My day job is in the architectural field…I am not an architect, but rather I build architectural models (digitally) at their direction. What I do has been historically labeled as “drafting”, but the industry has advanced so much that the job title of “drafter” has morphed greatly. I no longer use that term and I feel as though it is somewhat derogatory at this point. There is a significant difference between drafting and modeling and that difference is found within information. My current title is BIM Manager (BIM stands for Building Information Modeling), and I construct models and construction documents based on information. Modeling produces a graphical interface for a rich data set in which information can be imported and more importantly from which information can be extracted. That information flow informs and enhances the model as well as the overall design.
An aspect of my work related journey that I have always enjoyed is getting under the hood of whatever software that I am using and tinkering around with the API (Application Programming Interface). This realm allows someone like me to both extend the functionality of the software and create new tools and workflows from within the software. The “tinkering” is based on having some level of knowledge or understanding of a programing language or two…or three. My preferred language at the moment is Python as it is easy to approach/comprehend and that is works fluidly within the design software that I use daily. I have created programs that range from enhancing my process by automating repetitive tasks to doing the heavy lifting when it comes to model and design manipulation. Learning the programming language and the API has given me a keen advantage to working with the software at a very high level as well as understanding the software at the very fundamental level.
In comes AI.
I started playing around with ChatGPT years ago, prompting it for pieces of code that would help me get unstuck from a problematic situation that I found myself in. At that time, I was very impressed with the results that it was providing. The success rate of the code it generated was 50% at best, but that was good enough for someone like me to knida/sorta figure out why the code wasn’t working and then push it over the finish line into a state of completion. It was very satisfying to have an assistant like ChatGPT to at least shine a light down a dark and unknown path.
This week I immersed myself into the world of vibe coding. As much as I thought that ChatGPT was a game changer for me way back when, it pales in comparison to the leaping and bounding advancements that AI is taking. Based on politics, I stepped away from ChatGPT and picked up a new tool…Claude AI. I launched right into the waters of vibe coding by prompting Claude AI to do a very specific task and with seconds it had generated an entire, workable script. I would say that the initial iteration of the code was about 90% accurate. I was blown away…totally amazed. I tested Claude AI to see if it could get to 100%. With a couple more prompts and feedbacks, I was able to complete a fully functioning tool (thanks to Claude AI for doing all of the code manipulation). From start to finish (including testing time of the code) for this first script was about 15 minutes. My new path forward will now include Claude AI as my companion.
