Introduction
Suddenly, the algorithm started pushing no code AI agent videos into my YouTube feed. I had been looking at AI at a very high level for the past few years and have used Chat GTP and some of the art generation AI tools. I found them useful for research, but I did not want ChatGPT representing me to the real world. I figured out a way to use it for basic definitions and explanation of concepts, but I did all the interpretive writing myself.
However, the algorithm began presenting me videos on no code AI autonomous agents. These agents were supposed to remove the drudgery from my life and handle all my daily repetitive and unfulfilling tasks . The promise is that I can accomplish this without any coding skills.
Therefore, I decided to embark on the journey of creating an autonomous AI agent that would upload my art to my various websites, create a caption for the art and write a brief description of the art. I use Procreate to create a series of drawings that I would eventually like to see people turn into T-shirts. However, I need to perform the arduous work of putting them on my various websites.
Purpose
This is a series of blogs that chronicles my progress in understanding, autonomous artificial intelligent agents. I want to be clear that I have not figured this out. The whole purpose of these blogs is to document my process. I will guarantee you that there will be hidden surprises and unintended consequences that I will uncover. I fully expect to bang my head up against the wall on multiple occasions while trying to figure this out. Therefore, if something I write seems a bit goofy, it is probably because it is goofy. Please bear with me while I figure this out.
No Code Beginnings
The algorithm sent me a variety of different platforms that claimed they could do no code agents. These included Bolt, Claude, Relevance, Google, and Microsoft. However, there are many other no code platforms that you can choose to experiment with. I decided to start with Claude 3.5 sonnet.
It all starts with a prompt. This is the prompt that I fed Claude to begin the process.
“I would like to create an AI agent that will go to a folder of art, give each piece of art title, or, write out brief description of each piece of art, and then upload the art to my website”
This set off a series of questions and responses. However, the result was a series of code snippets. This code needed to be implemented and integrated into a variety of APIs to make the project work. I have attached the iterations. Claude created the code for me, but I had no idea what it said or how it worked. I discovered that I had to create an infrastructure for hosting the AI agent. Claude was not going to do it for me. Therefore, I created another chat about agent hosting infrastructure requirements. I asked Claude to explain agent hosting infrastructure requirements. The next blog “Claude Speaks I” details in the entire conversation
Prompt
I drew this very basic diagram. Please let me know if it is an accurate representation of the basic architecture of the autonomous, artificial intelligent agent
Claude returns the following;
Yes, this is a good basic representation of an AI agent's architecture! Let me break down the key components you've illustrated:
1. AWS Environment (outer box):
- Correct as it represents the cloud hosting platform where your agent would run
- AWS is a good choice as it provides all necessary services for hosting AI agents
2. AI Agent Software (circle):
- Correctly positioned as the core component
- This would contain the agent's logic, decision-making capabilities, and processing functions
3. Hosted Compute and Data Storage (bottom rectangle):
- Accurately shown as the foundation
- This could represent AWS services like:
- EC2 for compute
- S3 for data storage
- RDS for structured data
- DynamoDB for fast access to dynamic data
4. API (connection to external):
- Correctly shown as the interface between your agent and external resources
- This represents how your agent would communicate with external services
5. External Applications and Resources (right box):
- Accurately represents external services your agent might interact with
- Could include APIs for various services, websites, databases, etc.
The diagram captures the essential flow: the agent runs on cloud infrastructure, processes information, and interacts with external resources through APIs.
To make it even more complete, you might consider adding:
1. Security layers (IAM, VPC)
2. Load balancing if scaling is needed
3. Monitoring and logging components
4. Database connections
5. User interface or interaction points
But for a basic architectural overview, your diagram effectively communicates the core concepts!
Obviously, this is not as simple as entering a prompt and Claude creating a functional agent.
One of the main conclusions I derived from this exercise is that it’s not as simple as YouTube AI hustlers would like you to think. Stay tuned for the next episode in my autonomous AI agent quest.
