TC: 1sec./move Book:Noomen 3moves
xxx games (Clean List update xx-xx-2026) Total: xxx games
=========================================================================
IPMAN CHESS – THE AI TRANSITION UPDATE
=========================================================================
The chess engine world is currently undergoing its biggest transition since the invention of bitboards and NNUE. Due to the rapid rise of AI-generated code and externally trained neural networks, some prominent testers have decided to close their top rating lists. Ipman Chess will not stop. We choose to look forward.
To address the strong feedback within the community and to keep the performance of traditional,human-written source code pure, I am temporarily introducing a separate category:
*** The AI & Experimental Transition List ***
Engines that rely heavily on AI code-generation, AI-coding assistants, or external neural network recipes (such as Coda, Gaia, Caipora, and Triumviratus) will be moved to this dedicated list.
Please note: this is a TEMPORARY separation. My 40 years of experience in engine testing has taught me that every technological leap—from Deep Blue to the first neural networks—initially faces heavy resistance, only to become the new standard later on.
As AI assistance becomes deeply integrated into all software development, these lists will inevitably merge back together.Until then, Ipman Chess will keep testing to witness and record history live: to see which engine—be it human, hybrid, or pure AI—will push the boundaries of computer chess further, and whether Stockfish can defend its legendary number one spot.
Maybe a change at the top will not happen soon, but as a neutral tester,I am excited to see what the future brings!
================================================================================
Classification Meta-Labels for the AI & Experimental Transition List:
[Agentic AI] = Code generated completely by an AI system (e.g., Claude Code).
[AI-Assisted] = Human-written code/base heavily modified or expanded by AI coding tools.
[NNUE-Ext-Recipe] = Human-written code, but neural network relies heavily on external AI training structures.
• Coda - Label: [Agentic AI]
Description:Developed entirely through human-AI collaboration; every single line of code was generated by Claude Code.
• Incipit - Label: [Agentic AI]
Description: Built from scratch using agent-driven AI coding throughout, focusing on independent knowledge generation.
• Caipora - Label: [Agentic AI]
Description: A project completely generated by Claude Code under human direction, utilizing a threat-aware NNUE structure.
• Triumviratus - Label: [AI-Assisted]
Description: Built upon a historical Stockfish source code base, but heavily modified, optimized, and expanded with AI development assistants.
• Zigqueen - Label: [AI-Assisted]
Description: Core implementation and features heavily driven by autonomous AI coding agents under human architectural review.
• GaiaChess - Label: [NNUE-Ext-Recipe]
Description: Code handcrafted by human authors, but moved to this list due to its heavy reliance on external AI training frameworks (Bullet/LC0 dataset).
IPMAN CHESS – THE AI TRANSITION UPDATE
=========================================================================
The chess engine world is currently undergoing its biggest transition since the invention of bitboards and NNUE. Due to the rapid rise of AI-generated code and externally trained neural networks, some prominent testers have decided to close their top rating lists. Ipman Chess will not stop. We choose to look forward.
To address the strong feedback within the community and to keep the performance of traditional,human-written source code pure, I am temporarily introducing a separate category:
*** The AI & Experimental Transition List ***
Engines that rely heavily on AI code-generation, AI-coding assistants, or external neural network recipes (such as Coda, Gaia, Caipora, and Triumviratus) will be moved to this dedicated list.
Please note: this is a TEMPORARY separation. My 40 years of experience in engine testing has taught me that every technological leap—from Deep Blue to the first neural networks—initially faces heavy resistance, only to become the new standard later on.
As AI assistance becomes deeply integrated into all software development, these lists will inevitably merge back together.Until then, Ipman Chess will keep testing to witness and record history live: to see which engine—be it human, hybrid, or pure AI—will push the boundaries of computer chess further, and whether Stockfish can defend its legendary number one spot.
Maybe a change at the top will not happen soon, but as a neutral tester,I am excited to see what the future brings!
================================================================================
Classification Meta-Labels for the AI & Experimental Transition List:
[Agentic AI] = Code generated completely by an AI system (e.g., Claude Code).
[AI-Assisted] = Human-written code/base heavily modified or expanded by AI coding tools.
[NNUE-Ext-Recipe] = Human-written code, but neural network relies heavily on external AI training structures.
• Coda - Label: [Agentic AI]
Description:Developed entirely through human-AI collaboration; every single line of code was generated by Claude Code.
• Incipit - Label: [Agentic AI]
Description: Built from scratch using agent-driven AI coding throughout, focusing on independent knowledge generation.
• Caipora - Label: [Agentic AI]
Description: A project completely generated by Claude Code under human direction, utilizing a threat-aware NNUE structure.
• Triumviratus - Label: [AI-Assisted]
Description: Built upon a historical Stockfish source code base, but heavily modified, optimized, and expanded with AI development assistants.
• Zigqueen - Label: [AI-Assisted]
Description: Core implementation and features heavily driven by autonomous AI coding agents under human architectural review.
• GaiaChess - Label: [NNUE-Ext-Recipe]
Description: Code handcrafted by human authors, but moved to this list due to its heavy reliance on external AI training frameworks (Bullet/LC0 dataset).
P.S. Please note that classifying these engines is an ongoing challenge. With so many new engines emerging, it is impossible for a tester to know if AI or AI-assist was used unless it is explicitly stated in the repository's GitHub or Readme. Furthermore, even established developers of traditional engines may choose to integrate AI assistance behind the scenes without public mention. Therefore, this list relies entirely on the transparency of the authors and the best available information.