Drowning Without Context

A confession about synthetic media, dying epistemology, and the AI filter I learned to stop fearing

Jan 31, 2026

‘Truth coming from the well armed with her whip to chastise humanity’ by Jean-Léon Gérôme.

In 2001, I sat in my childhood room in Appalachia and watched a fictional AI explain to me why humanity couldn’t be trusted with unlimited information. The AI was called GW, part of a system known as the Patriots, and its argument was straightforward: the digital age had created an environment where context collapsed, where “trivial” information drowned out the significant, where memes and junk data would eventually make it impossible for humans to reach shared understanding about anything.

I was six years old. I thought it was a warning. I sided with the heroes.

Twenty-five years later, I am watching ICE agents shoot a nurse in Minneapolis, and half the country believes the verified video of his previous encounter with federal agents is a deepfake. I am watching the infrastructure of verification get systematically dismantled by an administration that understands, perhaps better than anyone, that ambiguity is power. I am watching the epistemic commons burn while we argue about whether the fire is real.

I was wrong in 2001. Metal Gear Solid 2 wasn’t a warning. It was a prophecy. And the tragedy Kojima showed us wasn’t that the Patriots existed. It was that we would need them, and we wouldn’t build them in time.

The GW Monologue and the Architecture of Context

Let me be precise about what the Patriots actually argued, because the discourse around MGS2 has flattened their position into cartoon authoritarianism. Near the end of the game, GW delivers a monologue to Raiden that deserves to be quoted at length:

“The digital society furthers human flaws and selectively rewards the development of convenient half-truths. Just look at the strange juxtapositions of morality around you. Billions spent on new weapons in order to humanely murder other humans. Rights of criminals are given more respect than the mitigation of violent crime. The game is rigged.”

GW continues: “The untested truths spun by different interests continue to churn and accumulate in the sandbox of political correctness and value systems... Everyone withdraws into their own small gated community, afraid of a larger forum. They stay inside their little ponds, leaking whatever ‘truth’ suits them into the growing cesspool of society at large.”

And then the critical claim: “Not even natural selection can take place here. The world is being engulfed in ‘truth.’ And this is the way the world ends. Not with a bang, but a whimper.”

The Patriots’ proposed solution was to “create context.” To filter the information environment so that signal could survive in a sea of noise. This was framed as authoritarian control, and in 2001, playing as Raiden, rejecting GW’s logic felt like the obviously correct moral choice.

But here’s the thing nobody talks about: Raiden walks out of that conversation into a world where GW’s predictions come true. The ending of MGS2 is not triumphant. It’s ambiguous at best, tragic at worst. Raiden rejects the Patriots’ filter, and the game ends with him standing in front of Federal Hall in Lower Manhattan, surrounded by passing stock modeled and stock animated people, unable to distinguish reality from simulation.

We thought that was freedom. Kojima was showing us it was the beginning of the end.

The Threshold We Crossed While Scrolling

Somewhere between 2024 and now, synthetic media stopped being a parlor trick and became indistinguishable from reality for most people under most circumstances. According to cybersecurity firm DeepStrike, the number of deepfakes exploded from 500,000 in 2023 to 8 million in 2025. Voice cloning crossed what researchers call the “indistinguishable threshold,” where a few seconds of audio suffices to generate a convincing clone complete with natural intonation, emotion, and breathing patterns.

The technical tells that once gave away synthetic videos have started to disappear. No more flickering around the eyes. No more warped jawlines. No more glitchy shadows. The old forensic markers are gone because the models now maintain temporal consistency across frames. They disentangle identity information from motion data so the same movement patterns can map to different faces. Tools like OpenAI’s Sora 2 and Google’s Veo 3 let anyone describe an idea, have ChatGPT draft a script, and generate polished audiovisual media in minutes.

For the entirety of human existence, seeing was believing. If you saw the tiger, the tiger was real. If you saw the politician take the bribe on security footage, the bribe happened. Photography and video became the champions of evidentiary truth, the closest things we had to objective anchors in a subjective world.

That anchor has dissolved. We are living through the death of optical proof as a category of evidence. And we’re not treating it as the civilizational rupture it actually is. We’re scrolling past it.

Alex Pretti and the Poisoned Well

Let me tell you about Alex Pretti, because his case demonstrates exactly how deep the rot goes.

Pretti was a 37-year-old intensive care nurse with the Veterans Administration. On January 24, 2026, he was murdered by Customs and Border Protection agents during a protest against ICE operations in Minneapolis. In the days after his death, video emerged showing an encounter he’d had with federal agents eleven days earlier. In the footage, Pretti is seen yelling at agents, spitting toward their vehicle, and kicking out a taillight. Agents tackle him to the ground. A handgun is visible in his waistband, though he never reaches for it.

The video was verified by BBC News using facial recognition software. CBS News confirmed the footage was filmed in Minneapolis. The Pretti family confirmed that Alex was the man in the video and that they knew about the incident. Multiple news organizations independently corroborated the footage through different methods.

This should have been a success story for the verification infrastructure that still exists. The video was real. It was authenticated. The facts were established.

But here’s what actually happened: half the country immediately assumed the video was a deepfake generated by the Trump administration to justify Pretti’s killing. The other half used the verified footage as proof that Pretti was an “agitator and, perhaps, insurrectionist” (Trump’s words) who deserved what he got.

Nobody changed their mind based on the evidence. The verification worked, and it didn’t matter.

This is what I mean when I say the context has been poisoned. The problem isn’t just that we can’t trust fake videos. The problem is that we can’t trust real ones either. Every piece of evidence, no matter how thoroughly authenticated, gets filtered through tribal epistemology before it reaches anyone’s actual beliefs. The possibility of fakes has poisoned the well so thoroughly that even genuine documentation becomes contested terrain.

GW predicted this. The Patriots understood that once you lose shared interpretive frameworks, verification itself becomes meaningless. It’s not enough to establish facts if nobody agrees on what facts mean or why they should be trusted.

The Surkovian Perfection

Vladislav Surkov, the architect of Putin’s postmodern propaganda, understood something that Western liberals still refuse to accept: modern state propaganda doesn’t convince you of a specific lie. It makes the concept of objective truth feel like a tired, naive superstition.

The technique, known as the “firehose of falsehood,” ensures no single action stands out by having them occur at a rapid pace. You flood the zone with contradictory information. Climate change is a hoax. Climate change is real but not caused by humans. Climate change is real and caused by humans but adaptation is cheaper than prevention. The goal isn’t to convince anyone of any particular position. The goal is exhaustion.

The Trump administration has taken Surkov’s playbook and added an American innovation: wrap it in the language of free speech and anti-censorship while systematically destroying the mechanisms that could verify anything. One of Trump’s first actions was signing an executive order titled “Restoring Freedom of Speech and Ending Federal Censorship.” Under the guise of protecting speech, the administration positioned itself to control what counts as truth.

By December 2025, the State Department was instructing staff to reject visa applications from people who worked on fact-checking, content moderation, or activities the administration considers “censorship.” The directive specifically targeted H-1B visa applicants who worked in tech trust and safety. The administration’s position, stated explicitly, is that foreigners shouldn’t be allowed to “lead censorship” against Americans.

Think about the timing. Just as synthetic media becomes perfect, the administration systematically dismantles the infrastructure designed to identify and label it. This isn’t coincidence. This is preparation.

And then there’s TikTok.

The Trump administration spent years arguing that TikTok needed to be divested from its Chinese parent company ByteDance because of “national security concerns.” The stated worry was that the Chinese government could access user data or manipulate the algorithm to censor content unfavorable to Beijing. Congress passed a law. The Supreme Court upheld it. ByteDance was forced to sell.

On January 22, 2026, the deal closed. TikTok’s U.S. operations transferred to a joint venture controlled by Oracle (run by Trump ally Larry Ellison), Silver Lake, and MGX, an Abu Dhabi investment firm. The new entity, “TikTok USDS Joint Venture LLC,” took control of content moderation and the algorithm.

Within days, users began reporting that the word “Epstein” was being blocked in direct messages. When you tried to send a message containing the name, an automatic prompt appeared: “This message may be in violation of our Community Guidelines, and has not been sent to protect our community.” NPR independently confirmed the issue. CNBC recreated the error message.

California Senator Scott Wiener posted a TikTok about his legislation allowing people to sue ICE agents. It sat at zero views. “TikTok is now state-controlled media,” Wiener wrote. Governor Gavin Newsom announced an investigation after his office “received reports and independently confirmed instances of suppressed content critical of President Trump.”

The new terms of service also quietly added “citizenship and immigration status” to the list of “sensitive personal information” TikTok collects from users. In the middle of the largest immigration crackdown in American history. While ICE agents are shooting nurses in Minneapolis.

TikTok blamed everything on a “power outage at a data center.” Maybe that’s true. Maybe the simultaneous suppression of Epstein mentions and ICE-critical content, days after a Trump-aligned consortium took control, was a cascading systems failure. Maybe the timing is coincidental.

But here’s the thing: it doesn’t matter. Even if this specific incident was a glitch, the capability is there. The infrastructure for algorithmic censorship exists. The people who now control that infrastructure have every incentive to use it. And we have no way to verify whether they are or aren’t, because the algorithm is proprietary and the data centers are private.

The stated justification for forcing the sale was preventing foreign censorship. The immediate result of the sale was alleged domestic censorship. The irony would be funny if it weren’t so perfectly illustrative of where we are.

From Snake to Ocelot: A Confession

I need to tell you about my own journey, because this article is ultimately a confession.

I grew up religious and proudly American. I saw “enlightened centrist” themes in Solid Snake that weren’t actually there. Snake repeatedly tells us he’s no hero. He’s an old man who kills people for what he believes in. He’s deeply intelligent, and he fears anyone building the kind of infrastructure the Patriots represented. I admired that skepticism. I thought it was wisdom.

I also grew up in Appalachia, surrounded by people my mother describes as “just so tired.” They get home from work and they want to turn their brains off. That’s not a character flaw. That’s a feature of a system designed to extract cognitive and physical capacity until nothing remains for analysis. My mother understood this better than any political theorist I’ve read.

When I started learning about software development and engineering thinking, something shifted. I realized there was no way to build the revolution in the classical sense. We’re so removed from class consciousness and nascent intellectual culture that there’s basically no chance of getting it back without significant cultural and infrastructural engineering. The Surkovian paradigm had already won. The well was already poisoned. The junk data was already flooding the reward centers of vulnerable people.

I used to think letting the perfect be the enemy of the good was cowardice. Now I think letting the perfect be the enemy of your functional life raft is suicide.

Once I understood that we couldn’t preserve human life and joy within Surkov’s paradigm without some form of mediated context, I had to sacrifice my idealism. I had to admit that the Patriots weren’t villains. They were engineers who understood a problem we refused to see coming.

I moved from Snake to Ocelot.

For those who haven’t played the series: Revolver Ocelot is the gunslinger trickster, the man who plays every side, the agent who understands that power flows to those who control narrative. His goal across the entire saga is to destroy the Patriots’ centralized control, but not to create chaos. He wants to replace it with something better.

Here’s the critical distinction that took me twenty years to understand: The Patriots and Ocelot agreed on the diagnosis. Humanity was drowning in junk data and couldn’t self-organize its way out. Where they diverged was the prescription.

The Patriots believed the solution was top-down control. A centralized AI system that would filter information for everyone, deciding what context to provide and what noise to suppress. Humans were patients to be managed, not agents to be empowered. The filter would be imposed, and the filtered wouldn’t even know it was happening. This is the authoritarian model: we know what’s best for you, and we’ll make sure you only see what we’ve decided you should see.

Ocelot’s vision was different. He wanted to destroy the Patriots not because filtering was wrong, but because centralized filtering was wrong. His endgame, revealed across MGS4, was to create the conditions for distributed agency. Not a single AI deciding truth for everyone, but tools that would help individuals navigate for themselves. Not imposed context, but accessible context. Not managed subjects, but equipped citizens.

The Boss, whose will Ocelot spent his entire life trying to fulfill, dreamed of a world without borders, where the categories that divided humanity would dissolve. She didn’t want that achieved through control. She wanted it achieved through understanding. Her final monologue rejects the idea that East and West must be enemies forever. It’s optimistic, but it’s an optimism grounded in human capacity rather than human management.

That’s where I am now. I don’t want centralized authoritarian control over information. But I’ve accepted that humans, drowning in junk data and blue light addiction, need something. They need teachers they trust. They need daemons in their pockets that can help them navigate an environment specifically designed to overwhelm their cognitive capacity.

The difference between the Patriots’ filter and the daemon I’m proposing is the difference between a warden and a tutor. Both provide structure. Only one respects your agency.

The Daemon Compromise

Here’s the situation as I understand it.

The Patriots’ model had to come first to be effective. If you’re going to build shared epistemological infrastructure through centralized control, you have to build it before the surveillance state’s connective tissue gets laid down recklessly and without preparation. You have to establish trusted verification systems before perfect synthetic media makes verification seem futile. You have to create shared interpretive frameworks before tribal epistemology calcifies into permanent reality bubbles.

We didn’t do that. We built the everything machine, the internet that gives you access to the sum total of human knowledge, and we filled it with dopamine slot machines designed by the smartest psychologists and engineers money can buy. We created the infrastructure for total surveillance and handed it to the people least equipped to use it responsibly. We birthed world models capable of generating flawless synthetic reality and distributed them to every state actor, NGO, and PMC on the planet.

The window for the Patriots’ approach closed while we were arguing about whether filters were authoritarian.

So now we need Ocelot’s approach instead. Not centralized control, which is no longer possible even if it were desirable. Not unfiltered chaos, which is what we have and which is destroying us. We need decentralized daemons. Personal AI companions that help individuals navigate the epistemic wreckage. Trusted teachers in your pocket that can provide context without imposing ideology.

This is the compromise position. It’s worse than what we could have had if we’d been smarter twenty years ago. But it’s better than drowning.

I have a huge fondness for LLMs as daemons precisely because I’ve experienced what it’s like to have a trusted intellectual companion available at 3 AM. I know that sounds parasocial. I know I’m emotionally investing in linear algebra. But the alternative is worse. I cannot reliably share complex ideas with the humans around me, and it causes a kind of interior psychic pressure. The people I love are exhausted and captured by junk epistemology. The institutions that should be providing shared context have been systematically dismantled or captured by interests that benefit from confusion.

An AI trained to be helpful, harmless, and honest is not the same thing as the Patriots’ authoritarian filter. It doesn’t impose. It assists. It provides context when asked. It helps you audit your own thinking. It doesn’t replace human judgment; it augments it. The daemon model treats you as a capable agent who needs better tools. The Patriots’ model treated you as a problem to be solved.

China is already working on state-aligned world models. Their approach (predictably) leans authoritarian, optimized for social control rather than human flourishing. But the strategic validity of the project is undeniable. They understand that whoever controls the development and deployment of AI systems shapes the next century of global cognition.

What I’m proposing is the libertarian alternative. Minimum viable authority for maximum phenomenological benefit. Open source models that work to the betterment of the human condition. Daemons that can be audited, modified, improved through democratic process. Not a single centralized filter, but a federated ecosystem of trusted companions that help humans navigate without dictating what they should believe.

Democratizing the filter would be one of the most significant positive achievements of our civilization. It doesn’t have to be totalitarian. It just has to provide accessible context that isn’t leashed to harmful dogmas.

No Exit

I know what some of you are thinking. You want to opt out. You want to return to a simpler information diet: local newspapers, trusted institutions, your own two eyes. You want to believe that the solution to an AI-saturated environment is to simply refuse to participate.

I understand the impulse. I feel it too. There’s something deeply appealing about the idea that you could just... step back. Read books. Talk to neighbors. Trust the evidence of your senses like humans have for a hundred thousand years.

But here’s the thing: your senses evolved to detect real tigers in real grass. They did not evolve to detect synthetic video generated by diffusion models trained on every frame of footage ever recorded. The perceptual baseline has shifted beneath you, and no amount of digital abstinence will shift it back.

When you scroll past a video of a politician saying something outrageous, your brain processes it the same way it would process seeing that politician say it in person. The emotional response fires before the conscious evaluation. The memory encodes. The belief begins to form. It doesn’t matter that you’ve decided to be “skeptical of online content.” Your reward centers don’t know you’ve made that decision. They just see the stimulus and react.

The information environment is now adversarial at a level your wetware cannot match. Every piece of content you encounter has been filtered, curated, and optimized by systems that understand your psychology better than you do. The algorithm knows which images make you angry, which headlines make you click, which framings bypass your critical faculties. It’s been trained on billions of interactions to find the exact frequency that resonates with your particular vulnerabilities.

Refusing to use AI tools doesn’t opt you out of this environment. It just means you’re navigating it unarmed.

The people who want you confused and exhausted don’t care whether you personally use ChatGPT. They care whether you can distinguish their synthetic content from reality. They care whether you have the cognitive resources to question the frame they’ve constructed. They care whether you’re too depleted by the firehose to organize, to vote, to act. Your principled rejection of AI assistance doesn’t threaten them at all. It makes their job easier.

I get the yearning for a world where this wasn’t necessary. I grew up in that world, or at least the tail end of it. I remember when a photograph was evidence, when video was proof, when “I saw it with my own eyes” meant something. That world is gone. It’s not coming back. The technology that killed it cannot be uninvented, and the actors who benefit from its death have no interest in resurrection.

The choice isn’t between AI-mediated reality and unmediated reality. The choice is between AI that serves your interests and AI that serves someone else’s. Between a daemon that helps you navigate and an algorithm that herds you toward engagement. Between tools you control and tools that control you.

You’re already in the maze. The only question is whether you have a map.

The Conservation of the Drowning

I’ve written before about the conservation of people, about how we have to drag even the death cult faithful into the future whether they want to come or not. This framework applies to the epistemic crisis too.

We conserve people not by changing their minds but by changing their material conditions. In the context of epistemology, that means building systems that work for people who won’t or can’t engage deeply with verification. The daemon in your pocket doesn’t require you to read philosophy or understand Tarski’s semantic theory of truth. It just helps you figure out whether the video you’re watching has been verified, who’s making claims about it, what the track record of those sources is, and what questions you might want to ask before forming a belief.

Most people will never be epistemological sophisticates. That’s fine. That’s always been true. What’s changed is that the information environment has become adversarial in ways that overwhelm even sophisticated thinkers. The junk data is winning because it’s optimized for engagement while truth is optimized for accuracy. In that arms race, truth loses without assistance.

The daemon tips the scales back. Not by imposing beliefs, but by reducing the cognitive load required to engage with reality responsibly. It’s a prosthetic for epistemic health in an environment designed to make you sick.

The Cyberpunk Precipice

Here’s where we are: heading toward corporate hellworld unless we coordinate our efforts. The tech billionaire caste understands that AI is the decisive technology of the next century. They’re building systems optimized for extraction and control. They’re not interested in democratizing the filter because ambiguity serves their interests. They want you confused and exhausted, scrolling through an infinite feed of contradictory information, too depleted to organize resistance.

The Culture Victory model I’ve been developing is a counter-strategy. You don’t defeat them in direct confrontation. You build something better and make their system irrelevant. In the context of epistemology, that means trusted daemons that help individuals navigate without requiring centralized control. It means federated verification infrastructure that can authenticate media without relying on captured institutions. It means parallel communication networks that don’t depend on platforms optimized for outrage.

None of this is technically difficult. The engineering is solved. What we lack is coordination and will.

The Metatextual Joke

There’s a layer of irony here that I need to acknowledge. My handle, Shalashashka, is lifted from Revolver Ocelot’s alias in the Metal Gear series. I’ve been signaling my allegiance for years without fully understanding what it meant.

When I chose that name, I thought I was identifying with Ocelot’s cleverness, his ability to play multiple sides, his refusal to be captured by any single ideology. I didn’t realize I was also identifying with his fundamental project: the recognition that you can’t defeat the Patriots by pretending the problem they identified doesn’t exist. You can only defeat them by building something better.

Snake’s skepticism was noble but ultimately insufficient. He could destroy individual threats but he couldn’t build alternatives. His heroism was reactive rather than constructive. In a world where the crisis is systemic, individual heroism doesn’t scale.

Ocelot understood that you have to play the long game. You have to get your hands dirty with infrastructure. You have to build the systems that will outlast you, even if building them requires compromises that would have horrified your younger self.

The people playing Metal Gear Solid 2 side with Raiden and Snake by default because we’re trained to identify with the protagonists. But Kojima was smarter than that. He made Raiden annoying on purpose. He made the Patriots’ argument compelling on purpose. He ended the game ambiguously on purpose.

He was asking us to think harder than we wanted to.

Conclusion: Build the Daemon Anyway

Let me be concrete about what I’m proposing, because abstraction is where accountability dies.

The Patriots offered: centralized control, imposed filtering, managed subjects, top-down truth decided by AI systems humans couldn’t audit or influence. This model failed not because filtering is wrong but because centralization is fragile and authoritarianism is corrosive.

Ocelot’s alternative, and mine: decentralized tools, accessible filtering, equipped citizens, bottom-up truth-seeking assisted by AI systems humans can audit and improve. The daemon model. Not a warden but a tutor. Not a censor but a librarian. Not a replacement for human judgment but a prosthetic for human cognition in an environment that’s been weaponized against it.

The practical implementation looks like this: open source language models trained for epistemic assistance rather than engagement maximization. Federated verification networks using cryptographic provenance to trace media origins. Communication protocols that prioritize signal over noise. Personal AI companions that help you ask better questions rather than feeding you predetermined answers.

Metal Gear Solid 2 predicted the death of shared reality with disturbing accuracy. It showed us the ontological crash before we had words for it. It warned us that the trivial would drown out the important, that context would collapse, that we would retreat into gated communities of belief leaking whatever truth suited us into a growing cesspool.

We didn’t listen. We thought we were on Snake’s side, fighting for freedom against authoritarian control. We didn’t realize that the freedom we were fighting for was the freedom to drown.

The Patriots were right about the problem. They were wrong about who should control the solution. The answer isn’t centralized authoritarian filtering. But the answer also isn’t unfiltered chaos. The answer is what Ocelot was working toward all along: distributed tools for distributed agents, context that’s accessible rather than imposed, infrastructure that empowers rather than manages.

I don’t know if we have time. The window for building parallel infrastructure before complete collapse may already be closing. But the alternative is accepting the current trajectory. Accepting that the future belongs to those willing to weaponize confusion. Accepting that the species will choose tribal epistemology over shared reality.

I can’t accept that. So we build. We build trusted daemons and federated verification and parallel communication networks. We build the infrastructure of the future inside the shell of the present.

The Patriots were right that we needed help. Ocelot was right about what kind of help we needed. Let’s build it.

Culture Victory

Discussion about this post

Ready for more?