LLM-Powered Search, Type-Safe GraphQL & Docs That Test Themselves

👋 Intro:

Hi! This week I continued reading Hands-On Large Language Models and also finished the new book Docs as Tests: A Strategy for Resilient Technical Documentation. It shares an interesting idea on how to make documentation more stable by combining it with tests. If you're curious, check it out → Docs as Tests – Intro.
I also wrote some blog posts and played further with LLM tooling.

📌 New Posts:

I wrote a blog post where I compared the state-of-the-art models for doing research with OpenAI. Ideally, I’d always use o3 for everything, but I don’t have the $200 subscription. You can check out my findings →
LLM-Powered Search Comparison: O4-mini-high vs O3 Deep Research

I also continued with my series on how to use GraphQL in a Vue project →
Type-Safe GraphQL Queries in Vue 3 with Codegen

Also funny I used o3 to solve a hard chess problem → https://alexop.dev/posts/how-03-model-tries-chess-puzzle/

🔗 3 Links I Loved:

A complete course on how to use Nuxt as a full-stack app – and it's free! I’ll definitely watch it slowly over time. Huge thanks to everyone involved in creating such a gem.

Cursor added an interesting new guide on how to use something like cursor.ai together with tools like Figma and Linear (MCPS) to ship faster.

If you still don’t know how to use tools like Cursor or Copilot to speed up your workflow, Daniel Kelly explains it all in a 3-minute video.

⚡ One-Minute Tip:

One fun thing you can do with Copilot or Cursor while working on a project is use a prompt like:

This surprisingly works well – you get multiple-choice questions that help you check if you understand the project you're working on.

A follow-up idea: if you're using something like VitePress, you can add a small interactive Vue component at the end of a section as a quiz.

That’s everything for this newsletter.
Wishing you all a good week!