blog@​arhan.sh:~$

From NAND to AI

!! UNPUBLISHED !! • more posts


  • cache pairs of acceleration vectors
  • bfs for precomputing fitnesses, diagonal distance, take into account obstacles, the noop case
  • screen memory for extra temporary swap space
  • coefficients to avoid overflow
  • lcg for random and scale down
  • heap memory over read bug
  • efficient collision detection
  • fitness stepWeight
  • dynamic mutation rate by regressing binomial distributions for different brain steps
  • choose population count based on remaining memory
  • unsigned division for more range for fitness
  • raster bit packing
  • string ascii packing
  • throw out part of the OS
  • A=-A ROM optimization
  • preceding syntax
  • vm escape bug