How to Build Effective AI Agents (without the hype)

47 thoughts on “How to Build Effective AI Agents (without the hype)”

1. Watch part 2 here, where you'll learn how to code these patterns in pure Python: https://youtu.be/bZzyPscbtI8

2. Thanks!

3. Thanks for that. great video. i am trying to learn all this and playing catch up. Wish i had done sooner. so when you seek specific answers, all you get is catchy Breaking News style headlines to get your attention.
   or it wil be a title that is super engaging but not the matter. you are the 2nd person who seems to be quite genuine. well, atleast you said that using your ip would cost. the other one didnt say that and i wondered whats the catch lol. its a business I get it! for nopw, soaking in as much knowledge as fast as I can! thx. sorry about the typos, My correct my spelling agent is too autonomous and not working today! lol

4. It is a bit too abstract to me. I am in minute 11:24 and you are still talking about prompt engineering, you have not mentioned embeddings or maybe feeding your own LLM is not the goal here. I am off. Interesting though.

5. Thanks for this tutorial! 👍🙏

6. @daveebbekar, around 7:46, you said whenever LLM needs it, you can try to retrieve it. In that context, you said as the scale of db increases, you never know what you get from RAG. Even on a personal side, if I have, let's say, 100GB documents, do I have this issue with RAG? What's the best or most accurate (less hallucinated) way of getting relevant information about my document?

7. Top explanation! Thanks buddy!

8. This will change in the future right? I imagine a world in which hallucinations wont be a thing and ai will be fully integrated. How long do you think it will take for that to happen?

9. Amen! Everyone around is AI this AI that, it's JUST an automation. Thank you for this video!

10. HAHAHA THANK YOU for telling the truth. Sometimes its really frustrating the stuff that creators say about AI agents talking about their automations as if they're AI agents

11. thnq

12. Thanks really

13. Misleading title, this is only for developers.

14. I want to build AI Agent, but I don't know Python.. is Python must to build an AI agent ?

15. Nice try keep trying

16. This is great, I have a concept and need your to help me implement. I am no engineer but a visionary !

17. Nice explaination Dave, learned a lot

18. My first YouTube comment after 20 years. You can imagine how much I loved this video. Super Thanks!

19. Thanks!🙏

20. A very interesting video.

21. Thanks for being honest

22. Great overview!

23. I gave up at 7:23. Lots of high-level talking about what you can do and not 1 single example to help ground theory into practice. 😢

24. Awesome video! I love to hear from similar minded devs. Start simple and build up only as needed. I see code bases that are wayyyy more complicated than they need to be.
    I also love the recommendation to build without the frameworks.

25. This is great intro into the topic, thank you very much!

26. thank you for just saying upfront: this is really hard.

27. Thank you brother

28. I love that you open up by defining the key terms that you're about to discuss.
    This is always very much appreciated as it prevents discussions from devolving into argument due to semantic breakdown, or presentations leaving the receiver confused about the information.
    Something the world needs more of.

29. WOW! Incredibly sharp deep dive in a short amount of time! Super helpful and exactly what I'm looking for. Really appreciate it. I'll check out the other videos, too.

30. About that hype stuff: Let's be honest. People these days get excited quickly. Be new models, new tools, new platforms – it's mind boggling to me how quickly are becoming huge fans of anything and anyone and also start throwing money and things quickly. That's probably the reason while Youtube is getting filled by overhyped content so much and every single of those "mainstream AI blokes" as I would call them basically sells some random course or "Skool" community. Most of them don't even care about creating a proper course anymore – they just hype up some promises, ramp up weird Fomo tactics to get people paying for their promise and that's it. The knowledge level mostly is just insane. Every other n8n course maybe is on a level that some trial and error and some documentation reading could give you in 2 weeks max. But why not pay 79$ per month for a skool community where most of the people just keep asking the same low level questions again every day.

    Sorry for the rant, but I got triggered by your intro because I'm just thinking the same while diving into the agent topic a lot lately. On the flipside it's tough to find any solid community to discuss this things on a deeper level. Discuss memory difficulties for example – tought to find solutions for and could really profit from some solid knowledge exchange.

    Might need to build an Internet filter next that gets rid of all the hype noise to finally hear the signal again 😉

31. This was one of the clearest breakdowns I’ve seen on AI agent design. The section on feedback loops and routing really clicked. I’ve seen platforms like Trillet apply some of these patterns effectively on the voice side.

32. wowowow!!!

33. Awesome video! What tools do you use to design your architecture/software to discuss with engineers/clients?

34. Orange who?

35. BeLikeNative helps me write clear, hype-free content when building and explaining AI agents. I trust BeLikeNative to refine my drafts, fix grammar, and save time on technical writing.

36. 1. Uncertainty has a cost in terms of risks and you want to minimize it.
    2. Solution of a problem is more important than the tools used to solve it.
    3. Scaling matters, cherry picked demo may not scale or generalize.
    4. Complexity has a cost which in most cases is paid in terms of transparency.
    5. Hype can help to win once but not always
    6. Good old fashion screw-driver may be better than Swiss army knife if you just need to unscrew.

37. Thank you so much for your videos. Planning to watch and use in pratice your videos. Do you mind sharing any websites were I can learn Python quicker so I can work on creation of my own AI systems? I have an idea for AI and I feel like I need to work on that. Would appreaciate your help 🥺

38. 1. what are you going to use to build AI system? [ coding / workflow coding[
    2. Building block: the augmented LLM [In–> LLM –> out]
    3. Workflow: prompt chaining
    4. workflow: Routing [if /cases A, we call B]
    5. workflow: parallelization [independent of each other async –> same output]
    6.workflow: Orchestrator-workers
    7.workflow: Evaluator-optimizer

39. Mann I love u 😂 nice job! Im from Brazil, will bring these lessons to my people 😊❤

40. Excellent video and explanation!!! Thanks for sharing this.

41. I think I will just hire a human. They are expensive, but they have real intelligence, which AI does not. Computers are REALLY STUPID. (You can call it AI, but it is still just computer programming.)

42. You became too technical for me to know how to use AI.

43. I am having a bit of difficulty in distinguishing the orchestrator and the router. Could you please provide more examples or any links to read?

44. Priceless info.. thank you!

45. Really helpful content, thanks for sharing

46. Thank you.

47. Finally someone cutting through the noise this is the kind of practical agent building we need I’ve been using Genum AI to manage prompt flows and it’s helped a ton with reliability over flash anyone else keeping it simple and solid like that?

Comments are closed.