How I'd Learn AI Agents FAST if I Had to Start Over (Full Roadmap)

Thank you for this video, powerful concept simplified.

I would suggest using nocodb as "local airtable" to manage personal/sensitive data (user memory) or other airtable related use cases.

It's a no-brainer (.exe installer+application) to setup your own local airtable in windows os with rest API, webhook automation, mcp compatible and no api rate limits (your pc spec is your bottlenecks).

i’ve been really enjoying, and applying your videos. thank you!

this blueprint is so good that I'm crying for pydantic_ai blueprint too..

If it requires me to pay for xyz service then I'm not giving you 28 minutes. Either local way or the highway.

AI agents are code under the hood and people been integrating AI agents in business operations before no code developed their node. Am I wrong? Graphs just make things easier to visualize.

great video please do more that you could actually sell to companies please

Spicy Mango - oh my god, we've invented Talkie Toaster

TOASTER: Howdy doodly do!  How's it going?  I'm Talkie -- Talkie Toaster,
  your chirpy breakfast companion.  Talkie's the name, toasting's the
  game.  Anyone like any toast?

LISTER: Look, I don't want any toast, and he (indicating KRYTEN) doesn't want any toast.  In fact, no one around here wants any toast. Not now, not ever. NO TOAST.

TOASTER: How 'bout a muffin?

Bro do you think is posible to use a gpt-4o-mini with agent with multiple tools ? i cant make it works....

great breakdown Cole. Now I want some spicy mango salsa.

It's openai caching most probably to save some computing cost. We've had VERY similar answers once in different accounts to the same question

Awesome content, as usual.

Have you tried PocketFlow?

@ColeMedin your sponsors are the most relevant advertisements I have ever seen in any marketing channel. They are incredible. But … because they are so good, I find the context switching incredibly disturbing. Maybe try and put the ad at the end of the video if the sponsor allows you, I would still watch because they are so good!

Great videos. I still need a more dumbed down version, with click along examples.

Really interesting video. Thank you! Now off to cook some spicy mango shrimp!

Over the years?

How is that possible when AI agents really began getting decent in 2025

Hi @ColeMedin, You just fried my Brain with one Sentence: "Agents are Graphs under the hood"...
Can you elaborate on that?

He’s almost there, citing Neuro linguistic programming techniques in the way humans communicate & retain.. it’s called “ chunking “. I know nothing about coding but I know behaviour & communication- there are in fact only 3 filters where communication outputs .. just like humans 😊. Once coding no longer needs coders, that’s when next level skill sets take over .. then all these videos will be dominated by strategy modelling.. in the same way I can install a strategy into some body else using NLP

Every time I come here and see this channel only have 122k subs, is the reason why I felt like we have yet to step into an AI boom yet, not enough people are paying attention to this type of tech and yet it will be one of the most impactful tech to ever hit our lives in the next few years. MCP + A2A really started the game changing process, similar to how web 2.0 kicked off really because the establishment of REST API spec.

🌶️🥭😂

Damn how fast and quickly your Channel grew this must be a record 😂🎉

Spicy mangos will bedefinitely your fate Cole 😁

Your content is great, bro. Question I have for you - why is the graph structure more useful than something like agno or similar frameworks? It's hard to imagine a use case where the added complexity is necessary, but I might just be lacking imagination

Cole -- Great content; I'm enjoying your videos and getting myself up to speed. Do you have an opinion and/or framework for multi-tenant AI agent workflows? Thanks.

Great breakdown Cole. BTW I think we're overdue for a Spicy Mango MCP server.

Just cant get excited for this kinda stuff... I truly believe it will all be built in and just work with the core LLM's with a single prompt before we know it.

What is the difference between an app and an agent? Is it complexity?

This really helped me conceptualize the workflow I wanted into the nodes. I knew what I wanted and could explain it but didn't know how to translate that into the specific nodes and sub nodes. Now I have a better understanding of how to tackle it, thanks!

training does not have enough data or your prompt are just in the same category, likely not enough training

Mr. Cole, you are one of the most highly recommended channels for AI automation. Keep up the great work.

Can you create another version based on adk?

Great setup!
Couple comments.
Why not make guardrails a part of the prompt? Those can be an instruction too, like do a chain of thoughts, make a plan, critique plan, execute, evaluate result against initial requirements.

And a tricky one.
N8n agent tends to eat away errors, both from model and from mcp server. Openai has rate limits on api calls, and if you happen to get one in the middle of agent run it's just gone. Similar happens with mcp calls failing, this one just silently continues to run even after the failure.
How do you handle these?

Also, n8n agent doesn't seem to work well with binary data, like downloading a file from Google drive as a tool.

Also, how do you track and monitor such agents? Langfuse doesn't work with n8n yet, what does?

can you cover a video for a tool named prompthancer its like making prompts another level such as when i feed the prompt into windsurf after making a component (asks befor continuing) which stops the AI from gettin confused

I've never been able to get an auto fixing output parser running in n8n - I thought it always had some kind of bug that I can't recall rn. Feel free to make a video about this. I guess a few people will appreciate it.

You have helped me move into a brand new career field, thank you for your simplicity

Great hob on the video, Cole! Packed full of very useful information! Jay

Who's the spicy mango company sponsoring that LLM?

I am curious what your reason is to use n8n instead of Kestra? Is Kestra missing something that you need?

Bro you are a treasure!

Thanks for the great work!!

I need more !!!

We want more spicey mango chicken

Always good stuff man

keep up the good work