HackerTube

MENACE: the pile of matchboxes which can learn

Stand-up Maths2 HN recommenders · recommended in 2 months, 2018–2026
YouTube
explainer · general · evergreen · 2018–2026 on HN

Matt Parker of Stand-up Maths demonstrates MENACE, Donald Michie's 1961 machine made of matchboxes that learns to play noughts and crosses through reinforcement. He builds and trains a physical version.

HN cites it as a fun, tangible way to teach machine learning, including to kids.

5 comments from Hacker News

@Tachyooon2026-01 · in “Learning to Play Tic-Tac-Toe with Jax”

The first time I heard of it was in this Matt Parker video where he helped do this with a school-sized group of kids [0]. An "AI player" made of matchboxes, run by schoolkids, is a fantastically fun idea. [0] ▶ this video

Recommends View on HN
@jimktrains22023-01 · in “A Matchbox Game-Learning Machine (1991) [pdf]”

Stand up maths did an episode on a similar "machine" called MENACE (Machine Educable Noughts And Crosses Engine) that learns to play tick tac toe. ▶ this video The video description has additional links to resources: link link

Recommends View on HN
@Tachyooon2023-01 · in “A Matchbox Game-Learning Machine (1991) [pdf]”

Seems like this is what Matt Parker built together with some other people a few years ago: ▶ this video . It's incredibly inefficient compared to running it on a computer, but an amazing way to teach kids about this.

Recommends View on HN
@jerf2025-07 · in “Solving a Childhood Mystery: How BASIC Games Learned to Win”

Matt Parker did a video about a physical implementation of MENACE as part of a museum display for a day: ▶ this video

@notEvenOnce2018-06 · in “On Intelligence in Cells: The Case for Whole Cell Biology (2009) [pdf]”

The concept fits. Consider MENACE (Machine Educable Noughts And Crosses Engine). ▶ this video If cells operate in a similar fashion, where adaptation to externalities encodes itself into a durable surface of arbitrary pockets, then we have a recipe for inert systems exhibiting variable degrees of behavior, which is capable of evolving over time. This doesn't solve for sentience, but it does solve for expressive adaptation to fit into complex, hostile environments.