72 lines
2.9 KiB
TeX
72 lines
2.9 KiB
TeX
%% SCRAP: archive/quality/validation/phase-2-executive-summary
|
|
%% SOURCE: docs/working/archive/quality/validation/phase-2-executive-summary.md
|
|
%% STATUS: HISTORICAL
|
|
%% FITS: none
|
|
%% EDITORIAL: lifted — prose rewritten to press voice
|
|
|
|
\section{Phase 2 Experiment Executive Summary (November 2025)}
|
|
|
|
Dated 20 November 2025, this document captured the design for a coupled
|
|
multivariate dynamics experiment intended to supersede the Phase 1
|
|
factorial DoE. The experiment was designed but the Phase 1 baseline used
|
|
a fixed heartbeat; Phase 2 was to capture adaptive heartbeat behavior.
|
|
The work informed subsequent L8 attractor and window-scaling campaigns.
|
|
|
|
\subsection{The Paradigm Shift}
|
|
|
|
Phase 1 measured isolated factor effects under a fixed 1\,ms heartbeat
|
|
interval. Phase 2 recognized that the system is a coupled multivariate
|
|
dynamical system: the heartbeat rate (\texttt{tick\_ns}) is itself
|
|
modulated by the inference engine in response to workload, coupling all
|
|
metrics through a single feedback variable.
|
|
|
|
\subsection{Per-Tick Observables}
|
|
|
|
Seven metrics were to be captured per heartbeat tick via a circular
|
|
buffer (100\,K slots, approximately 400\,KB memory):
|
|
|
|
\begin{itemize}
|
|
\item \texttt{tick\_interval\_ns} — elapsed time since previous tick
|
|
\item \texttt{cache\_hits\_delta} — cache lookup successes
|
|
\item \texttt{bucket\_hits\_delta} — secondary lookup successes
|
|
\item \texttt{hot\_word\_count} — words above heat threshold
|
|
\item \texttt{avg\_word\_heat} — mean execution heat
|
|
\item \texttt{window\_width} — current rolling window size
|
|
\item \texttt{word\_executions\_delta} — total word executions
|
|
\end{itemize}
|
|
|
|
\subsection{Experimental Design}
|
|
|
|
The design called for five elite configurations (identified from the
|
|
$2^7$ DoE), 150--200 runs each, across 100\,000+ ticks per run, for
|
|
750 total runs and an estimated 6--8 hour wall-clock runtime.
|
|
|
|
\begin{center}
|
|
\begin{tabular}{lp{6cm}}
|
|
\toprule
|
|
Dimension & Weight \\
|
|
\midrule
|
|
Stability & 30\% — convergence and reproducibility \\
|
|
Convergence & 25\% — speed and reliability \\
|
|
Coupling & 20\% — heartbeat response to load \\
|
|
Resilience & 15\% — noise resistance \\
|
|
Jitter & 10\% — timing cleanliness \\
|
|
\bottomrule
|
|
\end{tabular}
|
|
\end{center}
|
|
|
|
The configuration with the highest weighted composite score was to become
|
|
the baseline for MamaForth and PapaForth.
|
|
|
|
\subsection{Implementation Pipeline}
|
|
|
|
Five phases were planned: heartbeat instrumentation
|
|
(\texttt{HeartbeatTickSnapshot} struct and circular buffer injection),
|
|
per-run stability fingerprinting, per-configuration fingerprint aggregation,
|
|
golden configuration selection, and visualization with R/ggplot2.
|
|
|
|
%% TODO(bob): confirm whether the golden configuration was ever selected
|
|
%% and what label it carries in the current codebase. Cross-reference
|
|
%% against CLAUDE.md HeartbeatTickSnapshot note and
|
|
%% docs/03-architecture/heartbeat-system/instrumentation-plan.md.
|