diff --git a/lib/aos/coreboot.c b/lib/aos/coreboot.c
index 7a65772..a473cf5 100644
--- a/lib/aos/coreboot.c
+++ b/lib/aos/coreboot.c
@@ -211,10 +211,6 @@ errval_t coreboot(coreid_t mpid,
     err = invoke_kcb_identify(kcb_cap, &kcb_id);
     if(err_is_fail(err)) return err_push(err, LIB_ERR_FRAME_IDENTIFY);
 
-    // TODO rueegges: does freeing the capability slot work?
-    // err = cap_delete(kcb_ram_cap);
-    // if (err_is_fail(err)) return err_push(err, LIB_ERR_CAP_DELETE);
-
     // - Get and load the CPU and boot driver binary.
     struct mem_region * bootdriver_module = multiboot_find_module(bi, boot_driver);
     if (bootdriver_module == NULL) return LIB_ERR_COREBOOT_FIND_MODULE;
diff --git a/lib/aos/ump_chan.c b/lib/aos/ump_chan.c
index c9fd08f..9668757 100644
--- a/lib/aos/ump_chan.c
+++ b/lib/aos/ump_chan.c
@@ -98,7 +98,6 @@ static bool ump_ring_try_read_next(struct ump_ring_state *ring_state, void *buf,
  * @param recv_chan channel to read the message on
  */
 static void ump_try_read(struct ump_recv_chan *recv_chan) {
-    // TODO rueegges: is this locking required?!? if writing the state is a single assembly instruction then not, otherwise yes
     // make sure we own the read metadata, otherwise we have nothing to do anyways
     thread_mutex_lock(&recv_chan->recv_register_lock);
     if(
@@ -168,7 +167,6 @@ static void ump_try_read(struct ump_recv_chan *recv_chan) {
  * @param send_chan channel to send the message on
  */
 static void ump_try_send(struct ump_send_chan *send_chan) {
-    // TODO rueegges: is this locking required?!? if writing the head is a single assembly instruction then not, otherwise yes
     // check if there is anything to do.
     thread_mutex_lock(&send_chan->send_queue_lock);
     if (send_chan->send_queue_head == NULL) {
@@ -446,7 +444,7 @@ void ump_recv_header (
     assert(chan != NULL);
 
     thread_mutex_lock(&chan->recv_register_lock);
-    // TODO rueegges: caller has to ensure it only registers at most once?! Or should we just make it a noop?
+    // NOTE rueegges: caller has to ensure it only registers at most once?! Or should we just make it a noop?
     assert(chan->state == UMP_RECV_STATE_IDLE);
 
     chan->recv_header_callback = callback;
@@ -467,7 +465,7 @@ void ump_recv_payload (
     assert(chan != NULL);
     
     thread_mutex_lock(&chan->recv_register_lock);
-    // TODO rueegges: caller has to ensure it only registers at most once?! Or should we just make it a noop?
+    // NOTE rueegges: caller has to ensure it only registers at most once?! Or should we just make it a noop?
     assert(chan->state == UMP_RECV_STATE_PAYLOAD_IDLE);
 
     chan->recv_payload_callback = callback;
diff --git a/usr/init/main.c b/usr/init/main.c
index 465dcc4..fc00ff1 100644
--- a/usr/init/main.c
+++ b/usr/init/main.c
@@ -125,7 +125,6 @@ bsp_main(int argc, char *argv[]) {
     );
     cpu_idcache_wbinv_range((uintptr_t)urpc_addr, MON_URPC_SIZE);
 
-    // TODO rueegges: can we get the mpid in a better way?
     err = coreboot(1, "boot_armv8_generic", cpu_driver_name, "init", urpc_frame_id);
     if (err_is_fail(err)) return err_push(err, MON_ERR_SPAWN_CORE);
     // err = coreboot(2, "boot_armv8_generic", cpu_driver_name, "init", urpc_frame_id);