to parallelize for exceptionally large systems
Other benefits include:
* Unambiguous data modeling
* Semantic field meanings
* Simpler auditing
* Easier to parallelize for exceptionally large systems
I look forward to this getting into standard libraries, and for GPU manufacturers to start leveraging last year's breakthrough in matrix multiplication. arstechnica.com/information-...
Matrix multiplication advancement could lead to faster, more efficient AI models
At the heart of AI, matrix math has just seen its biggest boost “in more than a decade.”…
arstechnica.com
February 11, 2025 at 11:00 PM
I look forward to this getting into standard libraries, and for GPU manufacturers to start leveraging last year's breakthrough in matrix multiplication. arstechnica.com/information-...