Hi, a candidate for a Brazilian Superfinal: Caipora 3.8, an original UCI NNUE engine from Brazil, written from scratch in Rust, public and GPL-3.0: https://github.com/ComputecOrg/caipora/ ... s/tag/v3.8 (Windows and Linux builds, network embedded).
Brazil also has Tucano, so the country may now qualify with two engines. Our own estimate puts Caipora at about 3600 on the CCRL Blitz scale (gauntlet against Stash 33 and 37), and an independent test by Sylwy here on TalkChess (4'+2", still in progress) has it at 31/42 in a field that includes Shredder 14 and HIARCS 15.4. Both are only early numbers, of course.
Transparency note: the code is written by Claude Code (AI) under my direction, as declared in AI_USAGE.md in the repository.
If it qualifies on your list, a Caipora vs Tucano Superfinal would be a lot of fun to watch. Thanks for running the series!
Chess Engine Nations Cup — National Superfinals Series
Moderator: Ras
-
Zigoto
- Posts: 7
- Joined: Thu Oct 01, 2026 5:43 pm
- Full name: Matheus de Carvalho Jesus
-
goni-K26
- Posts: 268
- Joined: Mon Mar 09, 2026 4:16 pm
- Full name: Agron Kovaci
Re: Chess Engine Nations Cup — National Superfinals Series
Hi Zigoto,
Thank you for your interest and for suggesting Caipora!
It looks like a very promising project with strong early numbers.
However, Caipora cannot be included in the Chess Engine Nations Cup for two reasons based on my tournament rules:
1. Eligibility Rules: According to my entry criteria, a country needs at least two qualifying engines on my personal rating list (built from a 1.85M+ game database).
For Brazil, the gap between #2 (Minke) and #3 (Tucano) was larger than 70 Elo.
According to rule [3b], this triggered a direct Superfinal between Cinder and Minke, meaning the Brazilian cycle for this series is already officially finalized and closed.
2. AI Usage Policy: As explicitly stated in your repository's AI_USAGE.md, the source code for Caipora is written by an AI coding agent (Claude Code).
My testing framework is strictly dedicated to evaluating and ranking human-written chess software architectures.
For this reason, I do not include or test engines that rely on AI-generated source code.
Also, I apologize for previously testing Coda Chess Engine in my database, as I wasn't aware at the time that it was an AI-assisted project.
Thanks again for the transparency regarding your development process, and I appreciate you following the series!
Best regards,
Agron Kovaci (goni-K26).
Thank you for your interest and for suggesting Caipora!
It looks like a very promising project with strong early numbers.
However, Caipora cannot be included in the Chess Engine Nations Cup for two reasons based on my tournament rules:
1. Eligibility Rules: According to my entry criteria, a country needs at least two qualifying engines on my personal rating list (built from a 1.85M+ game database).
For Brazil, the gap between #2 (Minke) and #3 (Tucano) was larger than 70 Elo.
According to rule [3b], this triggered a direct Superfinal between Cinder and Minke, meaning the Brazilian cycle for this series is already officially finalized and closed.
2. AI Usage Policy: As explicitly stated in your repository's AI_USAGE.md, the source code for Caipora is written by an AI coding agent (Claude Code).
My testing framework is strictly dedicated to evaluating and ranking human-written chess software architectures.
For this reason, I do not include or test engines that rely on AI-generated source code.
Also, I apologize for previously testing Coda Chess Engine in my database, as I wasn't aware at the time that it was an AI-assisted project.
Thanks again for the transparency regarding your development process, and I appreciate you following the series!
Best regards,
Agron Kovaci (goni-K26).