Recurse
W2/6 D5/5

Created 2024-04-05 / Edited 2024-04-05

Virtual errands, GPU token generation. Oh right, Alloy

Token tree 3x3 cat sat on the

  • General Status:
    • Virtual today
    • Mostly doing some errands and general reading
    • Hung out with J some, looking at LLM stuff and showing a GPU cloud executor
  • Alloy+LLM+Rails Project:
    • I'm giving local-intel-gpu one more last chance
    • ... I got it working on an A100 and it was a lot faster! But like... 10x, not instant
    • I generated a large tree but it had some unescaped character that I haven't found yet so I don't have the actual rendered graphviz
    • I remembered to turn off the rented instance :)
    • I started going back and updating Alloy knowledge, esp around time-domain which was added since I learned it
    • Next I need to start doing real Alloy models of things I think about or systems I use, and start improving the tooling around that
    • I'm starting to think about Alloy tooling, both direct and LLM; like in addition to Rails in/out I'm thinking of PlantUML in/out. A lot of Alloy users report using it as a thinking tool. I use PlantUML as a thinking tool. So ...