1% Illustrative diagram: the six-stage autonomous discovery pipeline of WP34 (implemented in experiments/automated_discovery).2\begin{tikzpicture}[font=\small, >=Latex,3 st/.style={draw, rounded corners=3pt, fill=blue!7, minimum height=1.55cm, text width=3.3cm, align=center},4 art/.style={draw, dashed, rounded corners=3pt, fill=gray!6, text width=3.3cm, align=center, font=\footnotesize}]5 \node[st] (s1) at (0,0) {1. Literature extraction\\ \footnotesize structured DB, verified BibTeX};6 \node[st] (s2) at (4.3,0) {2. Hypothesis generation\\ \footnotesize 25 structured records, falsifiable};7 \node[st] (s3) at (8.6,0) {3. Symbolic investigation\\ \footnotesize SymPy / PARI exact checks, identity search};8 \node[st] (s4) at (8.6,-3.4) {4. Experimental evaluation\\ \footnotesize Rust engine, SAT/SMT, benchmarks, seeds};9 \node[st] (s5) at (4.3,-3.4) {5. Formal verification\\ \footnotesize Lean~4 \code{bv\_decide}, no \code{sorry}};10 \node[st] (s6) at (0,-3.4) {6. Scientific evaluation\\ \footnotesize 6-way classification, discovery log};11 \node[art] (a1) at (0,1.7) {\code{literature/db.yaml}, \code{references.bib}};12 \node[art] (a2) at (4.3,1.7) {\code{hypotheses/H*.md}};13 \node[art] (a3) at (8.6,1.7) {\code{results/theory/*}, \code{identity\_search.json}};14 \node[art] (a4) at (8.6,-5.1) {\code{results/**/*.json} with provenance block};15 \node[art] (a5) at (4.3,-5.1) {\code{formal/} theorems, \code{build.log}};16 \node[art] (a6) at (0,-5.1) {\code{discovery\_log.jsonl}, Table~\ref{tab:hypotheses}};17 \draw[->] (s1) -- (s2); \draw[->] (s2) -- (s3); \draw[->] (s3) -- (s4); \draw[->] (s4) -- (s5); \draw[->] (s5) -- (s6);18 \draw[->, dashed] (s6.north) to[out=70,in=-110] node[pos=0.5, left, font=\footnotesize]{refine / reject} (s2.south);19 \draw[->, dashed] (s4.west) to[out=160,in=-20] node[pos=0.5, above, font=\footnotesize, xshift=-2mm]{counterexample} (s3.south west);20 \foreach \a/\s in {a1/s1,a2/s2,a3/s3,a4/s4,a5/s5,a6/s6} { \draw[gray, dotted] (\a) -- (\s); }21\end{tikzpicture}22