Hello friend! It's Daniel here, author of Building AI Agents with LangGraph js and LangChain for JavaScript Developers.
It’s hard to believe we’re already in the last month of 2025. For me, this year flew by: new job, new country, small kid, new apartment. Staying busy really does make time move faster.
To close the year on a high note, next week I’ll be interviewing Chris Coyier (founder of CSS - Tricks and CodePen). I’m incredibly excited about this conversation. Just reply to this email if there’s a question you’d like me to ask Chris.
That being said, let's learn!
💰 TOON (Token-Oriented Object Notation), by Johann Schopplich, is a new data format built to help LLMs use fewer tokens. It stores the same info as JSON but in a much snappier, compact way, saving costs. Check out this example, where 680 chars of JSON were translated into just 286 chars of TOON.
💡 In case you needed extra proof that MCP will be a huge thing, just check out this video by Addy Osmani on how the Google Chrome team implemented an MCP server for the Dev Tools and all the great things it can do.
📺 Speaking of MCP, Paolo Ricciuti, former guest at the JS Craft Podcast and core contributor to Svelte.js, gave a really good talk at the MCP Summit on his TMCP project, a TypeScript-first approach to building Model Context Protocol (MCP) servers without the heavy Node/Express baggage. You can watch his full talk here.
✍️ Saying we don’t need to learn to write because LLMs can do it for you is like saying we don’t need to go to the gym because machines can lift for us. Writing is for the mind what exercise is for the body. Some good ideas here.
😉 Don’t rely on AI for everything! Stick to traditional, “boring” code whenever you can. It’s cheaper and more reliable. Use AI/LLMs only for tasks that truly need them, like natural language processing (NLP).
🎓 Watched this 0 hype, very grounded talk from one of the minds behind RAG on AI agents. My main takeaways: 1. systems, not models; a well-designed RAG system & decent LLM beats a great LLM & poor system and 2. build systems that can explain why they generated a particular answer. What a refresh from "earn 30k/month with this no-code AI agent" bs style videos! Full video here.
🍿 Large language models deeply explained is a great, succinct video explaining how large language models work. If you're still confused about LLMs and haven't had the time to read through academic papers or sit through hours-long lectures, this is a good video to get started.
🎬 Check out this quick video by IMB. It shows you the difference between RAG and fine-tuning, why each matters, and the trade-offs, all in plain English. Long story short: use RAG when you want to add highly dynamic data to the LLM, while fine-tuning changes the model weights so that it's adapted for things such as setting a specific tone.
🎧 This conversation between Tim Ferris and Arnold Schwarzenegger made my day. One more proof that bodybuilders are some of the smartest athletes, not mindless muscle bros.
💪 Speaking of bodybuilders, this podcast with Dorian Yates, who won six consecutive Mr. Olympia titles, is excellent! It seems that when it comes to building muscle mass, the same principles apply as in passive investing: keep it simple, take fewer actions, and put in as much as you can.
📝 This year, I decided to do a yearly review in the style of Tim Ferriss. Hope I will have an article out of it soon.
🇪🇸 It's not always sunny here in the South of Spain, but being able to work out outdoors at the seaside in shorts and a T-shirt in December makes it all worth it. Happiness++!
📕 Of you’re a JavaScript developer and want to level up your AI game, I have made two books that may save you time, energy, and even money: Building AI Agents with LangGraph.js and LangChain for JavaScript Developers.
Well, that’s all for now! Happy Christmas holiday present shopping, and see you next time. Be kind, keep coding, and keep learning!
📖 Build a full trivia game app with LangChain
Learn by doing with this FREE ebook! This 35-page guide walks you through every step of building your first fully functional AI-powered app using JavaScript and LangChain.js
📖 Build a full trivia game app with LangChain
Learn by doing with this FREE ebook! This 35-page guide walks you through every step of building your first fully functional AI-powered app using JavaScript and LangChain.js