**The Problem with LLMs**

As I sit down to write about my experiences with Large Language Models (LLMs), I'm reminded of the blog post by Vijay Khanna that sparked our exploration of these tools six months ago: "From Idea to App in 7 Hours." The ability to rapidly develop new features and applications using LLMs is indeed a powerful one. However, as we've come to realize, this power comes with significant ethical concerns that cannot be ignored.

**The Plagiarism Problem**

At its core, the fundamental issue with LLMs is plagiarism. These models are trained on vast amounts of text data and can produce meaningful responses when prompted. But in doing so, they inevitably rely on theft – specifically copyright violation. LLMs need to "eat" and consume large bodies of copyrighted works, which may include source code, images, videos, or other forms of intellectual property. This creates a predicament for developers who want to use these tools: they are essentially participating in piracy.

But that's not the only issue. Plagiarism also requires dishonesty – passing off someone else's work as one's own. LLMs do both of these things, making them inherently prone to plagiarism. If you're concerned about copyright, then using LLMs is a problem. It doesn't matter if it's open-source code or the work of struggling artists; the fact remains that LLMs break source code licensing and perpetuate intellectual property theft.

**The Consequences of Using LLMs**

I've been experimenting with LLMs for a month now, despite my reservations about plagiarism. However, I've come to realize that the lies we tell ourselves are more insidious than a willingness to dip our toes into grey-area theft. The first benefit I've witnessed is accessibility in foreign languages. Translators can review translations instead of translating CSV files, making it easier for users who speak different languages to access the app.

Another significant advantage is accessibility for individuals with disabilities. As someone who has struggled with eye injuries and can no longer visually tokenize source code, LLMs have been a game-changer for me. They allow me to focus on higher-level thinking and design without getting bogged down in tedious coding tasks.

However, I've also witnessed some disturbing trends among developers using LLMs. There's a spectrum of approaches, ranging from cautious developers who refuse to touch LLMs to those who enthusiastically adopt them without hesitation. The latter group often finds themselves overworked, exhausted by the relentless pace required to keep up with LLM improvements.

**The Psychological Landscape**

As I've observed this phenomenon, I've come to realize that attachment and addiction are major concerns in the world of programming. Attachment refers to an emotional investment in the act of coding itself – the joy of creating something new, the satisfaction of solving complex problems. LLMs, however, take away these little pleasures by automating many tasks.

Addiction, on the other hand, is a performance-enhancer effect that makes developers feel superhuman when using LLMs. It's similar to steroids or vyvanse – once you get hooked, it's hard to stop. I've seen friends who have been overusing LLMs struggle with the consequences of their addiction.

**The Future of LLMs**

While I acknowledge the benefits that LLMs can bring, I'm concerned about the distribution models for these tools in 2026. There's an opportunity for lock-in, where proprietary models become so entrenched that they create a data gatekeeping and walled garden effect. This could lead to a new era of intellectual property control, with shareholders profiting at the expense of customers.

**Conclusion**

As I conclude this essay, I'm left with more questions than answers. Should we use LLMs in our nonprofit organization? Is it worth considering their accessibility benefits for individuals with disabilities? The higher-level ethical concerns are outside my jurisdiction, but I hope this piece provides some food for thought and encourages management positions to consider the topic of LLMs from multiple angles.

In the end, the changes that have come to the profession of programming are permanent. We need to adapt to these new tools, rather than trying to resist them or pretend they don't exist. As we move forward, let's be mindful of the potential pitfalls and work towards creating a more responsible, equitable, and accessible world – one where intellectual property is respected and developers can thrive without sacrificing their values.