Source Authorship Forensics

Amira — human-written, or AI-generated?

A source read of FauziAkram/amira, contrasted with Stockfish.

Author Fauzi Akram (ChessTubeTree)Language C++ (single file)Scope ~2,500 LoC · 264 commitsEval Hand-crafted (HCE)Analyzed 2026-07-11Stockfish similarity Independent →
Verdict · AI-Assisted

A self-described human + LLM collaboration, built on camera.

Amira is open about its making: the README describes it as “the result of a collaborative effort between human and LLM,” developed as an exercise in building a competitive engine — and grown live on the author’s ChessTubeTree YouTube “Coding a Chess Engine” series.

It is a deliberately single-file, dependency-free C++ engine with a classical hand-crafted evaluation (piece-square tables, mobility, king safety, a pawn-structure cache) and a modern search (PVS, aspiration windows, quiescence). Human-directed, LLM-assisted, and its own contained thing.

Confidence: High — the human + LLM collaboration is stated in the README.

Part of a series applying the same lens — from the AI-pasted Luna and agent-built Owen, through a growing set of AI-assisted engines, to many human-written ones.  See the index →

1Evidence

#SignalWhat it showsWeight
1README: “a collaborative effort between human and LLMSelf-disclosed AI-assisted authorshipNuance
2Grown live on a YouTube “Coding a Chess Engine” seriesA human teaching/directing the build in publicStrong
3264 commits with a PR-based release workflow (v1.84, v1.85…)Sustained, versioned human developmentStrong
4Single-file, dependency-free; classical HCE; no AI-residue markers, no SF codeIts own contained, independent designSupport

2Compare & contrast with Stockfish

Where it resembles Stockfish
Bitboards; Zobrist hashing; transposition table
Modern search: iterative deepening, PVS, aspiration, quiescence
Versioned, incrementally-improved development
Where it differs (largely independent)
Single-file, dependency-free by design
Hand-crafted evaluation (HCE), not NNUE
Own contained architecture, built as a teaching project
No Stockfish code or techniques copied

Bottom line

Amira is an AI-assisted engine — self-described as a human + LLM collaboration — and a distinctive one: a single-file, dependency-free, hand-crafted-evaluation engine grown live on a YouTube series. Human-directed throughout, it is one of the more independent engines on the Stockfish axis: it shares the standard search stack but keeps its own classical, self-contained character.