[JS-Craft Newsletter #57] My interview with Josef Strzibny, orchestrating AI Agents with JavaScript code and LangGraph.js

DanielSeptember 5, 2024February 26, 2025

Hey! It’s Daniel here, the author of 📘 LangChain for JavaScript Developers and the guy behind js-craft.io.

For this edition of the Js-Craft newsletter, I’ve decided to try a simpler format. This idea was inspired by Michael Karnjanaprakorn's newsletter, the founder of Skillshare.

I hope that by using this new format, I can make the newsletter feel more personal. Plus, it will require less editing, so I should be able to send it twice per month. Let me know if you like this format—there’s a poll at the end of the newsletter.

Here we go:

1. I've recorded a new podcast episode with Josef Strzibny. We talked about how he wrote his two books, Deployment From Scratch and Kamal Handbook, which generated over $50k in sales. Be sure to check also this talk of Josef at the Balkan Ruby conference.
   

2. Speaking of the Js-Craft podcast, in case you missed it, the first episode featured Jacob Lee, lead maintainer of LangChain.js. Highly recommend it!

3. This week, I re-read The Pragmatic Programmer. Great book! In chapter 1.5, the technologies a software developer knows are seen as stocks in an investment portfolio. Each new technology we learn adds potential value to our portfolio. My current investment is LangChain.

4. In any job or industry, it's important to look up to those ahead of you and ask yourself if you admire them. If you work hard and do well, you’ll eventually find yourself among them—if not becoming like them.

5. I’ve posted a poll on Twitter about which LLMs and LangChain-related topics you’d like to learn more about. Curious to hear your thoughts:

   • How to build an LLM from scratch with JavaScript
   • AI agents and LangGraph.js
   • Evals for LLMs - LangSmith.js
   • Advanced RAG

6. I learned that we can view AI Agents as LLMs running in a do-while loop until they reach a given objective. This interview with Harrison Chase, the founder of LangChain, has fantastic insights. By the way, one of the example apps we build in the LangChain for JavaScript Developers book focuses on using AI Agents.

7. I’ve started experimenting with LangGraph.js. I’ll post more about this soon, but in the meantime, you can check out this short clip showing how LangChain, LangSmith, and LangGraph work together.

8. Maybe it’s just my limited experience, but I’ve noticed that the less technical someone is, the more likely they are to claim they understand AI and LLMs. It’s crucial to educate ourselves on how AI models actually work so we can call BS when needed.

9. On the topic of understanding LLMs, check out this Decoding the Decoder LLM talk by Ishan Anand from this year’s AI Engineer World's Fair.

10. By the way, I’ll have Ishan as a podcast guest next week! Feel free to reply to this email with any questions you’d like me to ask him.

11. Keep in mind that LLMs mimic the user’s style. If you misspell or use strange casing in your prompt, you increase the chances of unwanted results!

12. The official LangChain Twitter account shared an article I wrote on Routing LLM prompts with JavaScript. Made my day!

13. I don’t think there’s an AI bubble. LLMs and AI are incredible! The "bubble" is caused by people who don’t understand how they work. Learn the mechanisms, recognize real use cases, and be prepared to call out BS when someone invokes AI as a magic solution.

14. Last week, I spent some time with my wife and daughter at the seaside on a "workation." I was lucky enough to have this view from my desk:

    

How do you like the current format?

• 🙂 It's nice, keep it
• 🙁 The older format was better, revert back

That’s all for this edition, folks! Be kind, be smart, and keep coding!