diff --git a/.gitignore b/.gitignore
index 204888b..1491482 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,3 +1,7 @@
 **/.*
 result
 gate
+compact/log.txt
+compact/gate-compact
+compact/cnw.c
+compact/gate-compact.c
diff --git a/compact/bottom.c b/compact/bottom.c
new file mode 100644
index 0000000..b00eab7
--- /dev/null
+++ b/compact/bottom.c
@@ -0,0 +1,196 @@
+ return out;
+}
+
+// human attempt, 54 gates
+cell conway_clayton(cell in[9]) {
+    // pairs
+    cell p00 = ~(in[0] | in[1]);
+    cell p10 = ~(in[2] | in[3]);
+    cell p20 = ~(in[5] | in[6]);
+    cell p30 = ~(in[7] | in[8]);
+    cell p01 = in[0] ^ in[1];
+    cell p11 = in[2] ^ in[3];
+    cell p21 = in[5] ^ in[6];
+    cell p31 = in[7] ^ in[8];
+    cell p02 = in[0] & in[1];
+    cell p12 = in[2] & in[3];
+    cell p22 = in[5] & in[6];
+    cell p32 = in[7] & in[8];
+    // halfs
+    cell h00 = p00 & p10;
+    cell h10 = p20 & p30;
+    cell h01 = (p00 & p11) | (p10 & p01);
+    cell h11 = (p20 & p31) | (p30 & p21);
+    cell h03 = (p02 & p11) | (p12 & p01);
+    cell h13 = (p22 & p31) | (p32 & p21);
+    cell h02 = (p02 & p10) | (p00 & p12) | (p01 & p11);
+    cell h12 = (p22 & p30) | (p20 & p32) | (p21 & p31);
+    // neighbors
+    cell n2 = (h01 & h11) | (h02 & h10) | (h00 & h12);
+    cell n3 = (h03 & h10) | (h13 & h00) | (h02 & h11 ) | (h12 & h01);
+    // rule
+    cell out = n3 | (in[4] & n2);
+    return out;
+}
+
+typedef struct {
+    cell* cells;
+    size_t cells_len;
+    size_t width;
+    size_t height;
+} board_t;
+
+void fatal(const char* message) {
+    fprintf(stderr, "fatal: %s\n", message);
+    exit(1);
+}
+
+board_t board_new(size_t width, size_t height) {
+    if (width % 64 != 0) { fatal("board width must be multiple of 64"); }
+
+    size_t cells_len = (size_t)(width / 64) * height;
+    cell* cells = malloc(cells_len * sizeof(cell));
+    if (cells == NULL) { fatal("not able to allocate memory for board "); }
+
+    board_t board = { cells, cells_len, width, height };
+    return board;
+}
+
+// https://en.wikipedia.org/wiki/Xorshift
+// constant is frac(golden_ratio) * 2^64
+// global state bad cry me a river
+uint64_t rand_state = 0x9e3779b97f4a7c55;
+
+uint64_t rand_uint64_t() {
+    uint64_t x = rand_state;
+    x ^= x << 13;
+    x ^= x >> 7;
+    x ^= x << 17;
+    rand_state = x;
+    return x;
+}
+
+cell rand_cell() {
+    // return 0x0101010101010101;
+    return rand_uint64_t();
+}
+
+void rand_board_mut(board_t *board) {
+    for (size_t i = 0; i < board->cells_len; ++i) {
+        board->cells[i] = rand_cell();
+    }
+}
+
+void board_debug(board_t *board) {
+    size_t cells_per_row = board->width / 64;
+    for (size_t i = 0; i < board->height; ++i) {
+        for (size_t j = 0; j < cells_per_row; ++j) {
+            cell current_cell = board->cells[i * cells_per_row + j];
+            for (int k = 0; k < 64; k++) {
+                if ((current_cell >> k) & 1) {
+                    printf("█ ");
+                } else {
+                    printf("  ");
+                }
+            }
+        }
+        printf("|\n");
+    }
+    for (size_t i = 0; i < board->width * 2; ++i) {
+        printf("-");
+    }
+    printf("\n");
+}
+
+void board_step_scratch_mut(
+    board_t *board,
+    board_t *scratch_left,
+    board_t *scratch_right
+) {
+    size_t cells_per_row = board->width / 64;
+    for (size_t i = 0; i < board->height; ++i) {
+        cell msb_prev = (board->cells[i * cells_per_row + cells_per_row - 1] >> 63) & 1;
+        for (size_t j = 0; j < cells_per_row; ++j) {
+            size_t idx = i * cells_per_row + j;
+            cell cell_curr = board->cells[idx];
+            cell cell_shift = cell_curr << 1;
+            cell msb_curr = (cell_curr >> 63) & 1;
+            scratch_left->cells[idx] = cell_shift | msb_prev;
+            msb_prev = msb_curr;
+        }
+    }
+    for (size_t i = 0; i < board->height; ++i) {
+        cell lsb_next = board->cells[i * cells_per_row] & 1;
+        for (size_t j = cells_per_row; j-- > 0; ) {
+            size_t idx = i * cells_per_row + j;
+            cell cell_curr = board->cells[idx];
+            cell cell_shift = cell_curr >> 1;
+            cell lsb_curr = cell_curr & 1;
+            scratch_right->cells[idx] = cell_shift | (lsb_next << 63);
+            lsb_next = lsb_curr;
+        }
+    }
+}
+
+void board_step_mut(
+    board_t *board,
+    board_t *s_left, // scratch
+    board_t *s_right,
+    board_t *s_out
+) {
+    board_step_scratch_mut(board, s_left, s_right);
+
+    size_t step = board->width / 64;
+    size_t wrap = board->cells_len;
+
+    for (size_t i = 0; i < board->cells_len; i++) {
+        cell in[9];
+        size_t i_top = (i + wrap - step) % wrap;
+        size_t i_bottom = (i + step) % wrap;
+        // top row
+        in[0] = s_left->cells[i_top];
+        in[1] = board->cells[i_top];
+        in[2] = s_right->cells[i_top];
+        // middle row
+        in[3] = s_left->cells[i];
+        in[4] = board->cells[i];
+        in[5] = s_right->cells[i];
+        // bottom row
+        in[6] = s_left->cells[i_bottom];
+        in[7] = board->cells[i_bottom];
+        in[8] = s_right->cells[i_bottom];
+        // update output
+        s_out->cells[i] = conway(in);
+    }
+
+    // double-buffering
+    cell* tmp_cells = board->cells;
+    board->cells = s_out->cells;
+    s_out->cells = tmp_cells;
+}
+
+#include <unistd.h>
+
+int main() {
+    size_t width = 512;
+    size_t height = 512;
+
+    board_t board = board_new(width, height);
+    board_t sl = board_new(width, height);
+    board_t sr = board_new(width, height);
+    board_t so = board_new(width, height);
+
+    rand_board_mut(&board);
+
+    for (size_t count = 0; count < 100000; count++) {
+        // vvv comment out for benchmarking
+        printf("\033[H");
+        board_debug(&board);
+        printf("Step: %zu\n", count);
+        sleep(1);
+        // ^^^ comment out for benchmarking
+        board_step_mut(&board, &sl, &sr, &so);
+    }
+
+    printf("done!\n");
+}
diff --git a/compact/compact.sh b/compact/compact.sh
new file mode 100755
index 0000000..e934bf7
--- /dev/null
+++ b/compact/compact.sh
@@ -0,0 +1,52 @@
+# cut out just the generated portion (simple enough, no need to modify python part)
+cat ../gate.c | head -n 190 | tail -n 183 > cnw.c
+
+defs=
+
+rm log.txt -f;
+
+for vr in $(awk -F' ' '{ print $2 }' cnw.c); do
+
+ printf -- "\n--> $vr <--\n" >> log.txt
+
+ # find all variables that only get used once
+ grep -n -E "[ (]~?$vr[ ;)]" cnw.c > tmp.grep;
+ if [ $(wc -l tmp.grep | sed "s/ [^ ]*$//") -eq 2 ]; then
+
+echo "tmp.grep: $(cat tmp.grep)" >> log.txt
+
+  # find the definition of the variable
+  def=$(sed -n "1s/:.*\$//p" tmp.grep)
+
+  defs="$defs\n${def}d"
+
+echo "def=$def" >> log.txt
+
+  # find the only use of the variable
+  use=$(sed -n "2s/:.*\$//p" tmp.grep)
+
+echo "use=$use" >> log.txt
+
+  # find just the definition's rvalue
+  def2=$(sed -n "${def}p" cnw.c | sed -E "s/ *cell [^ ]+ = //" | sed -E "s/;//" | sed "s/&/\\\&/g")
+
+echo "def2=$def2" >> log.txt
+
+  # replace the single occurence with def2
+  sed -i -E "${use}s/([ (]~?)$vr([ ;)])/\1($def2)\2/" cnw.c
+
+ fi
+
+ rm tmp.grep
+
+done
+
+sed -i "$(echo -e $defs)" cnw.c
+
+# combine together, because I don't know how to use the template
+cat top.c cnw.c bottom.c > gate-compact.c
+
+CC="${CC:-gcc}"
+
+# compile using gcc or your compiler
+$CC gate-compact.c -o gate-compact
diff --git a/compact/top.c b/compact/top.c
new file mode 100644
index 0000000..ba4271a
--- /dev/null
+++ b/compact/top.c
@@ -0,0 +1,9 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+typedef uint64_t cell;
+
+cell conway(cell in[9]) {
+
+