niles
Joined 16 years ago
Joined 16 years ago
It's a little hard to reason about the use case for "in memory" but file based, and what the limitations are. What specific product problem did you build this to solve?
Beyond that it would be helpful to have some hardware variation in the performance claims. Does zen5 avx512 make a difference? A node with optane (memory or disk)? Threads vs clock speed? ARM vs X86? Feels like this was built with a specific cloud native deployment in mind and a certain scale and speed of storage.
Anyone compare with the club-3090 dual setup?
Does this interfere with cache hits? Could a single conversation or task span multiple roles?
Why are you building this? Does this maximize my toxen value by saving the hard tasks for the hard model? Does it maximize cache hits as part of its scoring? Does it help agents develop a specialist mindset? Are you anticipating users will have many local models hot, or is this also a model load/unload controller?
https://www.cdc.gov/folic-acid/data-research/mthfr/index.htm...
"There's no better PC under $350 than this Ryzen 7 powerhouse with a CPU that's faster than a 10-core Apple M5 CPU, and that's no mistake
Deals
By Wayne Williams published November 4, 2025
Machenike AI Mini PC comes with 16GB of high-speed RAM and a 512GB M.2 SSD
"
https://play.google.com/store/apps/details?id=projekt.launch...
Has a bug with 120hz VRR or OneUI and also kinda abandoned.
24GB and 48GB cards are supposed to be coming.
https://play.google.com/store/apps/details?id=com.simon.harm...
For me, I expected each date to reflect the most important stories, but to exclude the prior ranges.
For example, somehow three people getting exchanged is the most important news of every time period?