A network, learning
A 105-parameter neural network learns to separate two kinds of points.
Vladimir Dolgopolov
I’m an AI engineer in Toronto and Head of Engineering at STRATIS, where we connect media and creative data to decisions. Outside work I build Crowns, a daily logic puzzle, and interactive experiments.
Puzzle
Today’s Crowns
One crown in every row, column and region, with no two touching.
Lab
Experiments
-
-
Three neighbours, one rule
Each cell looks at itself and its two neighbours, and one rule decides the next row.
-
Geometry without a 3D engine
A trefoil knot, shaded face by face on a flat 2D canvas. Drag to turn it.
Notes
Recent writing
Oct 2026 · Puzzles
One solution, guaranteed
How the Crowns generator turns a random board into a logic puzzle with exactly one answer.
Sep 2026 · Generative systems
Simple rules, complicated consequences
How three cells and eight decisions become a pattern that is hard to predict by eye.
Sep 2026 · Machine learning
What can 105 numbers learn?
Inside the lab’s small network: weights, backpropagation, and what training fit does and doesn’t mean.
