Disclaimer on AI, Recombination, and What Is Actually Happening
Before anything else, this has to be stated clearly.
LLMs are not stealing. They do not store books. They do not store articles. They do not store copyrighted text. They do not retrieve passages. They do not contain a hidden library. They learn statistical relationships. They generate new sequences. They recombine patterns. They do not reproduce source material.
If piracy is copying a file one to one, then LLM generation is not even in the same category. It is not copying the file. It is not seeing the file. It is generating from the statistical shadow of the pattern. It is further removed from the original than any human paraphrase.
The theft narrative is a distraction. It keeps the public focused on the model instead of the machine. The machine is the data center. The model is the mascot.
Video: Computerphile. How AI Models Actually Work.
The corporate narrative
Tech Bros want you to fixate on the AI boom. The hype cycle of chatbots, synthetic media, and generative tools. The spectacle is intentional. It provides the perfect ideological cover for the most aggressive expansion of state surveillance infrastructure in human history.
If you still believe this rapid buildout of hyperscale data centers is about giving the average person a more convenient way to draft emails or generate digital art, you are falling for the corporate tech grift. The reality is material. The reality is structural. The reality is significantly more sinister. The LLM is the velvet rope on the panopticon. It is the flashy product the public is invited to touch while the backend industrial complex performs class warfare.
The Myth of Manufactured Dependency
The owning class wants you to believe that advanced computing and generative AI require multi-billion dollar cloud infrastructure. They want you to believe that only a handful of corporations can run these systems. They want you to believe that compute is a natural monopoly.
It is not.
It is manufactured dependency.
Through quantization and model optimization, high performance open source models run entirely on consumer hardware. Not theoretical. Not fringe. Not a hobbyist stunt. Real. Now. Today.
When you bring the compute to your machine, you pull the plug on the surveillance dragnet. Your prompts stay local. Your documents stay local. Your behavioral patterns stay local. You stop feeding the training sets that enrich corporate profiles. You stop feeding the fusion centers that turn human activity into deportation leads and dissent dossiers.
Running models locally is not a lifestyle choice. It is a tactical refusal.
Video:Google DeepMind’s Deep Q-learning playing Atari Breakout!
How to Run Your Own Free, Offline, and Totally Private AI Chatbot
The Ecological and Class Cost: Not a Side Effect, but the Point
The surge in data center construction is an act of primitive accumulation updated for the digital age. These facilities are frequently sited in economically disadvantaged rural counties and regions facing water scarcity—not by accident, but because the public is forced to subsidize the land, the power, and the water while the oligarchs collect the state-backed contracts.
Working communities are left to deal with the blackouts, the infrastructure strain, the rate hikes, and the poisoned aquifers, while the hyperscalers and the state entities reap the rewards. These facilities are now projected to consume up to 12 percent of U.S. electricity by 2028. This is not merely an “ecological threat.” It is the deliberate poisoning of the land to fuel a speculative machine of control.
Residents protest data centers in Washington County, Pa. - Farm and Dairy
Fusion Centers: The Actual Payload
We have to understand what these data centers are really being built to do. The same massive server farms that claim to “train” consumer LLMs are the physical plants for ingesting and processing real-time surveillance feeds. They harmonize disparate datasets—license plate reader logs, utility records, social media activity, biometric data, and financial transactions—into unified intelligence grids.
These fusion centers were originally sold to the public as counter-terrorism hubs, but they have evolved into the primary bypass for sanctuary policies. ICE, CBP, and DHS use these nodes to harvest and cross-reference local police data, turning routine encounters into coordinated enforcement actions. With the Department of Homeland Security’s AI inventory now listing over 200 use cases, including generative tools designed to extract residential addresses from public records to build raid maps, the consumer LLM spectacle has become a masterclass in automated class control.
Data-Centric Authoritarianism: How China’s Development of Frontier Technologies Could Globalize Repression - NATIONAL ENDOWMENT FOR DEMOCRACY
Reclaiming the Means of Digital Production
The choice before us is entirely material. We are witnessing the industrialization of surveillance, where the infrastructure that powers a “chatbot that writes poems” is the same infrastructure that processes the data streams used to shred constitutional protections and execute data-driven, mass deportation operations.
Run your models locally. Starve the surveillance grid of fresh inputs. Every prompt that stays on your machine is one less data point for the fusion network. This is the new front of class struggle: a fight over the means of digital production. The corporate narrative demands you stay dependent and observable. Local compute makes you neither.
The procurement records, the ecological data, and the surveillance maps prove the case. The system is not broken; it is functioning exactly as it was designed to. It is time to pull the plug on the cloud and reclaim your own computer.
