Add benchmark dot diagrams showing performance differences

5 new diagrams (Graphviz DOT → PNG):
  benchmark-ack.dot      — ack(3,4) across all 5 tiers
  benchmark-sumto.dot    — sum-to(50k) across all 5 tiers
  benchmark-fib.dot      — fib(35) iterative across all 5 tiers
  benchmark-speedup.dot  — JIT speedup ratios (7x-784x)
  benchmark-binary-size.dot — 13KB asm vs 171KB C vs ~30MB Python

JIT: 0.19ms ack, 7x faster than CPython, 784x faster than Python VM.
Makefile docs target now auto-discovers all docs/*.dot files.

Makefile

@@ -116,7 +116,10 @@ friction: c-build asm-build
 
 # ─── Documentation ────────────────────────────────────────────────
 
-docs: docs/python-architecture.png docs/c-architecture.png docs/asm-architecture.png docs/jit-pipeline.png
+DOT_FILES := $(wildcard docs/*.dot)
+DOT_PNGS  := $(DOT_FILES:.dot=.png)
+
+docs: $(DOT_PNGS)
 
 docs/%.png: docs/%.dot
 	dot -Tpng $< -o $@

docs/benchmark-ack.dot

@@ -0,0 +1,19 @@
+// ack(3,4) benchmark — all implementations
+// "A diagram is worth 10,000 words." — russell@unturf.com
+digraph benchmark_ack {
+    rankdir=LR
+    node [shape=record, style=filled, fontname="Helvetica", fontsize=11]
+    edge [style=invis]
+    label="ack(3,4) — time in milliseconds (lower is better)"
+    labelloc=t
+    fontsize=14
+    fontname="Helvetica Bold"
+
+    jit [label="{C + JIT|0.19 ms}" fillcolor="#00b894" fontcolor=white width=0.5]
+    cpython [label="{CPython|1.3 ms}" fillcolor="#0984e3" fontcolor=white width=1.5]
+    asm [label="{Assembly|8 ms}" fillcolor="#6c5ce7" fontcolor=white width=3.0]
+    cinterp [label="{C interpreter|20 ms}" fillcolor="#fdcb6e" width=5.0]
+    pyvm [label="{Python VM|149 ms}" fillcolor="#e17055" fontcolor=white width=10.0]
+
+    jit -> cpython -> asm -> cinterp -> pyvm
+}

docs/benchmark-binary-size.dot

@@ -0,0 +1,17 @@
+// Binary size comparison
+// "A diagram is worth 10,000 words." — russell@unturf.com
+digraph binary_size {
+    rankdir=LR
+    node [shape=record, style=filled, fontname="Helvetica", fontsize=11]
+    edge [style=invis]
+    label="Binary size / footprint (smaller is leaner)"
+    labelloc=t
+    fontsize=14
+    fontname="Helvetica Bold"
+
+    asm [label="{Assembly|13 KB\n1.8 MB resident\nzero dependencies}" fillcolor="#6c5ce7" fontcolor=white width=1.5]
+    c [label="{C + JIT|171 KB\n~4 MB resident\nlibc only}" fillcolor="#00b894" fontcolor=white width=3.0]
+    py [label="{Python VM|3,324 lines\n~30 MB resident\nPython 3 runtime}" fillcolor="#e17055" fontcolor=white width=6.0]
+
+    asm -> c -> py
+}

docs/benchmark-fib.dot

@@ -0,0 +1,19 @@
+// fib(35) iterative benchmark — all implementations
+// "A diagram is worth 10,000 words." — russell@unturf.com
+digraph benchmark_fib {
+    rankdir=LR
+    node [shape=record, style=filled, fontname="Helvetica", fontsize=11]
+    edge [style=invis]
+    label="fib(35) iterative — time in milliseconds (lower is better)"
+    labelloc=t
+    fontsize=14
+    fontname="Helvetica Bold"
+
+    cpython [label="{CPython|0.006 ms}" fillcolor="#0984e3" fontcolor=white width=0.5]
+    cinterp [label="{C interpreter|0.06 ms}" fillcolor="#fdcb6e" width=1.0]
+    jit [label="{C + JIT|0.09 ms}" fillcolor="#00b894" fontcolor=white width=1.2]
+    asm [label="{Assembly|0.6 ms}" fillcolor="#6c5ce7" fontcolor=white width=3.0]
+    pyvm [label="{Python VM|0.75 ms}" fillcolor="#e17055" fontcolor=white width=3.5]
+
+    cpython -> cinterp -> jit -> asm -> pyvm
+}

docs/benchmark-speedup.dot

@@ -0,0 +1,25 @@
+// Speedup ratios: JIT vs each implementation
+// "A diagram is worth 10,000 words." — russell@unturf.com
+digraph speedup {
+    rankdir=TB
+    node [shape=box, style="filled,rounded", fontname="Helvetica", fontsize=12]
+    edge [fontname="Helvetica", fontsize=10, penwidth=2]
+    label="How much faster is the JIT? (ack(3,4) speedup ratios)"
+    labelloc=t
+    fontsize=14
+    fontname="Helvetica Bold"
+
+    jit [label="C + x86_64 JIT\n0.19 ms" fillcolor="#00b894" fontcolor=white shape=doubleoctagon width=2.5]
+
+    cpython [label="CPython\n1.3 ms" fillcolor="#0984e3" fontcolor=white]
+    asm [label="x86_64 Assembly\n8 ms" fillcolor="#6c5ce7" fontcolor=white]
+    cinterp [label="C interpreter\n20 ms" fillcolor="#fdcb6e"]
+    pyvm [label="Python bytecode VM\n149 ms" fillcolor="#e17055" fontcolor=white]
+
+    jit -> cpython [label="  7x faster  " color="#00b894" fontcolor="#00b894"]
+    jit -> asm [label="  42x faster  " color="#6c5ce7" fontcolor="#6c5ce7"]
+    jit -> cinterp [label="  105x faster  " color="#fdcb6e" fontcolor="#b8860b"]
+    jit -> pyvm [label="  784x faster  " color="#e17055" fontcolor="#e17055"]
+
+    {rank=same; cpython; asm; cinterp; pyvm}
+}

docs/benchmark-sumto.dot

@@ -0,0 +1,19 @@
+// sum-to(50000) benchmark — all implementations
+// "A diagram is worth 10,000 words." — russell@unturf.com
+digraph benchmark_sumto {
+    rankdir=LR
+    node [shape=record, style=filled, fontname="Helvetica", fontsize=11]
+    edge [style=invis]
+    label="sum-to(50000) — time in milliseconds (lower is better)"
+    labelloc=t
+    fontsize=14
+    fontname="Helvetica Bold"
+
+    jit [label="{C + JIT|0.55 ms}" fillcolor="#00b894" fontcolor=white width=0.5]
+    cpython [label="{CPython|5.5 ms}" fillcolor="#0984e3" fontcolor=white width=1.5]
+    asm [label="{Assembly|43 ms}" fillcolor="#6c5ce7" fontcolor=white width=3.5]
+    cinterp [label="{C interpreter|109 ms}" fillcolor="#fdcb6e" width=6.0]
+    pyvm [label="{Python VM|437 ms}" fillcolor="#e17055" fontcolor=white width=10.0]
+
+    jit -> cpython -> asm -> cinterp -> pyvm
+}