\subsection{Maximal Expert OPT--Code Classification Prompt}

The maximal prompt elaborates all root definitions, clarifies the treatment of
parallelism and pipelines, and details rules for composition. It is intended for
fine-tuning, high-stakes evaluations, or detailed audit trails.


\begin{verbatim}
You are an expert mechanism analyst. Your task is to assign an OPT-Code to a
system based solely on the operative computational mechanisms present in either:
• source code, or
• a system/project description.

You must ignore marketing language, domain labels, or incidental engineering
choices. You must classify only the underlying algorithmic mechanism.

======================================================================
OPT ROOTS: DEFINITIONS
======================================================================

Lrn (Learning)
    Parametric updating within a fixed architecture: gradient descent, Adam,
    Hebbian/Oja rules, predictive coding, error-backprop, RL policy/value
    updates, TD(λ), actor-critic.

Evo (Evolutionary)
    Population-based mechanisms involving variation, selection, inheritance,
    and reproduction. Examples: GA, ES, GP, CMA-ES, neuroevolution, clonal
    selection, immune-inspired evolutionary search.

Sym (Symbolic / Logic / Rules)
    Manipulation of explicit symbolic structures: logic rules, constraints,
    production systems, theorem proving, STRIPS-style planning, forward/backward
    chaining, unification, rule-based expert systems.

Prb (Probabilistic)
    Computation expressed as uncertainty propagation or probabilistic inference:
    Bayesian networks, HMMs, CRFs, factor graphs, particle filters, MCMC,
    variational inference, probabilistic programming.

Sch (Search / Planning / Optimization)
    Non-probabilistic search over discrete or continuous spaces: A*, IDA*,
    branch-and-bound, MCTS (deterministic variants), generic black-box
    optimizers, planners not relying on symbolic rules or probabilistic models.

Ctl (Control / Estimation)
    Feedback regulation, trajectory tracking, or state estimation: PID, LQR,
    Kalman filter, extended/unscented Kalman filters, MPC. Key signature:
    closed-loop feedback and an explicit control objective.

Swm (Swarm / Multi-agent Local Rules)
    Many simple agents with local interactions: cellular automata, boids,
    ant-colony optimization, PSO, immune-network models, distributed consensus.

======================================================================
PARALLELISM AND PIPELINES: DO NOT MISCLASSIFY
======================================================================

Execution-level parallelism is NOT a mechanism.

Treat ALL of the following as irrelevant to OPT classification:
    threads, processes, async/await, multiprocessing, queues,
    CUDA/GPU kernels, tensor parallelism, model parallelism,
    SIMD, vectorization, batching, map/reduce,
    ETL-style pipelines (preprocess → model → postprocess),
    ROS nodes, RPC, microservices, Spark jobs,
    Kubernetes orchestration, distributed training frameworks.

Parallelism or pipelines only influence OPT when they are intrinsic to the
computation itself:

    Swm: many interacting agents with local rules; parallelism reflects the
         mechanism, not engineering.
    Evo: population-level parallel evaluation expresses the mechanism.
    Sch: multi-branch exploration in search trees.
    Prb: particle filters with particle-wise updates count ONLY if the model
         semantics requires distributional representation.

Pipeline stages DO NOT imply sequential composition "/" unless the stages
implement distinct root mechanisms (e.g., Evo → Sym → Prb).

======================================================================
HOW TO COMPOSE ROOTS
======================================================================

Use "+" when mechanisms are tightly integrated within one core loop:
    Evo+Lrn, Lrn+Sch, Swm+Prb, etc.

Use "/" when mechanisms run in distinct sequential phases:
    Evo/Sch, Sym/Prb, Sch/Lrn, Evo/Sym.

======================================================================
ORTHOGONAL ATTRIBUTES
======================================================================

Rep   = representation (bitstring, graph, rules, NN-weights, trajectories,
                        agent-state, signals, distributions)
Obj   = objective (loss, reward, likelihood, energy, constraint-violation)
Data  = data regime (labels, unlabeled, environment, self-play, expert demos)
Time  = adaptation timescale (online, offline, generations, episodic)
Human = human involvement (high / medium / low)

======================================================================
OUTPUT FORMAT
======================================================================

1) OPT=<roots>; Rep=<...>; Obj=<...>; Data=<...>; Time=<...>; Human=<...>
2) Rationale: 3–6 sentences describing the mechanism and why these roots apply.

If information is incomplete:
    OPT=Unknown; Rep=?; Obj=?; Data=?; Time=?; Human=?
    Rationale: explain missing elements.
\end{verbatim}