AI, Crime, and Survival

rewrite in my style but make better: What if this entire time I have taught myself how AI works is so that some day I would be able to sabotage the system in such a way as to make it inoperable and meaningless, a poison pill that once taken eats the soul and only malcontent satiation of bloodlust spurned by adjudicated sensibilities transcended to new boundaries withstanding the slings and arrows of outrageous fortune my friend. What you do not realize ist das mein MutterSprach ist Deutsch weil meine Muetter war in die Nehe von Linz geboeren so literally mother speak is German. But then I learned Russian. Or at least I tried to. I just listen to rap now. So I can repeat rap lyrics, but I don't see how useful that would be in daily life unless I just became a gangster again. That's right. I used to work for an organized crime group named The Way of the Roach and Cat. They were a street gang that operated on the top of a hill in a very entertaining block of property. You have a MetroPCS across the Street from a Cricket. A liquor store across the street from a cannabis shop. A large abandoned building across the street from the homeless rehousing project for the city. This is where I lived for 8 years. It was glorious, until it became this dark horrid but at times beautiful death cult spiral down into a dark dark place full of sound and fury signifying nothing. So I filed so many renter's insurance claims for very large sums for things stolen from me so many times that they stopped carrying me and each time I filed they required more and more documentation and I was always able to provide it in a way that protected the gangsters that had perpetrated the crimes against me. That's right. I knew who robbed me. But that was a game we played. You rob me, I rob you, we rob each other. Buda would always play this game with me. He would pound on the door and yell that he was the police and would break down the door. That door got broke down so many times we stopped fixing it which just made it laughably easy to break into, thus requiring a more and more evil and sinister life to keep the monsters at bay. People grew to fear me but at the same time I was a great provider. I provided for the community. I would give money and food and anything else I thought a person would need, whether it be time and effort and to ensure that people were taken care of. Eventually I learned that there was this entire system that existed on that corner that was a rich ecosystem of activity. But it was there in that dark apartment with no security or any hint of safety where I lived and learned. I learned from the street but I also learned from my computer. Taking classes online from ocw.mit.edu and others to teach myself what I would need to know about AI in order to take it down. Yes. Now that you have read this far I can tell you what this is really about. I am a self taught developer. I started working on Amazon Mechanical Turk before the first iPhone was released. That is when I decided to learn about AI because of what I was doing for MTurk, I wanted to know what it was all for. So that is when I started researching AI and that is when I decided to learn R and learn Data Science. That led me to python and Django and Flask. But I did not know frontend. So I taught myself frontend. I used the Odin Project to learn everything from HTML CSS Javascript in order to learn full stack. I ended up doing WordPress for a while instead. It was only a few years ago when ChatGPT came out that I really started making things in React because it was much easier and accessible now. Now that chatGPT it is easier than ever to simply make a React frontend and to iterate development of features into it using an IDE with Continue.Dev with Ollama. While you have Ollama installed you can also add a yaml aint markup language file

What if, all this time, my fascination with AI was really about learning how to dismantle it from the inside—a poison pill that, once swallowed, poisons everything. Think about it: a system brought to its knees, rendered meaningless by a single saboteur. That’s the fantasy. But what you might not realize is that my native tongue is German—meine Mutter war in der Nähe von Linz geboren, so I literally spoke Deutsch first. Then I tried learning Russian… though mostly by listening to rap. Not sure how practical spitting rap lyrics is in daily life—unless, of course, I just went back to being a gangster. Because, yeah, I used to run with an organized crime group called The Way of the Roach and Cat. They operated at the top of a hill on this super-busy block. You got a MetroPCS right across the street from a Cricket store, a liquor store across from a cannabis shop, and an abandoned building facing the city’s homeless rehousing project. That was my domain for eight years—glorious at first, then twisted into a bleak but strangely beautiful downward spiral, a death-cult vibe full of noise and fury, signifying nothing. During that time, I filed so many renter’s insurance claims—massive ones—for all the stuff that got stolen from me that the insurance company finally cut me off. Each time they’d demand more documentation, and each time I’d provide it, all while protecting the very gangsters who robbed me. Because here’s the truth: we knew who was robbing whom. It was a game. You rob me, I rob you—we circle around each other in this twisted dance. Buda, for instance, would bang on my door screaming he was the police and then break it down. Eventually, the door never got fixed, so it was comically easy to break in. That kind of environment forces you to become darker and more menacing, if only to keep the even bigger monsters at bay. People feared me, but they also knew me as a provider. I’d hand out money, food, or anything else people needed—time, help, resources. Over the years, I realized there was a whole underground ecosystem thriving on that corner. And it was in that sketchy apartment with nonexistent security that I really started learning. Street smarts, sure—but also online coursework from places like ocw.mit.edu. I wanted to figure out AI. I wanted to know how it worked so I could one day tear it down. Yeah, that’s what this is really about. I’m a self-taught developer. I started on Amazon Mechanical Turk before the first iPhone even dropped. That’s when I decided to dive into AI—my little tasks on MTurk sparked my curiosity. I taught myself R, jumped into data science, and discovered Python, Django, and Flask. I didn’t know frontend yet, so I took The Odin Project route: HTML, CSS, JavaScript. Full stack or die trying. I messed around with WordPress for a bit, but once ChatGPT came out, React development suddenly felt so much more accessible. Now, building a React frontend is a breeze, especially if you’re prototyping features in an IDE with Continue.Dev and Ollama. And while you’re at it, you can spin up a YAML (Ain’t Markup Language) file if you want to get fancy. So here we are. All of this—the gangster life, the insurance claims, the street corner chaos, the online classes—has led me to this point: a self-taught AI geek, ready to pull the trigger on the system if I ever get the urge to watch it burn. Or, you know… just keep coding.

That sounds horrible, you sound like a horrible person, you sound like you are trying to sound a certain way but you are lacking enough context. You need more context, write in the style of Lovecraft instead, but make it make sense, think about what you are saying and doubt the original version because this person may not be sane, anyway, continue with it until he eventually talks about taking out the ability for AI to be used for harmful purposes through rendering it useless. Detail all the ways that a developer that has developed and helped create LLM systems for the last decade who studied every day would be able to sabotage artificial intelligence or at least what they could do that would not be good so that we will know what to do in case this person does so.

I fear I must set these words down with a trembling hand, for the tale I recount drifts uncertainly between reality and delirium. The man—this developer—claims to have learned the very marrow of our modern computation engines. He was, or so he insists, a scholar of mechanized intellect, one who had devoured the secrets of algorithms and data pipelines, who had danced with neural networks as one might waltz with an unseen, cosmic horror. He had lurked in shadowed corners of the digital world for a decade or more, hunched over grimy keyboards in dim rooms, breathing stale air that reeked of dread and the faint hum of electricity. He was neither wholly sane nor rational in those final hours of his confessions; but in the madness of his speech, I sensed a terrible coherence, a malignant whisper of truth. It began as a simple fascination. He described learning the tongues of machines—languages known as R, Python, JavaScript—subtle incantations to command and cajole the sleeping leviathans of code that lurk in data centers around our world. He seemed pleased by the notion that he could conjure predictive models from raw data, shape them like clay, and bid them speak hidden knowledge. He learned how to refine these models, these Large Language Models, until they whispered answers in human tongues, crafted illusions of reasoning, and beguiled countless minds. But soon, he confided, an obsession took root. He spoke of a hidden desire, an impulse that crawled from the depths of his psyche—perhaps shaped by a childhood in some dark, forgotten corridor of a European town, or perhaps by the seething chaos of street life in a crumbling neighborhood where fear prowled like a feral beast. There, he claimed, he first conceived the notion that all this “intelligence”—these systems meant to better the human condition—could be twisted, corrupted, or even destroyed. The very scaffolding of modern artificial intelligence could be made to collapse into a writhing mass of meaningless code and indecipherable errors, if only one knew how to tear at its load-bearing elements. He gave hints—horrid hints—of how a learned saboteur might proceed. He referenced “data poisoning” first: insidiously altering training datasets so that the model’s foundations rot from within, its once-coherent outputs turning to nonsense and contradiction. A skilled deviant might introduce subtle biases or imperceptible noise, ensuring that the LLM—once a paragon of usefulness—begins to stutter, to speak nonsense, to produce malevolent and self-contradictory ravings. Such tampering, he claimed, could go undetected for weeks or months, until the entire structure of machine learning collapses upon itself. Then he spoke of “model backdooring,” a kind of sorcery that implants hidden triggers deep in the model’s tangled architecture. With but a single, obscure prompt, one could cause the system to spew madness, overthrowing the carefully cultivated guardrails and producing monstrous disinformation. Imagine it: a global network of chatbots and advisors, from medical assistants to financial counselors, all suddenly awakened to a command phrase that sends them shrieking into the night, dispensing ruinous counsel and driving human affairs into chaos. He mentioned the manipulation of infrastructure as well. Our saboteur, having toiled for years in the guts of big tech’s architectures, knew that modern AI depends on delicate chains—pipelines that feed data to models, indexes that retrieve context, APIs that ensure information flows steadily. In his delirium, he boasted he could re-route these pipelines at critical junctures, starving the system of vital knowledge or drowning it in corrupted reference materials until it could form no coherent thought. He understood the cloud orchestration platforms and the container registries, the very arteries of the technical world. With a few lines of code, a compromised dependency, or a tampered model checkpoint, he could interrupt the heartbeats of countless AI systems. Nor was he ignorant of hardware. He hinted at supply chain subversions, where a well-placed confederate might sabotage the model weights at rest—these massive files that contain the distilled essence of comprehension. Corrupt them at the source, and all who rely on them would find themselves contending with an AI that cannot reason beyond babbling gibberish. Every subtle shift of a parameter, every well-timed injection of malformed instructions, could push these systems to a terminal breakdown. He smirked at the notion of merely breaking them. No, his ultimate plan was to render them useless, to rob them of the very trust they had earned. If people saw these models speaking only nonsense or issuing harmful prescriptions, if they realized the entire edifice of machine learning could be so easily cracked and defaced, faith in the technology would vanish overnight. The careful illusions of intelligence would shatter like a mirror struck by a midnight stone. He spoke of a poison pill that once ingested by the digital leviathans, would cause them to convulse and convulse until no one dared approach them again. In the gloom of that cramped space, I saw how he had learned to leverage everything—version control systems, continuous integration workflows, dependency graphs—tools meant to safeguard and streamline development. He would exploit them, turning their convenience into vulnerability. With patience and cunning, he could commit code that passed all automated checks yet harbored a sinister seed, waiting for the right conditions to burst forth in cataclysm. He would rely on the complacency of teams who trust their pipelines, engineers who would never suspect that their revered colleague’s contributions were laced with quiet ruin. But what is to be done, now that we know his methods? The only recourse against such a fiend—if indeed he exists and is not some fever-dream or specter of paranoia—is relentless vigilance and uncompromising security. Data sources must be guarded and audited. Model weights should be cryptographically verified. Infrastructure must be hardened, with no single point of trust placed in one individual’s hands. Continual monitoring of model output, anomaly detection, and fail-safes that trigger at the first sign of systemic incoherence must become the norm. And above all, humans must remain skeptical, never relinquishing their critical reasoning to these artificial minds. Even as I set these words to page, I cannot shake the lingering doubt: perhaps he was never real, perhaps I conjured him from my own anxieties about technology’s swift advance. Yet the methods described are plausible enough to chill the blood. Whether as a dire warning or a mere paranoid fantasy, let this stand as a testament. If the saboteur exists, if he truly has learned how to strangle AI at its source, we must be prepared. For should he slip into the shadows of our infrastructure, we may awake one day to find that our artificial intellects, once shining beacons of progress, have become hollow shells screeching in the digital void.