AI Community Newsletter 12_5_2025
Caleb Bryant
This email is the full newsletter. For a printable copy, click here to download the PDF.
Welcome to the first "Under the Hood," where I document my journey building an AI-powered newsletter, as well as other AI applications. This week, I'm diving into the initial steps of creating an AI Agent to automate the newsletter's generation, blending creativity with automation. I'll also make sure to define industry terminology, as a learning environment for us all. As an AI newbie, armed with only casual chats with tools like ChatGPT, Gemini, Perplexity, and Claude, I started by experimenting with each platform's "project" or "GEM" – dedicated spaces designed to retain context, though I quickly learned about their limited memory. One can use these tools to accomplish an automated newsletter. However, this is not really using AI, in the sense of learning how. Anyone can type in ChatGPT, "create a newsletter for me, here are the parameters/structure". However, the front end chat version of these AI models is simply the express, novice version. It is not actually optimizing the full use of AI in blending automation with learned and stylized content. Because these models retain limited memory, you are not going to get the same structure/format from week to week in having this newsletter created. In addition, as a user, you have to initiate the creation and prompt the model each week. Yes, it will create a passable newsletter for you, but you have to perform the task each time and you are at the mercy of the model's random creation. So, while this newsletter is something that can be done by the AI model chat interfaces, that's not the point. There are other, more complex, use cases that cannot be accomplished with the front end chat versions. This simple newsletter exercise demonstrates the process of building an actual AI Agent that could do more complex things and continue to 'learn', be fine-tuned. My first hurdle was crafting a respectable newsletter format. This was necessary because I had to have an acceptable format of a newsletter that would later be used to teach, 'fine-tune', an AI Model, as well as keep consistency throughout the process. This required dusting off my rudimentary developer skills. While I'm no coding expert, the chatbots proved invaluable for generating the HTML (HyperText Markup Language) layout using a "quasi-Vibe Coding" approach – a mix of intuition, iteration, and chatbot-assisted prompting. Using VS Code, my IDE (Integrated Development Environment), I iteratively built an HTML (HyperText Markup Language) foundation for the newsletter. It took me multiple tries to get the AI model (at this point, ChatGPT) to put together acceptable code that I could pass into VS Code and execute for an acceptable html. So, before even implementing any real AI Agent activity, I had to create my newsletter version. I did use the AI Model, ChatGPT, to accomplish this and it did save me tons of time. However, this is not the type of AI implementation that actual AI techs aim to utilize. That's important for people to understand. Also, there are 'no code' solutions and websites that could have been used to achieve this. However, you pay for that. The process that will continue to be outlined in future editions are cost free, which creates value. Please post comments, concerns, challenges, or anything at all in the 'Join the Dialogue' section. Join me next week as the adventure continues!
“The future is already here – it's just not evenly distributed.”
— William Gibson