The Struggle of a Dyslexic Developer
As a dyslexic developer, documenting projects has always felt like trying to solve an puzzle with missing pieces. Every new line of code I wrote brought with it a sense of ‘documentation dread’. I frequently leaned on excuses such as ‘it’s self-documenting code’ and techniques like ’living documentation’, consistently searching for creative detours around the documentation challenge.
But, as my career aspirations grew, I found myself at a juncture where avoiding this formidable task was no longer viable.
The Daunting Assignment
Recently, I tackled a short-term assignment that initially required the creation of a voluminous architecture document. Firm in my conviction that such grand documents are often more of a time sink than a useful tool—perhaps due to my personal aversion to both writing and reading them—I managed to persuade the team to reconsider this approach. Instead, we agreed on making a list of Architecture Decision Records (ADR), focusing on the individual, specific architectural choices that had to be made.
Yet, even with this compromise, the idea of detailing these 10 ADRs felt as daunting as an ascent up Mount Everest. Even though we had successfully captured the requirements and I had a clear grasp of the decisions involved, the sheer scale of the undertaking was overwhelming.
AI to the Rescue
Being a firm advocate for GPT in my development work, I realized its potential to help me overcome my literacy challenges. It all began with creating a prompt, embedded with our chosen ADR template. To this prompt, I added some clear directions like use American English and yield responses in raw markdown format.
With the initial setup complete, the next hurdle was describing the decision as I understood it. Rather than meticulously crafting a well-structured and cohesive narrative, I let my thoughts pour out onto the page. I tried to encapsulate the decision, including all the requirements and unique insights gathered during our refinement sessions. Every consideration, every detail discussed in our meetings, was thrown into this expansive, unstructured text. I paid no heed to crafting perfect sentences or structuring my thoughts – my focus was solely on capturing everything.
Once I unleashed my stream of thoughts onto the prompt, a striking transformation occurred. Individual ADRs started to come from the GPT model, each one taking about 10 to 20 minutes. The model functioned like a machine, quickly turning raw input into neatly formatted templates. What was once a daunting task now seemed like a smoothly flowing assembly line of ideas.
Transforming Mountains into Hills
ChatGPT, a nifty AI tool powered by its fourth-generation GPT model, served as my personal mountain guide, transforming the Everest-sized task in my head into a series of more manageable foothills. Instead of being faced with the daunting blank page, wondering how to express my thoughts, I was now presented with a series of AI-generated ADRs. It’s crucial to note, though, that these were just the foundations, rough sketches if you will, and not the finished structure.
Every ADR needed careful checking and changes to ensure they truly captured the essence of the decisions and requirements. It felt less like generating new content and more like refining a rough gemstone into a polished jewel. The work was far from over, but what this tool offered was a head start; a launching pad from which I could dive into the nitty-gritty of editing and refining. It gave me the momentum I needed to overcome the inertia that often comes with starting such a daunting task from scratch.
Looking to the Future
In the midst of working with GPT, I often find myself contemplating the future. Online, you’ll encounter the statement: “AI is not going to replace you, but someone using AI will.” This serves as a counter-narrative to the hype suggesting that AI will replace all knowledge workers, and it resonates with me. As it stands, I don’t believe AI is on the verge of replacing me. Yet, having witnessed the exponential growth of technology, who’s to say what tomorrow might bring? For me, the “but someone using AI will” portion of the statement seems a bare minimum. It allowed someone like me, inexperienced with documentation due to misplaced incentives resulting from my dyslexia, to take a substantial leap forward.
What I do know is this: AI is already helping me overcome my limitations and reach my full potential. It has become an invaluable tool in my developer’s toolkit. Even though I might have doubts, I’m looking forward to what the future of AI holds for us.