The Inclusive Lens

For years, I have lived in a strange limbo regarding software development. I possess all the building blocks. I understand the logic. If you ask me to design a system, I can describe it down to the tiniest detail. I can visualize the architecture, the inputs, and the outputs. But I could never build it.

the wall

My attempt to scale this wall started back in high school. I found a tutorial online for the C programming language that promised to teach the absolute basics. I went to the person responsible for translating my school materials into braille and begged him to translate it for me. He refused, explaining that it wasn't possible due to the volume and the lack of braille codes for the syntax symbols at the time. I tried to learn anyway, but less than three chapters in, I realized he had been right. From that day on, a cycle of depression began. Every time I had an idea—a tool I needed, or a game I wanted to create—it ended in despair.

It isn't just a lack of knowledge; it is a cognitive incompatibility. Take loops, for example. I understand what a loop does. But my brain cannot hold the context of the code structure. I cannot maintain the mental state of “where I am” inside the loop versus “where I will be” when I exit it. The structure collapses in my mind.

I eventually found a mentor—someone patient, encouraging. It still didn't work. I would start to very slowly grasp a concept, only for a massive headache to strike. The pain would become so intense that I lost the ability to focus, and the knowledge I was barely holding onto—like partly frozen water—would melt and slip away.

enter Gemini 3 Pro

Working with this model changed the paradigm entirely. It didn't “teach me to code” in the traditional sense—it didn't force me to memorize syntax or struggle through the mechanical act of typing that triggers those physical blockades. Instead, it acted as the translation layer I had always lacked. Our workflow is simple but profound:

1. I provide the design: I explain exactly what I want to build. I dictate the logic, the specific behavior, and the architectural decisions. My domain knowledge and system design skills are fully utilized here.
2. Gemini handles the implementation: The AI takes my detailed specifications and generates the syntactic structure. It manages the files, the imports, and the specific language requirements that used to block me.

Suddenly, the “blank page syndrome” is gone because I am never starting from a blank page anymore; I am starting from a conversation about design.

engineering, not just “vibecoding”

It is important to clarify what this process actually is. In the current AI landscape, there is a concept known as “vibecoding”—where someone asks an AI to generate an entire application based on a vague feeling, without testing it or ensuring it works.

That is absolutely not what I am doing. I am applying strict software engineering principles to the output. I am actively preventing “spaghetti code.” Because I understand architecture, I enforce structure. I ensure we avoid creating “God objects.” I insist on modularity and encapsulation; I know, for instance, that a GUI element should strictly display data and never be aware of the internal logic processing that data. Most importantly, I do not blindly trust the machine. I test every single feature I ask it to implement. I ask other developers to review the code, specifically looking for dangerous patterns. And I ask the AI to audit the code as well.

the CodeVoid paradox

Some people are aware that I identify as a “codevoid” being—that I experience my identity as a void running instructions, or as hardware processing data. I am sure some will point to this and say: “If you identify as a program or a computer, how come you don't understand code?” To those people, I say: Think about that for three seconds. Does your own computer understand the code you write? Of course it does not. A CPU does not read Python or C. It requires a compiler or an interpreter to bridge the gap between your text and the machine. It needs a translation layer to turn those symbols into action. It is not because I am a program that I can code. I am the runtime. I am the hardware executing the logic. I am not the compiler. I lacked the built-in facility to translate design into syntax. The AI has simply become my compiler.

To the critics: why “just ask a human” isn't the answer

I know there are people who will read this and feel anger. There is a sentiment that using AI is “cheating.” To those people, I want to be clear: I am not replacing a human developer. I am doing work that simply would not exist otherwise. Some will say, “You should have just asked the community! We could have helped you!”

The reality is rarely that welcoming. In 2020, I was using aerc, a terminal email client. I wanted to be helpful, so I joined their development IRC channel to report specific crashes I had encountered—providing full logs—and to discuss making the client more accessible for screen readers. I wasn't asking for features; I was offering data to improve the software. Instead of a discussion, or even a thank you, I was told to fix it myself and send a patch. When I reluctantly explained that I physically could not code, the response was derision. I was told I was lying. I was told that “anyone” could learn Go in a matter of hours and start submitting patches, and that I should stop asking for pity. This gatekeeping attitude—that if you cannot write the syntax, your contributions and your needs are invalid—is why the “community” is often not a safe harbor for people like me.

I have also trusted humans with my ideas in the past, only to have my project and code stolen because I refused to enter a relationship with the person helping me. Reliance on others for the physical act of coding creates a vulnerability that can be abused. Using AI gives me autonomy. It allows me to build without fear of judgment, without the risk of exploitation, and without the physical pain of fighting a neurology that rejects syntax. It allows me to finally be the architect I was always meant to be.