The thing I was most surprised by this week was the usability and existence of closed LLMs as opposed to the much more popular open LLMs. Most of the time people’s biggest concern with AI usage, or at least mine, is the ability to hallucinate, which is substantially less likely to happen if you can directly have control over the sources it tries to employ. By having a closed LLM to only use exactly the sources I provide makes certain types of work way more trustworthy. This doesn’t mean closed LLMs are the greatest resource of all time, if I have the sources I already know I need then it makes summarizing or studying them a lot faster and more trustworthy. However, open LLMs are still going to be extremely helpful when actually looking for sources or work that I don’t already have saved, so when looking for articles on JSTOR or other archives ChatGPT and Claude are still going to be my go-to’s. Outside of learning the basics of how an LLM works, the UT proper prompting guide was kind of interesting to me in how many ways they thought of giving prompts to get a specific answer they want but I didn’t necessarily find it helpful as my AI usage is already pretty accurate for what I use it for. The image generation on ChatGPT is also pretty good, much better than I was expecting it to be. Just look how trustworthy ChatGPT is the photo it generated for me when asked what he looks like in the real world.

University of Texas, 2025, UT Aspire Prompt Literacy
ChatGPT prompt: give me a picture of chatgpt as a physical robot