"Minigo" refers to two different programs. An early "minigo" page (~2006) here referred to a program by Mathijs Romans. The original description of this minigo is at the bottom.
More recently (2017+), Minigo refers to an open-source implementation of the alphazero algorithm. It was originally based on MuGo, an implementation of the first AlphaGo paper by Brian Lee, and was extended to the AlphaZero methodologies and trained at scale.
It reached professional strength (although it has not played competitively) and played hundreds of millions of games.
The code and details can be found on github
Game records and strength estimates can be found on cloudygo.com.
Data is available on a public Google Cloud Storage bucket at gs://minigo-pub , and the project is still under active development.
"Minigo" is a very small program that can play 9x9 go in a terminal. It's quite weak and not very fast, but its source code is only about 170 lines and a little over 3kB.
minigo webpage (which is no longer available)
I would be happy to receive comments.