This is a very bad chess bot I wrote in a few days, after getting inspiration from Neal Agarwal's password game. I didn't look anything up, so it's a simple materiel advantage lookahead with no hash table and no pruning. Jonathan Gordon contributed the piece weights, my initial guesses were Queen 25, Knight 10, Bishop 7, Rook 5, Pawn 1.
The config.yml can be pasted into a lichess-bot repo to hook it up to lichess, I have a bot account here but I'm not going to leave it running all the time as it's on my laptop.
It beat sargon-2ply once, so that's something. sargon-2ply was running on a raspberry pi and only used 19 seconds of its time, while chess used ~52 minutes at full throttle, but hey it's still a victory.
make
./chess - the engine will play a game against itself, searching for moves with depth 5
./chess --stockfish /path/to/engine - the engine will play a game against another UCI-compatible chess engine, such as Stockfish. It will attempt to hamper the other engine by asking it nicely to set its skill level to zero, and requesting moves with go depth 1.
./chess --uci - the engine will behave as a UCI-compatible engine which can be connected to some chess frontends or lichess-bot. It doesn't understand every option, but it will try to determine how deep to search based on its time remaining, if the information is provided. If time remaining is not provided, it will search using the depth option, if it is provided. Setting depth to anything above 6 will likely never complete.
Other engines have "Nodes/second" listed as some sort of performance metric, I'm not sure exactly what that measures, but I can count the number of outer loop iterations in engine, which comes out to ~6M moves/minute using all 4 cores on my laptop. This is probably way higher than whatever the actual "nodes/second" measures as other chess engines with lower values seem to perform much better.