diff --git a/compiler/lexer.py b/compiler/lexer.py
index 55c67cc..f8060c6 100644
--- a/compiler/lexer.py
+++ b/compiler/lexer.py
@@ -4,10 +4,12 @@ import collections.abc
 import enum
 import re
 from dataclasses import dataclass, field
+from typing import cast
 
 from .logger import Logger
 from .source import SourceLocation, Location
 from .typechecking import typecheck
+from .utils import implies
 
 logger = Logger(__name__)
 
@@ -34,9 +36,15 @@ class Tokens(enum.Enum):
     Parens_Right = re.compile(r"\)")
     Brace_Left = re.compile(r"\{")
     Brace_Right = re.compile(r"}")
+
+    KwLet = re.compile(r"\blet\b")
+
     Identifier = re.compile(r"[a-zA-Z_][a-zA-Z_0-9]*")
+
     Equal = re.compile(r"=")
+    Colon = re.compile(r":")
     Semicolon = re.compile(r";")
+
     Newline = re.compile(r"\n", flags=re.MULTILINE)
     EOF = re.compile(r"\Z")
     Blank = re.compile(r"[ \t]+")
@@ -45,6 +53,9 @@ class Tokens(enum.Enum):
     def __bool__(self):
         return True
 
+    def is_keyword(self) -> bool:
+        return self in [Tokens.KwLet]
+
 
 class Lexer(collections.abc.Sequence):
     def __init__(self, data: str):
@@ -81,15 +92,20 @@ class Lexer(collections.abc.Sequence):
                                            source=self.data),
                                        value=""
                                        )
+            token_kind: Tokens
             for token_kind in list(Tokens):
                 if token_kind == Tokens.Unknown:
                     continue
-                regex: re.Pattern = token_kind.value
+                regex = cast(re.Pattern, token_kind.value)
                 match = regex.match(self.data, self.begin)
                 if match is not None:
                     logger.trace(f"Got match: {match}")
                     result = match.group(0)
                     if best_result.kind == Tokens.Unknown or len(result) >= len(best_result.value):
+                        if not implies(best_result.kind.is_keyword(), token_kind.is_keyword()):
+                            logger.trace(
+                                f"Best match is a keyword ({best_result}) and current match ({token_kind}) is not, skipping")
+                            continue
                         loc = SourceLocation(
                             begin=Location(line=self.line, character=self.character),
                             end=Location(line=self.line, character=self.character + len(result))