Rexigon chess - new engine built with Kiro AI / Claude 3.5

[silentshark]

I imagine before long there will be thousands of these GenAI built engines appearing. So I had a go, using Kiro which uses Claude 3.5 as the brains. Been playing round with this since the summer.

I called the engine Rexigon, for no apparent reason. Built doing this 'vibe coding' thing where AI does most of the heavy lifting. Some bits and bobs

- currently incorporating the Pesto PST's (only) for eval.
- playing around 2400 ELO (CCRL scale)
- nothing innovative, uses all the usual techniques - I added one thing at a time, then tested, looking for obvious regressions. Here's the order of stuff I added:
- alpha-beta search with basic PST's
- quiescence
- basic move ordering
- LMR and NMP
- hash tables
- draw scoring
- better PST's and scoring interpolation between mid and endgame
- pesto's PST's
- static NMP
- killer heursitic
- SEE and drawscore fixing
- history heuristic
- more aggressive LMR
- time handling improvments
- improved SEE
- better time usage
- more LMR tweaks
- incremental PST's to speed things up
- singular extensions and multicut - then removed it
- NMP improvement
- LMP

I found that often the vibe coded stuff is not great, it missed stuff or just gets things wrong.

Anyway, I can pop onto github if anyone wants a peruse. I have a hunch that with a bit more work (eval is just Pesto PST right now so could be improved), it will likely catch my engine, Francesca, which took many years of blood, sweat and tears to develop! That was back when source code for chess programs wasn't really available - think there was an early version of GnuChess then Crafty, but nothing much!

Cheers, all,
Tom

[smatovic]

2400 CCRL Elo with PeSTO eval, nice.

What kind of board representation does the engine use?

I would add it to the YATT thread, but the engine did not develop the evaluation on its own.

Giovanni Lavorgna showed a way how to use a driver script to automate the vipe coding process:

GeminiChess, an LLM built engine
viewtopic.php?t=85490

Jost Triller showed how to generate Stockfish patches:

Re: ChatGPT usage in computer chess?
viewtopic.php?p=981272#p981272

I think this is where it starts to get interesting, a test-framework feedback loop with generative AI for chess engine development.

--
Srdja

[glav]

I imagine before long there will be thousands of these GenAI built engines appearing. So I had a go, using Kiro which uses Claude 3.5 as the brains. Been playing round with this since the summer.

I called the engine Rexigon, for no apparent reason. Built doing this 'vibe coding' thing where AI does most of the heavy lifting. Some bits and bobs

- currently incorporating the Pesto PST's (only) for eval.
- playing around 2400 ELO (CCRL scale)
- nothing innovative, uses all the usual techniques - I added one thing at a time, then tested, looking for obvious regressions. Here's the order of stuff I added:
- alpha-beta search with basic PST's
- quiescence
- basic move ordering
- LMR and NMP
- hash tables
- draw scoring
- better PST's and scoring interpolation between mid and endgame
- pesto's PST's
- static NMP
- killer heursitic
- SEE and drawscore fixing
- history heuristic
- more aggressive LMR
- time handling improvments
- improved SEE
- better time usage
- more LMR tweaks
- incremental PST's to speed things up
- singular extensions and multicut - then removed it
- NMP improvement
- LMP

I found that often the vibe coded stuff is not great, it missed stuff or just gets things wrong.

Anyway, I can pop onto github if anyone wants a peruse. I have a hunch that with a bit more work (eval is just Pesto PST right now so could be improved), it will likely catch my engine, Francesca, which took many years of blood, sweat and tears to develop! That was back when source code for chess programs wasn't really available - think there was an early version of GnuChess then Crafty, but nothing much!

Cheers, all,
Tom

Congratulations! I would be very interested too in the source code and in any info you feel to share about the project.

[silentshark]

hello,

have uploaded the source files and windows executable to github.

Program is as above, with some simple king safety and pawn structure stuff.

I guess it plays around 2500, using CCRL scale

https://github.com/silentshark-uk/rexigon

Feedback / thoughts welcome

Regards,
Tom

[Frank Quisinsky]

Hi Tom,

often unclear ...
If the sources are ready, the program is ready for compilation.

Examples:
Hobbes: https://github.com/kelseyde/hobbes-chess-engine (from the author from Calvin)
Anura: https://github.com/Vast342/Anura (from the author of Clarity)
Starlynn: https://github.com/zzzzz151/Starlynn (from the author of Starzix)

From all this "newer" engines no compiles are available.
I am thinking, not ready because programmer is working on it.

If the sources available under "tags" I am thinking ...
OK, perhaps I can ask Jim for a compile.

Example:
https://github.com/Bobingstern/Tarnished/tags

Here I am thinking, OK engine a longer time available, sources available unter "tags", so I can asked Jim.

Best
Frank

Not easy at all to managed an overview, because the programmers uses different possible ways.

[Frank Quisinsky]

Hi Tom,

why I wrote that ...

If no sources under "tags", perhaps people like Jim Ablett are not sure ...
Can I create the compiles or better not.

Loud thinking, not sure what Jim is thinking if he have time and searching for sources for the compilation.

Rexigon: https://github.com/silentshark-uk/rexigon (no entry under "tags").

Best
Frank

[smatovic]

[...]
https://github.com/silentshark-uk/rexigon
[...]

I guess what Frank confused, is that you put the .exe in the source repository.

You can make git "tags" for each release of your source in github, and than publish an "release" and add binaries to that release.

--
Srdja

[Frank Quisinsky]

Ah ...
*.exe is in source directory ...
Not my day!

I don't now why:
SG Wattenscheid 09 won 2:0 vs. ASC 09 Dortmund in football League 5 (Oberliga).
Now 12 wins and 4 draws, my club will go in League 4 ... and in 5 years the new German football champion with a bit luck.
Normally all is fully OK with this weekend.

Much more important:
Thank you Tom!!

Best
Frank

[silentshark]

Ah ...
*.exe is in source directory ...

Yeah, sorry, my lack of understanding of github. I just had a go at publishing a release and sticking the windows binary in there too!

Regards,
Tom

[Frank Quisinsky]

Hi Tom,

great ...

I am looking in the evening.
32Mb for Hash!

Maybe with a newer release of Rexigon a xb parameter for hash will be available.
That would be great ...

Good luck and many fun with your new Rexigon!

Best
Frank

A good game from Francesca vs. MChess in Paderborn (WM, many years ago)
Never forget it ... you made me and Kai very happy as operators.
Go Rexigon go ...