Llama: update history

package.json

@@ -25,6 +25,7 @@
     "remark-rehype": "^11.0.0",
     "sharp": "^0.33.1",
     "shiki": "^0.14.7",
+    "shikiji": "^0.10.2",
     "tailwindcss": "^3.4.0"
   },
   "devDependencies": {

src/content/llama/a-history-of-llamas.md

@@ -191,3 +191,13 @@ how much impact they have in a retrospective:
    to user evaluation, it beats Mixtral for all uses.
  - **2024-04-18**: Meta releases LLaMA3 8b and 70b. 70b is the new best open
    model, right around Claude3 Sonnet and above older gpt4 versions!
+ - **2024-05-13**: OpenAI releases gpt-4o. This multimodal model is able to take
+   in different forms of input and output, like speech and text, no model-chain
+   required. It also beats all other models comfortably
+ - **2024-05-28**: Gemini-pro-1.5 is second only to gpt-4o, all other gpt-4
+   models are ranked lower, but not for english
+ - **2024-05-29**: Mistral releases codestral, a 22B closed model. User reviews
+   have put it above gpt4o and opus for coding (me too)
+ - **2024-06-20**: Anthropic releases Sonnet 3.5. Despite being the medium
+   model, it's ranked higher for coding on lmsys than gpt-4o and user reviews
+   vastly prefer its voice

src/content/unix/networking-from-scratch.md

@@ -6,10 +6,10 @@ updateDate: 'Apr 22 2024'
 
 # Why learn Low-Level Networking?
 
-Networking is one of the most complicated and practical standards for ever
-created for digital communication. This unfortunately means that when we connect
-to a network, it's quite hard to understand what's actually happening and
-knowing this is very helpful for debugging.
+Networking is one of the most complicated and practical standards ever created
+for digital communication. This unfortunately means that when we connect to a
+network, it's quite hard to understand what's actually happening, making it hard
+to pick up network debugging skills.
 
 In this article we'll go through setting up a simple wired network using the
 `ip` tool on Linux. We'll also do a quick overview of next steps.