Phil 9.5.2025

Human munitions:

  • At dawn on May 8, 2023, a 17-year-old Russian teenager named Pavel Solovyov climbed through a hole in the fence of an aircraft plant in Novosibirsk, Russia. He and two friends were looking for a warplane that could be set on fire. An anonymous Telegram account had promised them one million rubles, around $12,500, to do so — a surreal amount of money for the boys.

Tasks

  • Bills – done
  • Clean – done
  • Weed?
  • Dishes – done
  • LLC call – done
  • Dentist
  • Load up truck

SBIRs

  • 2:00 meeting today
  • Send Matt the code review paragraph. Done
  • Thinking more about the maps as a W2V approach. I think I’m going to make an X by Y (by more?) grid that has vector “labels” that can also be arbitrary size. Then pick a random starting point and do a random walk for a number of steps. That set of vectors becomes the input for the skip-gram calculation. Once the model is trained, re-run the random walk data to get the new vectors and see if the embeddings match the relationship of the original grid. The nice thing is that we can start very simply, with the index for each cell as the input, and a 2-neuron final layer that should approximate the XY. Then we start playing with the size of the “index” and the size of the final layer as independent variables