diff --git a/Makefile b/Makefile
index 10966d1..89f18d5 100644
--- a/Makefile
+++ b/Makefile
@@ -2,7 +2,8 @@ TARGET = foxdos
 
 FILES = config.s \
 		src/int21.s \
-		src/kernel.s
+		src/kernel.s \
+		src/vga.s
 
 .PHONY: all qemu clean
 all: obj/boot.o obj/kernel.o $(TARGET)
diff --git a/config.s b/config.s
index 5d830b2..511d33f 100644
--- a/config.s
+++ b/config.s
@@ -1,4 +1,4 @@
-NUMSEG equ 0 ; TODO build system should calculate this value
+NUMSEG equ 1 ; TODO build system should calculate this value
 
 ; might be good to move one of these into the high memory area
 K_ADDR    equ 0x500  ; kernel load address
diff --git a/src/boot.s b/src/boot.s
index ef1b0ee..eaac5a6 100644
--- a/src/boot.s
+++ b/src/boot.s
@@ -1,4 +1,4 @@
-; mothdos boot sector. should load kernel into segment 0x50 at 0x00. max size
+; foxdos boot sector. should load kernel into segment 0x50 at 0x00. max size
 ; would be 64k but i doubt a dos needs more than that. stack is placed at
 ; 0x10500. i would suggest changing this to load the kernel into extended
 ; a20-gate memory so as not to overwrite the boot sector with larger kernels
diff --git a/src/kernel.s b/src/kernel.s
index 405553f..ebf0418 100644
--- a/src/kernel.s
+++ b/src/kernel.s
@@ -1,6 +1,12 @@
 %include "config.s"
-[org K_ADDR]
+
 kernel_entry: ; MUST be a short jump due to loader config
-	jmp $ ; TODO stub because os doesnt actually exist yet
+	mov si, hello_string
+	mov bl, 0x70
+	call print_string
+	jmp $
+
+hello_string: db "hello world!", 13, 10, '$'
 
 %include "int21.s"
+%include "vga.s"
diff --git a/src/vga.s b/src/vga.s
new file mode 100644
index 0000000..fbf1559
--- /dev/null
+++ b/src/vga.s
@@ -0,0 +1,46 @@
+; print a string to the screen
+; inputs:
+; DS:SI: pointer to '$'-terminated string
+; BL: attribute
+; outputs:
+; none
+print_string:
+	push ax
+	push bx
+	push cx
+	push dx
+	push bp
+	push es
+
+	; get cursor position in DX
+	mov ah, 0x03
+	mov bh, 0x00
+	int 0x10
+
+	; calculate string length by iterating over it until we reach '$'
+	; this sucks but we have to do it because the bios expects to be passed the string length directly
+	xor cx, cx
+	push si
+.size_loop:
+	inc cx
+	inc si
+	cmp byte [si], '$'
+	jnz .size_loop
+	pop si
+	mov bp, si
+
+	; print string and update cursor position
+	mov ax, ds
+	mov es, ax
+	mov ah, 0x13
+	mov al, 0x01
+	mov bh, 0x00
+	int 0x10
+
+	pop es
+	pop bp
+	pop dx
+	pop cx
+	pop bx
+	pop ax
+	ret