Xyra ✨
@xyratech.bsky.social
she/her. Tech projects and political quote tweeting. Trans fucking rights.
Where’s my open source metaverse?
Once-a-month longer updates on my projects: https://xyra.beehiiv.com/
Where’s my open source metaverse?
Once-a-month longer updates on my projects: https://xyra.beehiiv.com/
Experimenting with AI reliability in multi step tasks…
November 25, 2025 at 1:52 AM
Experimenting with AI reliability in multi step tasks…
While 2 x 9699 GBP = 19398 GBP (25011 USD) is a lot of money to spend, if you wanted 1TB of VRAM to run DeepSeek R1 with Nvidia hardware (13x NVIDIA H100 SXM 80GB), you’d be looking at a cost of around 350,000 USD 💰
March 5, 2025 at 10:47 PM
While 2 x 9699 GBP = 19398 GBP (25011 USD) is a lot of money to spend, if you wanted 1TB of VRAM to run DeepSeek R1 with Nvidia hardware (13x NVIDIA H100 SXM 80GB), you’d be looking at a cost of around 350,000 USD 💰
This picture reminds me of something…
March 4, 2025 at 4:00 PM
This picture reminds me of something…
OK, it's INCREDIBLY slow (a token output every 2 minutes), but I just got DeepSeek’s R1 671B model (dynamic quantised to 2.51-bit) running on a MacBook Pro M3 with 36 GB of RAM.
February 2, 2025 at 8:08 PM
OK, it's INCREDIBLY slow (a token output every 2 minutes), but I just got DeepSeek’s R1 671B model (dynamic quantised to 2.51-bit) running on a MacBook Pro M3 with 36 GB of RAM.
I got an LLM to make me a React app.
Why is my node_modules half a gigabyte?!
Why is my node_modules half a gigabyte?!
February 1, 2025 at 3:25 PM
I got an LLM to make me a React app.
Why is my node_modules half a gigabyte?!
Why is my node_modules half a gigabyte?!
Absolutely wild.
What happens when models are consistently over 84%?
What happens when models are consistently over 84%?
February 1, 2025 at 3:16 PM
Absolutely wild.
What happens when models are consistently over 84%?
What happens when models are consistently over 84%?