5acaa5553d
Two fixes for clean local builds: 1. journal/.latexmkrc — sets pdf_mode=1, max_repeat=5, silence_logfile_warnings=1, force_mode=1. Without this, latexmk bails on first-pass undefined-reference errors before getting to the second pass that would resolve them. Standard issue with pdfLaTeX in nonstopmode. 2. Replaced the literal U+1F98E lizard glyph in 2026-04-20-evening-mega-session.tex with a textual description. pdfLaTeX with default inputenc can't render U+1F98E; XeLaTeX or LuaLaTeX could but switching toolchains is a bigger change than the issue warrants. The lizard stays in markdown / Pluto where it works fine. Final state: 26 pages, 744 KB, no Reference/LaTeX/Unicode errors in the log. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
152 lines
6.7 KiB
TeX
152 lines
6.7 KiB
TeX
% ---------------------------------------------------------------------------
|
|
% 2026-04-20 evening — Lab journal scaffold, full Julia migration, app v1
|
|
% Live / B-style narrative entry. A-pass markers throughout.
|
|
% ---------------------------------------------------------------------------
|
|
|
|
\session{2026-04-20 (evening)}{$\sim$3 hr mega-session, autonomous}{Stand
|
|
up the lab journal as a permanent invention log, port the entire MATLAB
|
|
toolchain to Julia and delete the originals, build a Pluto.jl predicate
|
|
explorer as the FRET-adjacent UI v1. All in one go in auto mode.}
|
|
|
|
\section{2026-04-20 (evening) — Journal, Julia migration, app v1}
|
|
\label{sec:20260420-evening}
|
|
|
|
\subsection*{Origin of the session}
|
|
|
|
Dane (post-dinner) green-lit three tracks discussed in the afternoon:
|
|
\begin{enumerate}
|
|
\item LaTeX lab journal as a multi-modal invention log; daily
|
|
entries; A-style for retroactive, B-style for live with
|
|
\texttt{\textbackslash apass\{\}} markers.
|
|
\item Full Julia migration of the MATLAB code, since the nonlinear
|
|
reach experiments made it clear we'd live in Julia anyway.
|
|
Delete MATLAB.
|
|
\item ``App v1'' as a Pluto.jl notebook — a stand-alone read-only
|
|
renderer of \texttt{predicates.json} with the boolean $\to$
|
|
halfspace bridge made visible. Edit UX present but
|
|
non-functional.
|
|
\end{enumerate}
|
|
|
|
Direction at start: ``be ambitious, be creative, leave Easter eggs,
|
|
have fun.'' Auto mode on.
|
|
|
|
\subsection*{What landed}
|
|
|
|
\begin{enumerate}
|
|
\item \textbf{Journal scaffold}: \texttt{journal/} with
|
|
\texttt{preamble.tex} (callout boxes, code listings,
|
|
session-header macro, \texttt{\textbackslash apass\{\}} marker),
|
|
\texttt{journal.tex} (top-level aggregator),
|
|
\texttt{journal/README.md} (format conventions).
|
|
\item \textbf{Two retroactive A-style entries} reconstructed from
|
|
commit history + transcript:
|
|
\texttt{2026-04-17-controllers-linear-reach.tex} and
|
|
\texttt{2026-04-20-predicates-boundaries-julia-nonlinear.tex}.
|
|
Both at full invention-log depth.
|
|
\item \textbf{This entry} — B-style narrative with pointers.
|
|
\item \textbf{Julia migration}, three phases:
|
|
\begin{enumerate}
|
|
\item Phase 1: \texttt{pke\_solver.jl},
|
|
\texttt{plot\_pke\_results.jl},
|
|
\texttt{main\_mode\_sweep.jl}. Mode-sweep results
|
|
match MATLAB to 3--4 decimals on every state.
|
|
\item Phase 2: \texttt{reach\_linear.jl},
|
|
\texttt{reach\_operation.jl},
|
|
\texttt{barrier\_lyapunov.jl},
|
|
\texttt{barrier\_compare\_OL\_CL.jl},
|
|
\texttt{load\_predicates.jl}. Per-halfspace margins
|
|
match MATLAB exactly.
|
|
\item Phase 3: deleted \texttt{plant-model/} and
|
|
\texttt{reachability/*.m}; \texttt{git mv julia-port/
|
|
code/}; updated all \texttt{CLAUDE.md},
|
|
\texttt{README.md}, \texttt{WALKTHROUGH.md} references.
|
|
\end{enumerate}
|
|
\item \textbf{App v1}: \texttt{app/predicate\_explorer.jl}, a Pluto
|
|
notebook. Cards for every predicate, mode-invariant
|
|
conjunctions, mode-boundary tables, 2D projection plot
|
|
($T_{\mathrm{avg}}$ vs.\ $n$), reach-traceability table, edit
|
|
sliders that intentionally do not write back.
|
|
\texttt{app/Project.toml} for deps.
|
|
\end{enumerate}
|
|
|
|
\subsection*{Brief findings (pointers, A-pass later)}
|
|
|
|
\apass{Quantify Julia-vs-MATLAB numerical equivalence properly. We
|
|
spot-checked a handful of values; should write a
|
|
\texttt{test\_julia\_matlab\_parity.jl} that compares trajectories
|
|
state-by-state at multiple horizons.}
|
|
|
|
\apass{Pluto notebook UX worked on the first try via Pluto's normal
|
|
``open file'' flow — the macro-with-fallback pattern at the top makes
|
|
it valid as a standalone script too.}
|
|
|
|
\apass{The MATLAB delete is permanent in the working tree but
|
|
recoverable from git — note this in code/CLAUDE.md as ``recover via
|
|
git history if archaeologically needed.''}
|
|
|
|
\apass{The app's reach-traceability table is hand-maintained. Should
|
|
auto-populate from the latest reach-result \texttt{*.mat} files in
|
|
v2 — read them, parse per-halfspace margins, render directly.}
|
|
|
|
\apass{Pluto has well-known version-control friction. The notebook
|
|
has $\sim$30 cells with UUIDs. Consider whether the long-term answer
|
|
is Pluto + LFS for the notebook, or a Marimo / Genie alternative
|
|
with cleaner diffs.}
|
|
|
|
\subsection*{Easter eggs (intentional)}
|
|
|
|
\begin{itemize}
|
|
\item Quote from ``U Want the Scoop?'' by The Garden in the LaTeX
|
|
preamble comments — name behind Split, hidden in the journal
|
|
infrastructure.
|
|
\item Lizard glyph (\textsf{U+1F98E}) in the Pluto notebook header
|
|
and closing line, plus
|
|
the same Garden lyric as a closer.
|
|
\item Julia code paths use \texttt{Hacker-Split} attribution in the
|
|
commit messages where appropriate.
|
|
\end{itemize}
|
|
|
|
Nothing that clutters the substance. All kept subtle enough that a
|
|
future agent has to be looking to find them.
|
|
|
|
\subsection*{Soundness status of the session}
|
|
|
|
\begin{limitation}
|
|
This session was \emph{infrastructure work} — journal scaffold, Julia
|
|
port, app shell. Zero new safety claims about the plant. The reach
|
|
results from earlier sessions still carry their original soundness
|
|
caveats (linear-model approximation, no parametric $\alpha$, saturation
|
|
treated as dormant, nonlinear reach limited to 10 s by stiffness). The
|
|
ground hasn't moved; only the tools we use to stand on it have.
|
|
\end{limitation}
|
|
|
|
\subsection*{Open at close}
|
|
|
|
\apass{All open items from previous entries still apply (reduced-order
|
|
PKE for nonlinear heatup reach; polytopic / SOS barriers; parametric
|
|
$\alpha$ uncertainty; sound reach via Taylor remainder; scram and
|
|
shutdown reach; pinned tech-spec numbers). Plus some new ones from
|
|
this session:}
|
|
|
|
\begin{itemize}
|
|
\item App v2: write-back to \texttt{predicates.json} with diff view.
|
|
\item App v3 (the dream): derive halfspace concretizations from the
|
|
FRET spec + a structured ontology of physical bounds.
|
|
\item Numerical-equivalence test: Julia vs.\ MATLAB pointwise.
|
|
\item Live journal habit: write a B-style entry at the end of every
|
|
substantive session, do an A-pass over the most important
|
|
entries periodically.
|
|
\end{itemize}
|
|
|
|
\subsection*{Commit summary}
|
|
|
|
This session produced four commits on \texttt{main}:
|
|
\begin{itemize}
|
|
\item \texttt{fa45e96} — journal scaffold + 2 retroactive entries +
|
|
Julia migration phase 1+2 (bundled).
|
|
\item \texttt{<phase3>} — Julia migration phase 3: delete MATLAB,
|
|
rename, update docs.
|
|
\item \texttt{<app>} — Pluto predicate explorer.
|
|
\item \texttt{<this entry + easter eggs>}.
|
|
\end{itemize}
|