diff --git a/.editorconfig b/.editorconfig
index 0a31b58..1c51663 100644
--- a/.editorconfig
+++ b/.editorconfig
@@ -8,6 +8,6 @@ insert_final_newline = true
 indent_style = space
 indent_size = 2
 
-[Makefile]
-indent_style = tab
-indent_size = 4
+[meson.build]
+indent_style = space
+indent_size = 2
diff --git a/README.md b/README.md
index b282a8e..15b4d47 100644
--- a/README.md
+++ b/README.md
@@ -1,14 +1,3 @@
-# wolfscheme
+# wolflisp
 
-An experiment in how quick I can get from zero to a decent Lisp.
-Don't know why I called it `wolfscheme`.
-
-## Day 1:
-
-- Symbol interning
-- Garbage collector and cons pairs
-- Object printing
-
-## Day 2:
-
-- Tree-walking evaluator with a few primitives
+a lisp but wolfy
diff --git a/include/wolflisp.h b/include/wolflisp.h
new file mode 100644
index 0000000..05dff12
--- /dev/null
+++ b/include/wolflisp.h
@@ -0,0 +1,167 @@
+#ifndef WOLFLISP_H
+#define WOLFLISP_H
+
+#include <stddef.h>
+#include <stdint.h>
+
+/// * Behavioral macros
+#define GC_DEBUG 1
+
+/// * Type declarations
+typedef void V;
+typedef intptr_t I;
+typedef uintptr_t U;
+typedef char C;
+typedef uint8_t U8;
+typedef uint32_t U32;
+typedef int32_t I32;
+typedef size_t Z;
+
+// Object
+typedef U O;
+
+#define NIL ((O)0)
+#define IMM(x) ((O)(x) & (O)1)
+#define NUM(x) (((O)((I)(x) << 1)) | (O)1)
+#define ORD(x) ((I)(x) >> 1)
+
+// Pair
+typedef struct Pa {
+  O head, tail;
+} Pa;
+
+// Symbol
+typedef struct Sy {
+  U32 hash;
+  Z len;
+  U8 *data;
+} Sy;
+
+// Closure
+typedef struct Cl {
+  O args, body, env;
+} Cl;
+
+// Primitive
+typedef struct In In;
+typedef struct Pr {
+  const char *name;
+  O (*fn)(In *, O, O);
+} Pr;
+
+// Symbol table
+typedef struct St {
+  Z count;
+  Z capacity;
+  Sy **data;
+} St;
+
+#define HEAP_BYTES (1024 * 1024)
+#define TYPE_MASK 7
+
+enum {
+  TAG_MAN = 0,  // GC-managed object
+  TAG_IMM = 1,  // Immediate number
+  TAG_SYM = 2,  // Pointer to symbol
+  TAG_PRIM = 4, // Pointer to primitive
+};
+
+enum {
+  TYPE_NIL = 0,
+  TYPE_NUM = 1,
+  TYPE_SYM = 2,
+  TYPE_PRIM = 4,
+  TYPE_PAIR, // = 5,
+  TYPE_CLOS, // = 6,
+  TYPE_CODE, // = 7,
+  TYPE_FWD,  // = 8,
+  TYPE__MAX,
+};
+
+#define TAG_OF(x) (((U)(x)) & TYPE_MASK)
+#define UNTAG(x) (((U)(x)) & ~TYPE_MASK)
+#define TAG(x, t) (V *)(((U)(x)) | t)
+
+// GC-managed header
+typedef struct Gh {
+  U32 type;
+  U32 size;
+} Gh;
+
+#define BOX(x) ((O)(x))
+#define UNBOX(x) ((Gh *)(x))
+
+// GC space
+typedef struct Gs {
+  U8 *start, *end;
+  U8 *free;
+} Gs;
+
+// GC context
+typedef struct Gc {
+  Gs from, to;
+  struct {
+    Z count;
+    Z capacity;
+    O **data;
+  } roots;
+} Gc;
+
+// Interpreter context
+typedef struct In {
+  Gc gc;
+  St symtab;
+  O env;
+} In;
+
+/// * Function declarations
+
+// Get the type of an object
+I type(O obj);
+// Get the name of a type
+const char *typename(I t);
+
+// Add a root to a GC context.
+V gc_addroot(Gc *gc, O *root);
+// Mark the current root state in a GC context.
+I gc_rootmark(Gc *gc);
+// Reset the root state in a GC context to a previously marked state.
+V gc_rootreset(Gc *gc, I mark);
+// Perform a garbage collection in a GC context.
+V gc_collect(Gc *gc);
+// Allocate memory in a GC context.
+Gh *gc_alloc(Gc *gc, Z sz);
+// Initialize a GC context.
+V gc_init(Gc *gc);
+// Finalize a GC context.
+V gc_finalize(Gc *gc);
+
+// Initialize an interpreter context.
+V interp_init(In *in);
+// Finalize an interpreter context.
+V interp_finalize(In *in);
+
+// Evaluate a list of values.
+O interp_eval_list(In *in, O list, O env);
+
+// Evaluate an expression.
+O interp_eval(In *in, O obj, O env);
+
+// Intern a string
+Sy *intern(St *tab, const char *str, Z len);
+
+// Create a pair
+O pair_make(Gc *gc, O head, O tail);
+// Unwrap a pair
+Pa *pair_unwrap(O obj);
+
+V print(O obj);
+V println(O obj);
+
+O symbol_make(In *in, const char *str);
+O prim_make(In *in, const char *name, O (*fn)(In *, O, O));
+
+O list_assoc(O key, O alist);
+O list_reverse(O list);
+
+#endif
diff --git a/meson.build b/meson.build
new file mode 100644
index 0000000..7d5f4e7
--- /dev/null
+++ b/meson.build
@@ -0,0 +1,27 @@
+project(
+  'wolflisp',
+  'c',
+  meson_version : '>= 1.3.0',
+  version : '0.1',
+  default_options : ['c_std=gnu11', 'buildtype=debugoptimized', 'warning_level=3'],
+)
+
+inc = include_directories('include', 'src')
+src = [
+  'src/gc.c',
+  'src/interp.c',
+  'src/list.c',
+  'src/main.c',
+  'src/pair.c',
+  'src/prim.c',
+  'src/print.c',
+  'src/symbol.c',
+  'src/type.c',
+]
+
+exe = executable(
+  'wl',
+  src,
+  include_directories : inc,
+  install : true,
+)
diff --git a/Makefile b/old/Makefile
similarity index 63%
rename from Makefile
rename to old/Makefile
index 026da86..f2dfeb2 100644
--- a/Makefile
+++ b/old/Makefile
@@ -1,5 +1,5 @@
 CFLAGS	:= -std=gnu99 -Og -g -Wpedantic -Wall
-OBJS	:= symbol.o object.o gc.o print.o eval.o main.o
+OBJS	:= util.o symbol.o object.o gc.o print.o lex.o read.o eval.o main.o
 
 wscm: $(OBJS)
 	$(CC) $(OBJS) -o wscm
diff --git a/old/README.md b/old/README.md
new file mode 100644
index 0000000..b282a8e
--- /dev/null
+++ b/old/README.md
@@ -0,0 +1,14 @@
+# wolfscheme
+
+An experiment in how quick I can get from zero to a decent Lisp.
+Don't know why I called it `wolfscheme`.
+
+## Day 1:
+
+- Symbol interning
+- Garbage collector and cons pairs
+- Object printing
+
+## Day 2:
+
+- Tree-walking evaluator with a few primitives
diff --git a/eval.c b/old/eval.c
similarity index 94%
rename from eval.c
rename to old/eval.c
index c199cf7..58fa416 100644
--- a/eval.c
+++ b/old/eval.c
@@ -18,7 +18,7 @@ static O listrev(O list) {
   return prev;
 }
 
-O assoc(O v, O env) {
+O pair_assoc(O v, O env) {
   while (kind(env) == KIND_CONS) {
     C *c = uncons(env);
     O pair = c->car;
@@ -164,6 +164,12 @@ O mkprim(const char *name, O (*fn)(O, O)) {
 }
 
 V setupenv(O *env) {
+  O t = BOX(TAG(intern("t", 1), TAG_SYM));
+  *env = cons(cons(t, t), *env);
+
+  O nil = BOX(TAG(intern("nil", 3), TAG_SYM));
+  *env = cons(cons(nil, NIL), *env);
+
   *env = cons(mkprim("cons", prim_cons), *env);
   *env = cons(mkprim("car", prim_car), *env);
   *env = cons(mkprim("cdr", prim_cdr), *env);
@@ -195,7 +201,7 @@ O eval(O obj, O env) {
   I k = kind(obj);
 
   if (k == KIND_SYM) {
-    O pair = assoc(obj, env);
+    O pair = pair_assoc(obj, env);
     if (pair == NIL) {
       S *s = (S *)UNTAG(obj);
       fprintf(stderr, "error: undefined symbol '%.*s'\n", (int)s->len, s->data);
diff --git a/gc.c b/old/gc.c
similarity index 100%
rename from gc.c
rename to old/gc.c
diff --git a/old/lex.c b/old/lex.c
new file mode 100644
index 0000000..9e50cd9
--- /dev/null
+++ b/old/lex.c
@@ -0,0 +1,119 @@
+
+#include <err.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "wscm.h"
+
+// helpers
+static int iswhite(int ch) { return ch == ' ' || ch == '\t' || ch == '\n'; }
+
+static int isdelim(int ch) {
+  return ch == '(' || ch == ')' || ch == '\'' || ch == ';';
+}
+
+static inline void appendchar(Lx *lex, char ch) {
+  if (lex->cursor > LEXER_CAP) {
+    fprintf(stderr, "lexer buffer overflow");
+    abort();
+  }
+  lex->buffer[lex->cursor++] = ch;
+}
+
+static int getcws(Lx *lex) {
+  if (feof(lex->input))
+    return EOF;
+  for (;;) {
+    int ch = getc(lex->input);
+    if (iswhite(ch))
+      continue;
+    return ch;
+  }
+}
+
+static int scanword(Lx *lex) {
+  int ch = getc(lex->input);
+  for (;;) {
+    if (ch == EOF) {
+      if (lex->cursor == 0)
+        lex->kind = TOK_EOF;
+      appendchar(lex, 0);
+      return lex->kind;
+    } else if (iswhite(ch) || isdelim(ch)) {
+      ungetc(ch, lex->input);
+      appendchar(lex, 0);
+      return lex->kind;
+    } else {
+      appendchar(lex, ch);
+      ch = getc(lex->input);
+    }
+  }
+}
+
+static int scanstring(Lx *lex) {
+  int ch;
+  for (;;) {
+    ch = getc(lex->input);
+    switch (ch) {
+    case EOF:
+      goto eof;
+    case '\\':
+      ch = getc(lex->input);
+      if (ch == EOF)
+        goto eof;
+      switch (ch) {
+      case 'n':
+        appendchar(lex, '\n');
+        break;
+      case 't':
+        appendchar(lex, '\t');
+        break;
+      case '"':
+        appendchar(lex, '"');
+        break;
+      }
+      break;
+    case '"':
+      appendchar(lex, 0);
+      return (lex->kind = TOK_STRING);
+    default:
+      appendchar(lex, ch);
+    }
+  }
+
+eof:
+  errx(1, "unterminated string literal");
+}
+
+int nexttoken(Lx *lex) {
+  int ch;
+  lex->cursor = 0;
+
+  if (feof(lex->input)) {
+    lex->kind = TOK_EOF;
+    *lex->buffer = 0;
+    return 0;
+  }
+
+  ch = getcws(lex);
+  switch (ch) {
+  case ';':
+    for (; ch != '\n'; ch = getc(lex->input))
+      appendchar(lex, ch);
+    appendchar(lex, 0);
+    return (lex->kind = TOK_COMMENT);
+  case '(':
+  case ')':
+  case '.':
+  case '\'':
+    return (lex->kind = ch);
+  case '"':
+    return scanstring(lex);
+  default:
+    ungetc(ch, lex->input);
+    lex->kind = TOK_WORD;
+    return scanword(lex);
+  }
+
+  return 0;
+}
diff --git a/main.c b/old/main.c
similarity index 77%
rename from main.c
rename to old/main.c
index e4b1b7c..3f0e751 100644
--- a/main.c
+++ b/old/main.c
@@ -10,7 +10,7 @@ int main(void) {
   setupenv(&env);
 
   // IF
-  O code = cons(SYM("if"), cons(NIL, cons(NUM(100), cons(NUM(200), NIL))));
+  O code = readfile(stdin);
   addroot(&code);
 
   // collect();
diff --git a/object.c b/old/object.c
similarity index 100%
rename from object.c
rename to old/object.c
diff --git a/print.c b/old/print.c
similarity index 97%
rename from print.c
rename to old/print.c
index 906825e..a22ae71 100644
--- a/print.c
+++ b/old/print.c
@@ -5,7 +5,7 @@
 
 void print(O obj);
 
-void printcons(O obj) {
+static void printcons(O obj) {
   O c = obj;
   I f = 1;
 
diff --git a/old/read.c b/old/read.c
new file mode 100644
index 0000000..235cdde
--- /dev/null
+++ b/old/read.c
@@ -0,0 +1,156 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "wscm.h"
+
+static O read(Lx *lex);
+
+static void skipcomments(Lx *lex) {
+  while (lex->kind == TOK_COMMENT) {
+    if (!nexttoken(lex))
+      break;
+  }
+}
+
+static O makeobject(Lx *lex) {
+  if (lex->kind == TOK_WORD) {
+    char *tok = lex->buffer;
+    char *end;
+    long v = strtol(tok, &end, 10);
+    if (end != tok && *end == '\0') {
+      nexttoken(lex);
+      return NUM((I)v);
+    } else {
+      S *s = intern(tok, (I)strlen(tok));
+      O sym = BOX(TAG(s, TAG_SYM));
+      nexttoken(lex);
+      return sym;
+    }
+  } else if (lex->kind == TOK_STRING) {
+    // TODO: string type
+    char *tok = lex->buffer;
+    S *s = intern(tok, (I)strlen(tok));
+    O sym = BOX(TAG(s, TAG_SYM));
+    nexttoken(lex);
+    return sym;
+  }
+  return NIL;
+}
+
+static void lastcdr(O list, O cdr_val) {
+  O curr = list;
+  while (1) {
+    C *c = uncons(curr);
+    if (c->cdr == NIL) {
+      c->cdr = cdr_val;
+      return;
+    }
+    curr = c->cdr;
+  }
+}
+
+static O readlist(Lx *lex) {
+  nexttoken(lex);
+  skipcomments(lex);
+
+  if (lex->kind == TOK_RPAREN) {
+    nexttoken(lex);
+    return NIL;
+  }
+
+  O head = NIL;
+  while (lex->kind != TOK_EOF) {
+    skipcomments(lex);
+    if (lex->kind == TOK_RPAREN) {
+      nexttoken(lex);
+      break;
+    }
+
+    if (lex->kind == TOK_DOT) {
+      nexttoken(lex);
+      skipcomments(lex);
+      if (lex->kind == TOK_EOF) {
+        fprintf(stderr, "reader error: unexpected EOF after '.'\n");
+        abort();
+      }
+      O cdr_val = read(lex);
+      skipcomments(lex);
+      if (lex->kind != TOK_RPAREN) {
+        fprintf(stderr, "reader error: expected ')' after dotted pair cdr\n");
+        abort();
+      }
+      nexttoken(lex);
+      O normal = listreverse(head);
+      if (normal == NIL) {
+        fprintf(stderr, "reader error: '.' with no preceding elements\n");
+        abort();
+      }
+      lastcdr(normal, cdr_val);
+      return normal;
+    }
+    O elem = read(lex);
+    head = cons(elem, head);
+    skipcomments(lex);
+  }
+
+  return listreverse(head);
+}
+
+static O readquote(Lx *lex) {
+  nexttoken(lex);
+  skipcomments(lex);
+  O e = read(lex);
+  O qsym = BOX(TAG(intern("quote", -1), TAG_SYM));
+  return cons(qsym, cons(e, NIL));
+}
+
+static O read(Lx *lex) {
+  skipcomments(lex);
+
+  switch (lex->kind) {
+  case TOK_EOF:
+    return NIL;
+  case TOK_LPAREN:
+    return readlist(lex);
+  case TOK_QUOTE:
+    return readquote(lex);
+  case TOK_WORD:
+  case TOK_STRING:
+    return makeobject(lex);
+  case TOK_COMMENT:
+    nexttoken(lex);
+    return read(lex);
+  default:
+    nexttoken(lex);
+    return NIL;
+  }
+}
+
+O readfile(FILE *f) {
+  if (!f)
+    return NIL;
+
+  Lx lex;
+  lex.kind = TOK_EOF;
+  lex.cursor = 0;
+  lex.input = f;
+  lex.buffer[0] = '\0';
+
+  nexttoken(&lex);
+  skipcomments(&lex);
+
+  if (lex.kind == TOK_EOF)
+    return NIL;
+  return read(&lex);
+}
+
+O readstring(const char *s) {
+  if (!s)
+    return NIL;
+  size_t len = strlen(s);
+  FILE *f = fmemopen((void *)s, len, "r");
+  O res = readfile(f);
+  fclose(f);
+  return res;
+}
diff --git a/symbol.c b/old/symbol.c
similarity index 100%
rename from symbol.c
rename to old/symbol.c
diff --git a/old/util.c b/old/util.c
new file mode 100644
index 0000000..55fd0ff
--- /dev/null
+++ b/old/util.c
@@ -0,0 +1,16 @@
+#include "wscm.h"
+
+O listreverse(O list) {
+  O prev = NIL;
+  O curr = list;
+  O next;
+
+  while (curr != NIL) {
+    C *c = uncons(curr);
+    next = c->cdr;
+    c->cdr = prev;
+    prev = curr;
+    curr = next;
+  }
+  return prev;
+}
diff --git a/wscm.h b/old/wscm.h
similarity index 85%
rename from wscm.h
rename to old/wscm.h
index 351eea7..22b447a 100644
--- a/wscm.h
+++ b/old/wscm.h
@@ -131,8 +131,36 @@ O mksym(const char *str);
 O cons(O head, O tail);
 C *uncons(O obj);
 
+O listreverse(O list);
+
 V setupenv(O *env);
 O eval(O obj, O env);
 
 void print(O obj);
 void println(O obj);
+
+enum {
+	TOK_EOF = 0,
+	TOK_COMMENT = ';',
+	TOK_WORD = 'a',
+	TOK_LPAREN = '(',
+	TOK_RPAREN = ')',
+	TOK_STRING = '"',
+	TOK_QUOTE = '\'',
+	TOK_DOT = '.',
+};
+
+#include <stdio.h>
+
+#define LEXER_CAP 1024
+
+typedef struct Lx {
+	int kind;
+	int cursor;
+	FILE *input;
+	char buffer[1024];
+} Lx;
+
+int nexttoken(Lx *lex);
+O readfile(FILE *f);
+O readstring(const char *s);
diff --git a/shell.nix b/shell.nix
index 4870fe9..fcf4e51 100644
--- a/shell.nix
+++ b/shell.nix
@@ -1,7 +1,10 @@
 { pkgs ? import <nixpkgs> {} }:
 pkgs.mkShell {
+  name = "rufus";
   packages = with pkgs; [
     clang-tools
+    meson
+    ninja
     bear
     gdb
   ];
diff --git a/src/gc.c b/src/gc.c
new file mode 100644
index 0000000..b6c1342
--- /dev/null
+++ b/src/gc.c
@@ -0,0 +1,171 @@
+#include <assert.h>
+#include <inttypes.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include <wolflisp.h>
+
+#define ALIGN(n) (((n) + 7) & ~7)
+static inline I infrom(Gc *gc, V *ptr) {
+  const U8 *x = (const U8 *)ptr;
+  return (x >= gc->from.start && x < gc->from.end);
+}
+
+V gc_addroot(Gc *gc, O *root) {
+  if (gc->roots.count >= gc->roots.capacity) {
+    Z newcap = gc->roots.capacity == 0 ? 16 : gc->roots.capacity * 2;
+    O **newdata = realloc(gc->roots.data, newcap * sizeof(O *));
+    if (!newdata) {
+      fprintf(stderr, "fatal error: failed to expand roots array\n");
+      abort();
+    }
+    gc->roots.capacity = newcap;
+    gc->roots.data = newdata;
+  }
+  gc->roots.data[gc->roots.count++] = root;
+}
+
+I gc_rootmark(Gc *gc) { return gc->roots.count; }
+V gc_rootreset(Gc *gc, I mark) { gc->roots.count = mark; }
+
+static O copy(Gc *gc, Gh *hdr) {
+  assert(infrom(gc, hdr));
+  assert(hdr->type != TYPE_FWD);
+
+  // Copy the object to the to-space and leave a forwarding pointer behind
+  Z sz = ALIGN(hdr->size);
+  Gh *new = (Gh *)gc->to.free;
+  gc->to.free += sz;
+  memcpy(new, hdr, sz);
+  hdr->type = TYPE_FWD;
+  O *obj = (O *)(hdr + 1);
+  *obj = BOX(new);
+
+  // Return the new object
+  return *obj;
+}
+
+static O forward(Gc *gc, O obj) {
+  if (obj == NIL)
+    return NIL;
+  if (IMM(obj))
+    return obj;
+  if (TAG_OF(obj) != TAG_MAN)
+    return obj;
+  if (!infrom(gc, (V *)obj))
+    return obj;
+
+  Gh *hdr = UNBOX(obj);
+
+  // If the object to copy is already a forwarding pointer, return the object
+  // pointed to.
+  if (hdr->type == TYPE_FWD) {
+    O *o = (O *)(hdr + 1);
+    return *o;
+  } else {
+    return copy(gc, hdr);
+  }
+}
+
+#if GC_DEBUG
+static V printstats(Gc *gc, const char *label) {
+  size_t used = (Z)(gc->from.free - gc->from.start);
+  fprintf(stderr, "[%s] used=%zu/%zu bytes (%.1f%%)\n", label, used,
+          (Z)HEAP_BYTES, (double)used / (double)HEAP_BYTES * 100.0);
+}
+#endif
+
+V gc_collect(Gc *gc) {
+  U8 *scan = gc->to.free;
+
+#if GC_DEBUG
+  printstats(gc, "before GC");
+#endif
+
+  // Forward roots to the to-space.
+  for (Z i = 0; i < gc->roots.count; i++) {
+    O *o = gc->roots.data[i];
+    *o = forward(gc, *o);
+  }
+
+  // Scan to-space for objects to forward (breadth-first iteration)
+  while (scan < gc->to.free) {
+    Gh *hdr = (Gh *)scan;
+    switch (hdr->type) {
+    case TYPE_PAIR: {
+      Pa *obj = (Pa *)(hdr + 1);
+      obj->head = forward(gc, obj->head);
+      obj->tail = forward(gc, obj->tail);
+      break;
+    }
+    case TYPE_CLOS: {
+      Cl *obj = (Cl *)(hdr + 1);
+      obj->args = forward(gc, obj->args);
+      obj->body = forward(gc, obj->body);
+      obj->env = forward(gc, obj->env);
+      break;
+    }
+    case TYPE_FWD:
+      fprintf(stderr, "fatal GC error: forwarding pointer in to-space\n");
+      abort();
+    default:
+      fprintf(stderr, "GC warning: junk object type %" PRId32 "\n", hdr->type);
+    }
+    scan += ALIGN(hdr->size);
+  }
+
+  // Swap to- and from-spaces
+  Gs tmp = gc->from;
+  gc->from = gc->to;
+  gc->to = tmp;
+  gc->to.free = gc->to.start;
+
+#if GC_DEBUG
+  printstats(gc, "after GC");
+#endif
+}
+
+Gh *gc_alloc(Gc *gc, Z sz) {
+  sz = ALIGN(sz);
+  if (gc->from.free + sz > gc->from.end) {
+    gc_collect(gc);
+    if (gc->from.free + sz > gc->from.end) {
+      fprintf(stderr, "out of memory (requested %" PRIdPTR "bytes\n", sz);
+      abort();
+    }
+  }
+  Gh *hdr = (Gh *)gc->from.free;
+  gc->from.free += sz;
+  hdr->size = sz;
+  return hdr;
+}
+
+V gc_init(Gc *gc) {
+  gc->from.start = malloc(HEAP_BYTES);
+  if (!gc->from.start)
+    goto fatal;
+  gc->from.end = gc->from.start + HEAP_BYTES;
+  gc->from.free = gc->from.start;
+
+  gc->to.start = malloc(HEAP_BYTES);
+  if (!gc->to.start)
+    goto fatal;
+  gc->to.end = gc->to.start + HEAP_BYTES;
+  gc->to.free = gc->to.start;
+
+  gc->roots.capacity = 0;
+  gc->roots.count = 0;
+  gc->roots.data = NULL;
+  return;
+
+fatal:
+  fprintf(stderr, "failed to allocate heap space\n");
+  abort();
+}
+
+V gc_finalize(Gc *gc) {
+  free(gc->from.start);
+  free(gc->to.start);
+  free(gc->roots.data);
+}
diff --git a/src/interp.c b/src/interp.c
new file mode 100644
index 0000000..5056d16
--- /dev/null
+++ b/src/interp.c
@@ -0,0 +1,156 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include <prim.h>
+#include <wolflisp.h>
+
+V interp_init(In *in) {
+  memset(&in->symtab, 0, sizeof(St));
+  gc_init(&in->gc);
+
+  in->env = NIL;
+  in->env = pair_make(&in->gc, prim_make(in, "cons", prim_cons), in->env);
+  in->env = pair_make(&in->gc, prim_make(in, "head", prim_head), in->env);
+  in->env = pair_make(&in->gc, prim_make(in, "tail", prim_tail), in->env);
+  in->env = pair_make(&in->gc, prim_make(in, "quote", prim_quote), in->env);
+  in->env = pair_make(&in->gc, prim_make(in, "print", prim_print), in->env);
+  in->env = pair_make(&in->gc, prim_make(in, "println", prim_println), in->env);
+}
+
+V interp_finalize(In *in) {
+  free(in->symtab.data);
+  gc_finalize(&in->gc);
+}
+
+static O bind(Gc *gc, O params, O args, O env) {
+  I mark = gc_rootmark(gc);
+  gc_addroot(gc, &params);
+  gc_addroot(gc, &args);
+  gc_addroot(gc, &env);
+
+  O res = env;
+  gc_addroot(gc, &res);
+
+  while (params != NIL) {
+    if (type(params) == TYPE_SYM) {
+      O pair = pair_make(gc, params, args);
+      gc_addroot(gc, &pair);
+      res = pair_make(gc, pair, res);
+      break;
+    }
+
+    if (type(params) != TYPE_PAIR) {
+      fprintf(stderr, "error: expected proper list or symbol for parameters\n");
+      abort();
+    }
+
+    Pa *p = pair_unwrap(params);
+    O sym = p->head;
+
+    O val = NIL;
+    if (args != NIL) {
+      if (type(args) != TYPE_PAIR) {
+        fprintf(stderr, "error: too many parameters for arguments\n");
+        abort();
+      }
+      Pa *a = pair_unwrap(args);
+      val = a->head;
+      args = a->tail;
+    }
+
+    O pair = pair_make(gc, sym, val);
+    gc_addroot(gc, &pair);
+    res = pair_make(gc, pair, res);
+
+    params = p->tail;
+  }
+
+  gc_rootreset(gc, mark);
+  return res;
+}
+
+static O eval(In *in, O obj, O env);
+
+static O eval_list(In *in, O list, O env) {
+  I mark = gc_rootmark(&in->gc);
+  gc_addroot(&in->gc, &list);
+  gc_addroot(&in->gc, &env);
+
+  O head = NIL;
+  O curr = list;
+
+  gc_addroot(&in->gc, &head);
+  gc_addroot(&in->gc, &curr);
+
+  while (curr != NIL) {
+    Pa *p = pair_unwrap(curr);
+    O obj = eval(in, p->head, env);
+    head = pair_make(&in->gc, obj, head);
+    curr = p->tail;
+  }
+
+  O result = list_reverse(head);
+  gc_rootreset(&in->gc, mark);
+  return result;
+}
+
+static O apply(In *in, O fn, O args, O env) {
+  I ty = type(fn);
+  switch (ty) {
+  case TYPE_PRIM: {
+    Pr *pr = (Pr *)UNTAG(fn);
+    return pr->fn(in, args, env);
+  }
+  case TYPE_CLOS: {
+    Gh *hdr = UNBOX(fn);
+    Cl *cl = (Cl *)(hdr + 1);
+    args = eval_list(in, args, env);
+    O nenv = bind(&in->gc, cl->args, args, cl->env);
+    return eval(in, cl->body, nenv);
+  }
+  default:
+    fprintf(stderr, "tried to call non-function value\n");
+    abort();
+  }
+}
+
+static O eval(In *in, O obj, O env) {
+  I ty = type(obj);
+
+  if (ty == TYPE_SYM) {
+    O pair = list_assoc(obj, env);
+    if (pair == NIL) {
+      Sy *s = (Sy *)UNTAG(obj);
+      fprintf(stderr, "error: undefined symbol '%.*s'\n", (int)s->len, s->data);
+      abort();
+    }
+    return pair_unwrap(pair)->tail;
+  } else if (ty != TYPE_PAIR) {
+    return obj;
+  }
+
+  I mark = gc_rootmark(&in->gc);
+  gc_addroot(&in->gc, &obj);
+  gc_addroot(&in->gc, &env);
+
+  Pa *c = pair_unwrap(obj);
+  O fn = eval(in, c->head, env);
+  gc_addroot(&in->gc, &fn);
+
+  O res = apply(in, fn, c->tail, env);
+  gc_rootreset(&in->gc, mark);
+  return res;
+}
+
+O interp_eval_list(In *in, O list, O env) {
+  if (env == NIL)
+    env = in->env;
+  return eval_list(in, list, env);
+}
+
+O interp_eval(In *in, O obj, O env) {
+  if (env == NIL)
+    env = in->env;
+  return eval(in, obj, in->env);
+}
diff --git a/src/list.c b/src/list.c
new file mode 100644
index 0000000..8870337
--- /dev/null
+++ b/src/list.c
@@ -0,0 +1,35 @@
+#include <wolflisp.h>
+
+O list_assoc(O key, O alist) {
+  while (type(alist) == TYPE_PAIR) {
+    Pa *c = pair_unwrap(alist);
+    O pair = c->head;
+    if (pair == NIL) {
+      alist = c->tail;
+      continue;
+    }
+    if (type(pair) == TYPE_PAIR) {
+      Pa *kv = pair_unwrap(pair);
+      if (kv->head == key)
+        return pair;
+    }
+    alist = c->tail;
+  }
+  return NIL;
+}
+
+O list_reverse(O list) {
+  O prev = NIL;
+  O curr = list;
+  O next;
+
+  while (curr != NIL) {
+    Pa *c = pair_unwrap(curr);
+    next = c->tail;
+    c->tail = prev;
+    prev = curr;
+    curr = next;
+  }
+
+  return prev;
+}
diff --git a/src/main.c b/src/main.c
new file mode 100644
index 0000000..d024013
--- /dev/null
+++ b/src/main.c
@@ -0,0 +1,19 @@
+#include <wolflisp.h>
+
+int main(void) {
+  In interp;
+  interp_init(&interp);
+
+  // Build code
+  O s_println = symbol_make(&interp, "println");
+  O s_quote = symbol_make(&interp, "quote");
+
+  O data = pair_make(
+      &interp.gc, s_quote,
+      pair_make(&interp.gc, pair_make(&interp.gc, NUM(1), NUM(2)), NIL));
+
+  O code = pair_make(&interp.gc, s_println, pair_make(&interp.gc, data, NIL));
+  (void)interp_eval(&interp, code, NIL);
+
+  interp_finalize(&interp);
+}
diff --git a/src/pair.c b/src/pair.c
new file mode 100644
index 0000000..b860b0d
--- /dev/null
+++ b/src/pair.c
@@ -0,0 +1,28 @@
+#include "stdlib.h"
+#include <stdio.h>
+#include <wolflisp.h>
+
+O pair_make(Gc *gc, O head, O tail) {
+  I mark = gc_rootmark(gc);
+  gc_addroot(gc, &head);
+  gc_addroot(gc, &tail);
+
+  Z size = sizeof(Gh) + sizeof(Pa);
+  Gh *hdr = gc_alloc(gc, size);
+  hdr->type = TYPE_PAIR;
+
+  Pa *pair = (Pa *)(hdr + 1);
+  pair->head = head;
+  pair->tail = tail;
+
+  gc_rootreset(gc, mark);
+  return BOX(hdr);
+}
+
+Pa *pair_unwrap(O obj) {
+  if (type(obj) != TYPE_PAIR) {
+    fprintf(stderr, "expected pair, got %s\n", typename(type(obj)));
+    abort();
+  }
+  return (Pa *)(UNBOX(obj) + 1);
+}
diff --git a/src/prim.c b/src/prim.c
new file mode 100644
index 0000000..d51f38f
--- /dev/null
+++ b/src/prim.c
@@ -0,0 +1,53 @@
+#include <prim.h>
+#include <stdlib.h>
+#include <wolflisp.h>
+
+O prim_make(In *in, const char *name, O (*fn)(In *, O, O)) {
+  Pr *pr = malloc(sizeof(Pr));
+  pr->name = name;
+  pr->fn = fn;
+  O sym = BOX(TAG(intern(&in->symtab, name, 0), TAG_SYM));
+  O prim = BOX(TAG(pr, TAG_PRIM));
+  return pair_make(&in->gc, sym, prim);
+}
+
+O prim_cons(In *in, O args, O env) {
+  args = interp_eval_list(in, args, env);
+  O head = pair_unwrap(args)->head;
+  args = pair_unwrap(args)->tail;
+  O tail = pair_unwrap(args)->head;
+  return pair_make(&in->gc, head, tail);
+}
+
+O prim_head(In *in, O args, O env) {
+  args = interp_eval_list(in, args, env);
+  return pair_unwrap(pair_unwrap(args)->head)->head;
+}
+
+O prim_tail(In *in, O args, O env) {
+  args = interp_eval_list(in, args, env);
+  return pair_unwrap(pair_unwrap(args)->head)->tail;
+}
+
+O prim_print(In *in, O args, O env) {
+  args = interp_eval_list(in, args, env);
+  O arg = pair_unwrap(args)->head;
+  print(arg);
+  return NIL;
+}
+
+O prim_println(In *in, O args, O env) {
+  args = interp_eval_list(in, args, env);
+  O arg = pair_unwrap(args)->head;
+  println(arg);
+  return NIL;
+}
+
+O prim_quote(In *in, O args, O env) {
+  (void)in;
+  (void)env;
+
+  if (args == NIL)
+    return NIL;
+  return pair_unwrap(args)->head;
+}
diff --git a/src/prim.h b/src/prim.h
new file mode 100644
index 0000000..093e6e0
--- /dev/null
+++ b/src/prim.h
@@ -0,0 +1,8 @@
+#include <wolflisp.h>
+
+O prim_cons(In *in, O args, O env);
+O prim_head(In *in, O args, O env);
+O prim_tail(In *in, O args, O env);
+O prim_quote(In *in, O args, O env);
+O prim_print(In *in, O args, O env);
+O prim_println(In *in, O args, O env);
diff --git a/src/print.c b/src/print.c
new file mode 100644
index 0000000..418b653
--- /dev/null
+++ b/src/print.c
@@ -0,0 +1,70 @@
+#include <inttypes.h>
+#include <stdio.h>
+
+#include <wolflisp.h>
+
+void print(O obj);
+
+void print_pair(O obj) {
+  O c = obj;
+  I f = 1;
+
+  printf("(");
+  while (c != NIL && !IMM(c)) {
+    Gh *h = UNBOX(c);
+    if (h->type != TYPE_PAIR) {
+      printf(" . ");
+      print(c);
+      printf(")");
+      return;
+    }
+    Pa *p = (Pa *)(h + 1);
+    if (!f)
+      printf(" ");
+    f = 0;
+    print(p->head);
+    c = p->tail;
+  }
+  if (c != NIL) {
+    printf(" . ");
+    print(c);
+  }
+  printf(")");
+}
+
+void print(O obj) {
+  if (obj == NIL) {
+    printf("()");
+  } else if (IMM(obj)) {
+    printf("%" PRIdPTR, ORD(obj));
+  } else {
+    switch (TAG_OF(obj)) {
+    case TAG_SYM: {
+      Sy *s = (Sy *)UNTAG(obj);
+      printf("%.*s", (int)s->len, s->data);
+      break;
+    }
+    case TAG_PRIM: {
+      Pr *p = (Pr *)UNTAG(obj);
+      printf("<#primitive %s>", p->name);
+      break;
+    }
+    default: {
+      Gh *h = UNBOX(obj);
+      switch (h->type) {
+      case TYPE_PAIR:
+        print_pair(obj);
+        break;
+      default:
+        printf("<#obj type=%" PRId32 " @ %p>", h->type, (void *)h);
+        break;
+      }
+    }
+    }
+  }
+}
+
+void println(O obj) {
+  print(obj);
+  putchar('\n');
+}
diff --git a/src/symbol.c b/src/symbol.c
new file mode 100644
index 0000000..67d049c
--- /dev/null
+++ b/src/symbol.c
@@ -0,0 +1,84 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include <wolflisp.h>
+
+#define ALIGN(n) (((n) + 7) & ~7)
+
+static Sy *find(St *tab, U32 hash, Z len) {
+  if (tab->capacity == 0)
+    return NULL;
+
+  Z ix = hash % tab->capacity;
+  for (Z i = 0; i < tab->capacity; i++) {
+    Sy *s = tab->data[ix];
+    if (!s)
+      return NULL;
+    if (s->hash == hash && s->len == len)
+      return s;
+    ix = (ix + 1) % tab->capacity;
+  }
+  return NULL;
+}
+
+static V resize(St *tab) {
+  Z cap = tab->capacity == 0 ? 16 : tab->capacity * 2;
+  Sy **nb = calloc(cap, sizeof(Sy *));
+  for (Z i = 0; i < tab->capacity; i++) {
+    if (tab->data[i]) {
+      Sy *s = tab->data[i];
+      Z ix = s->hash % cap;
+      while (nb[ix])
+        ix = (ix + 1) % cap;
+      nb[ix] = s;
+    }
+  }
+
+  if (tab->data != NULL)
+    free(tab->data);
+
+  tab->capacity = cap;
+  tab->data = nb;
+}
+
+static U32 hashstr(const char *data, Z len) {
+  U32 hash = 2166136261u;
+  for (Z i = 0; i < len; i++) {
+    hash ^= (uint8_t)data[i];
+    hash *= 16777619u;
+  }
+  return hash;
+}
+
+Sy *intern(St *tab, const char *str, Z len) {
+  if (len == 0)
+    len = strlen(str);
+  U32 hash = hashstr(str, len);
+  Sy *sym = find(tab, hash, len);
+  if (sym)
+    return sym;
+
+  sym = aligned_alloc(8, ALIGN(sizeof(Sy)));
+  if (!sym) {
+    fprintf(stderr, "failed to allocate memory for symbol\n");
+    abort();
+  }
+
+  sym->data = malloc(len);
+  memcpy(sym->data, str, len);
+  sym->len = len;
+  sym->hash = hash;
+  if (tab->count + 1 > tab->capacity)
+    resize(tab);
+  Z idx = hash % tab->capacity;
+  while (tab->data[idx] != NULL)
+    idx = (idx + 1) % tab->capacity;
+  tab->data[idx] = sym;
+  tab->count++;
+  return sym;
+}
+
+O symbol_make(In *in, const char *str) {
+  return BOX(TAG(intern(&in->symtab, str, 0), TAG_SYM));
+}
diff --git a/src/type.c b/src/type.c
new file mode 100644
index 0000000..49ab75c
--- /dev/null
+++ b/src/type.c
@@ -0,0 +1,43 @@
+#include <inttypes.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <wolflisp.h>
+
+// clang-format off
+static const char *typenames[] = {
+  [TYPE_NIL] = "nil",
+  [TYPE_NUM] = "number",
+  [TYPE_SYM] = "symbol",
+  [TYPE_PRIM] = "primitive",
+  [TYPE_PAIR] = "pair",
+  [TYPE_CLOS] = "closure",
+};
+// clang-format on
+
+I type(O obj) {
+  if (obj == NIL) {
+    return TYPE_NIL;
+  } else if (IMM(obj)) {
+    return TYPE_NUM;
+  } else {
+    switch (TAG_OF(obj)) {
+    case TAG_SYM:
+      return TYPE_SYM;
+    case TAG_PRIM:
+      return TYPE_PRIM;
+    case TAG_MAN: {
+      Gh *hdr = UNBOX(obj);
+      return hdr->type;
+    }
+    default:
+      fprintf(stderr, "unknown pointer tag %" PRIdPTR "\n", TAG_OF(obj));
+      abort();
+    }
+  }
+}
+
+const char *typename(I t) {
+  if (t >= TYPE__MAX)
+    return "??";
+  return typenames[t];
+}