M
MNIST64
Handwritten digit recognition in 6502 assembly on an unmodified Commodore 64
Emerged finds new products worth a look, daily — what’s new today
If you have ever wanted to see what a transformer is actually doing, this one is a level you can walk around in while it reads your handwriting.
Doomformer compiles small integer models into DOOM levels: weights are lines on the floor, sums are light levels, activations are staircases. It handles MLPs, spiking networks and transformers with integer softmax and RMSNorm, and ships a museum map, a DrawPad you draw a digit on, and a level where the marine's policy network picks its moves. MNIST comes in at 97.16% against 97.58% for the float model, with every exposed value matched to a NumPy reference. Install with pip; GZDoom and Freedoom to play. No LayerNorm, and inference is slow.
More like this
Doom or Bloom
A short quiz that plots your AI outlook next to 50+ public figures
jdBasic
A modern BASIC where you pause a running program, change the code and carry on
Whiteboard
A desktop canvas for reviewing what your coding agent actually built
Tonefold
Open-source music sketchpad: AI writes parts into a piano roll you can edit
Nerdulator
Put in a number, date, colour or address and get every computable fact about it