I’m thinking of building an agent team platform based for consumers and enterprises for various tasks automation via mobile apps, browser and desktop apps.
Do the tasks in crewai have to be predefined? I wanted to try this framework but for my usecase I need agents to be able to figure things out during runtime meaning I need on the fly dynamic tasks. Seems like crewai follows a predefined workflow?
I really struggle on Cursor, can you please do a video in Replit? I am not a developer by trade and could not even get the 1st step working :( the pip install.
To build a great project It’s not gonna take 10 min vid or “5 easy steps” It’s gonna take months of work That is if u actually have like 5-10 yrs of experience.
I've worked in IT for 40 years so know my way around a lot of technologies. I like to watch these videos to get an idea of what can be done, but you're right - there's so much going on in the background here that is taken for granted. When I hear the presenter say "oh, you just need to add this" without showing it, I know there's going to be another video I need to watch. I've managed to set up ollama and Open Web-UI, and Stable Diffusion on my macbook, but it took a fair bit of bouncing around different videos and tutorials. But I'm grateful to all these guys like Tyler who are taking the time and effort to explain this stuff and to make it accessible.
@@andyhorne9747 I am loving the AI revolution really. If this was up to people like you who got a job while it was as easy as going for lunch, you wouldn't teach shit. Boomers think someone took their flesh out of them when they teach something. Go away mate. We are the future.
So, I have been in the software business for 30+ years. I say that to illustrate that I have seen a lot of change. I’d like to see a broad conversation on WHEN to use deterministic techniques and when to use probabilistic techniques. I just cannot buy the idea that AI tech is a magic silver bullet, and its the end of the programming world as we know it. Yes, there is a place for AI and probabilistic logic, but is it really for EVERY use case as many are suggesting?
There are generative AI, and there are predictive AI which have been around much longer. Back in 2016, predictive AIs with narrow use cases for object detection via computer vision were already hitting 90%+ accuracy. Consistency and reliability comes down to the architecture here. You are correct that generative models are over hyped, but, they are still interesting.
I created and deployed a 3,800 LoC project from scratch in under 1 week (didn’t write a single line, but actively architected the 3800loc via ~300 messages to AI). It was for a client who was ecstatic stating that I delivered above his expectations That was in November tho
Well, AI has almost all the knowledge that humans combined have. It's not limited by iq as well as not being limited by energy as us humans. Hence, i think it can learn most non physical tasks. What tasks do you think it couldnt handle?
@ What I am saying is, that runtime performance and cost will be faster and cheaper for applications when engineers use the the architecture that uses the least resources and is the most predictable (quality). Yes, AI is a great new tool, but its a part of a toolbox. It will be OVERUSED for a while and folks will build very expensive runtime software til they figure out what I am saying. The new AI coding assistants will certainly allow for more development on the “efficient frontier” using known techniques.
Join the Skool Community here:
www.skool.com/the-ai-agent-5174/about
I’m here to help you create AI Agents. They are here and here to stay 💪
your skool rocks !!! 🦾🤘
If you're reading this, you're already ahead of most people in the AI race
Why? There is no other better ones?
Thats awesome! Love the use of Firecrawl! 🔥
I’m thinking of building an agent team platform based for consumers and enterprises for various tasks automation via mobile apps, browser and desktop apps.
Thank you Tyler - Does any CLI exist like this for any TypeScript tools?
Very nice example and I am loving AgentStack. With firecrawl tool, how can we structure the inputs/task to take a deterministic website?
I’m really interested in learning more about Ai Agents
I am confused, what repo is the code in?
Do the tasks in crewai have to be predefined? I wanted to try this framework but for my usecase I need agents to be able to figure things out during runtime meaning I need on the fly dynamic tasks.
Seems like crewai follows a predefined workflow?
I really struggle on Cursor, can you please do a video in Replit? I am not a developer by trade and could not even get the 1st step working :( the pip install.
Hey mate, is it possible to create a real estate hunting multi agents?
where to import crewai?
How much would it cost ro hire someone to create something like this?
This was just a variable version of the how to on the website 😂
Phenomenal
To build a great project
It’s not gonna take 10 min vid or “5 easy steps”
It’s gonna take months of work
That is if u actually have like 5-10 yrs of experience.
I've worked in IT for 40 years so know my way around a lot of technologies. I like to watch these videos to get an idea of what can be done, but you're right - there's so much going on in the background here that is taken for granted. When I hear the presenter say "oh, you just need to add this" without showing it, I know there's going to be another video I need to watch.
I've managed to set up ollama and Open Web-UI, and Stable Diffusion on my macbook, but it took a fair bit of bouncing around different videos and tutorials.
But I'm grateful to all these guys like Tyler who are taking the time and effort to explain this stuff and to make it accessible.
@ damn 40 years?
U must be 60 now
@@andyhorne9747 I am loving the AI revolution really. If this was up to people like you who got a job while it was as easy as going for lunch, you wouldn't teach shit. Boomers think someone took their flesh out of them when they teach something. Go away mate. We are the future.
So, I have been in the software business for 30+ years. I say that to illustrate that I have seen a lot of change. I’d like to see a broad conversation on WHEN to use deterministic techniques and when to use probabilistic techniques. I just cannot buy the idea that AI tech is a magic silver bullet, and its the end of the programming world as we know it. Yes, there is a place for AI and probabilistic logic, but is it really for EVERY use case as many are suggesting?
There are generative AI, and there are predictive AI which have been around much longer. Back in 2016, predictive AIs with narrow use cases for object detection via computer vision were already hitting 90%+ accuracy.
Consistency and reliability comes down to the architecture here. You are correct that generative models are over hyped, but, they are still interesting.
I created and deployed a 3,800 LoC project from scratch in under 1 week (didn’t write a single line, but actively architected the 3800loc via ~300 messages to AI). It was for a client who was ecstatic stating that I delivered above his expectations
That was in November tho
Well of course you should use deterministic every time possible for cost and peace of mind.
Well, AI has almost all the knowledge that humans combined have. It's not limited by iq as well as not being limited by energy as us humans. Hence, i think it can learn most non physical tasks. What tasks do you think it couldnt handle?
@ What I am saying is, that runtime performance and cost will be faster and cheaper for applications when engineers use the the architecture that uses the least resources and is the most predictable (quality). Yes, AI is a great new tool, but its a part of a toolbox. It will be OVERUSED for a while and folks will build very expensive runtime software til they figure out what I am saying. The new AI coding assistants will certainly allow for more development on the “efficient frontier” using known techniques.
dont start technical right away, show the result first