diff --git a/src/se050_blake2s.c b/src/se050_blake2s.c
index bbce4f5..bc4c54d 100644
--- a/src/se050_blake2s.c
+++ b/src/se050_blake2s.c
@@ -154,7 +154,8 @@ int se050_blake2s_init(se050_blake2s_ctx_t *ctx, size_t outlen)
     }
     
     /* Initialize hash state with IV XORed with parameter block */
-    ctx->h[0] = BLAKE2S_IV[0] ^ 0x01010020 ^ outlen;  /* fanout=1, depth=1, outlen */
+    /* Parameter block: 0x01010000 ^ (0 << 8) ^ outlen = 0x01010000 ^ outlen */
+    ctx->h[0] = BLAKE2S_IV[0] ^ 0x01010000 ^ outlen;
     ctx->h[1] = BLAKE2S_IV[1];
     ctx->h[2] = BLAKE2S_IV[2];
     ctx->h[3] = BLAKE2S_IV[3];
@@ -169,7 +170,6 @@ int se050_blake2s_init(se050_blake2s_ctx_t *ctx, size_t outlen)
     ctx->f[1] = 0;
     ctx->buflen = 0;
     ctx->outlen = outlen;
-    ctx->last_node = 0;
     
     return 0;
 }
@@ -184,7 +184,7 @@ int se050_blake2s_init_key(se050_blake2s_ctx_t *ctx, size_t outlen,
         return -1;
     }
     
-    /* Initialize with key */
+    /* Initialize with key: 0x01010000 ^ (keylen << 8) ^ outlen */
     ctx->h[0] = BLAKE2S_IV[0] ^ 0x01010000 ^ (keylen << 8) ^ outlen;
     ctx->h[1] = BLAKE2S_IV[1];
     ctx->h[2] = BLAKE2S_IV[2];
@@ -194,17 +194,18 @@ int se050_blake2s_init_key(se050_blake2s_ctx_t *ctx, size_t outlen,
     ctx->h[6] = BLAKE2S_IV[6];
     ctx->h[7] = BLAKE2S_IV[7];
     
-    ctx->t[0] = 0;
+    ctx->t[0] = BLAKE2S_BLOCK_SIZE;  /* Key block already processed */
     ctx->t[1] = 0;
     ctx->f[0] = 0;
     ctx->f[1] = 0;
-    ctx->buflen = keylen;
+    ctx->buflen = 0;  /* Key already compressed */
     ctx->outlen = outlen;
-    ctx->last_node = 0;
     
-    /* Pad key to block size */
-    memset(ctx->buf, 0, BLAKE2S_BLOCK_SIZE);
-    memcpy(ctx->buf, key, keylen);
+    /* Compress key block */
+    uint8_t key_block[BLAKE2S_BLOCK_SIZE];
+    memset(key_block, 0, BLAKE2S_BLOCK_SIZE);
+    memcpy(key_block, key, keylen);
+    blake2s_compress(ctx, key_block);
     
     return 0;
 }
@@ -229,25 +230,21 @@ int se050_blake2s_update(se050_blake2s_ctx_t *ctx, const void *data, size_t len)
         p += left;
         len -= left;
         
-        /* Update counter */
+        /* Compress the block */
         ctx->t[0] += BLAKE2S_BLOCK_SIZE;
         if (ctx->t[0] < BLAKE2S_BLOCK_SIZE) {
             ctx->t[1]++;
         }
-        
-        /* Compress */
         blake2s_compress(ctx, ctx->buf);
         ctx->buflen = 0;
     }
     
     /* Process full blocks */
     while (len >= BLAKE2S_BLOCK_SIZE) {
-        /* Update counter */
         ctx->t[0] += BLAKE2S_BLOCK_SIZE;
         if (ctx->t[0] < BLAKE2S_BLOCK_SIZE) {
             ctx->t[1]++;
         }
-        
         blake2s_compress(ctx, p);
         p += BLAKE2S_BLOCK_SIZE;
         len -= BLAKE2S_BLOCK_SIZE;
@@ -268,7 +265,7 @@ int se050_blake2s_final(se050_blake2s_ctx_t *ctx, void *out, size_t outlen)
         return -1;
     }
     
-    /* Update counter with remaining data */
+    /* Update counter */
     ctx->t[0] += ctx->buflen;
     if (ctx->t[0] < ctx->buflen) {
         ctx->t[1]++;