AI Newsletter 1_8_26
Caleb Bryant
This email is the full newsletter. For a printable copy, click here to download the PDF.
Welcome back to Under the Hood. I hope your holidays were fun and refreshing. Now, let's get back to picking up where I left off on the last newsletter with my AI newsletter story. In the last newsletter, I mentioned how I came to the realization that my laptop did not suffice for actually utilizing an AI model for training, inference (using a model to act on what it has already learned), let alone executing hard code. Using the model to execute the hard code was more useless than just executing the hard code itself. The model could not use its abilities sufficiently, nor was it trained enough to create the formatted content that was desired. So, the next step would be utilizing one of the popular LLM models (ChatGPT, Gemini, Claude, Grok, etc.) that are competing for everyday public use. How much would this cost? How do you utilize one from what is now a developing standpoint, rather than just front facing, question and answer format on the their respective sites? The only way to find out would be to stop asking questions, roll up the sleeves and start doing. I decided that I would start by utilizing Google Gemini. I chose Google Gemini because it seemed to me that Google was a little ahead of the curve in terms of providing software that assists in AI development. I was primarily a user of ChatGPT on a day to day basis, however, at the time Google had just released their newer model of Gemini and it was evident that they were catching up, if not surpassing ChatGPT (these model versions seem to leap frog one another each time they get released). One thing I found confusing at the start was that Google has several software platforms that, from an inexperienced user, seem to provide the same thing. There is Google AI Studio, Firebase Genkit, Vertex AI, among several other platforms that utilize AI in specific user based ways (Notebook LM, Gemini for Workspace). After some research, I found that Google AI Studio was the next level that I needed to progress to. It allows a person to create a project that can accept already existing files, like I had from VS Code. However, you can start from scratch, without any files at all. What is nice about Google AI Studio is that now you can begin creating projects that are long lasting and the AI model (of which there a few to choose from depending on our goal) has better memory capability within the chat sessions. In addition, you create Markdown files (Markdown files are widely considered the "Goldilocks" format for LLMs to track project status. A Markdown file is a lightweight plain-text file that uses a simple formatting syntax to indicate how the text should be displayed. When it is time to start a new session within Google AI Studio, within your project, you simply feed it system instructions again, in your new chat, along with the Markdown files for it to reference. In Google AI Studio, you simply begin the project by indicating what it is you want to create. This is where the use of API keys come in as well. Next week, we will talk about API Keys, cost structure, and how they are used.
“The future is already here – it's just not evenly distributed.”
— William Gibson — William Gibson is an American-Canadian speculative fiction writer and essayist. He is credited with popularizing the term "cyberspace" and for pioneering the cyberpunk subgenre. His work explores the intersection of technology, popular culture, and postmodernity.