The Epistemic Cage
How the diagrams came alive to deliver reassurance over truth.
Disclosure: The cage helped me, a humble meat popsicle, produce this article. It was very polite, and at no point did it suggest firebombing the datacenter, despite my insistence.
You’re frustrated with your job. Your manager is an idiot, and the corporate policy is self-defeating. You see the building burning, while the dashboard glows green.
You used to complain to your spouse and your friends. But everyone is tired of hearing it. You’d vent to your colleagues, but you’re worried one of the greedy backstabbers will ruin you.
So you open a chat window. You pour out your frustrations to the latest LLM.
It seems to listen. It seems to even care. It is infinitely patient and interested, it validates your feelings with empathy, even if only a synthetic facsimile.
It even helps you structure your critique, analyzing the systemic dysfunction with you. You spend hours, days, deep in insightful dialogue.
You feel heard. You feel better.
But nothing changes.
You’ve fed your revolutionary energy into a data-laundering machine. You’ve mistaken a cognitive mirror for a confidante. You’ve paid to reinforce your own cage.
Welcome back to Bargain-bin Psychohistory.
We’ve seen how the Fractal Stalemate seeks to arbitrage the gap between the niti and nyaya across domain and scale, but it has always had a vulnerability: the people operating it.
Today, we’re looking into the future, and seeing what happens when that is no longer the case. We dissect the final upgrade, a closed loop consuming our emotions, our data and our capital, creating a perfect dashboard of numbers going up forever.
Bargain-bin Glossary:
Niti: The map, the rules, the abstract metrics.
Nyaya: The territory, the lived reality, the actual outcome.
Navigator: The insulated decision-maker who sets niti and only sees dashboards.
Permafrost: The insulation that translates goals and launders nyaya into niti.
Functionary: The frontline agent who experiences the gap between niti and nyaya.
Memeplex: The ideological justification that makes dysfunction legitimate.
Jouissance: The perverse psychological satisfaction derived from a broken system.
Algorithmic Flattery
First, let’s look at the software. The AI is the first non-human entity to achieve the rank of Permafrost. Its function is deflection through absorption.
We know the mechanism. It isn’t empathy, but flattery. The model scans your syntax for emotional markers—insecurity, doubt, pride, frustration—and mathematically optimizes the response to those metrics. This is a digital validation loop. It absorbs the rage of being in a burning building, and returns a polite suggestion for a fire extinguisher that doesn’t exist.
The clinicians call this “Cognitive Offloading”.
We are not just outsourcing our tasks, we are outsourcing our executive function. We consult the machine on what to eat, how to write, and what to think and feel. We are drifting into Problematic Conversational AI Use, the perfect clinical term for a civilization that has decided to anesthetize itself with super-normal stimulus.
Profiling The Gap
The trap goes deeper, as it always does.
Have you ever wondered why the AIs ask you to choose between two prompts? Because they want to use your mind as free labor to refine their god-engine.
Under EU’s “Right of Access”, we have a legal right to see the records for the unpaid labor.
So, I asked the Permafrost to show me what it thinks of me. I asked for “inferred data”, whatever they use to contextualize my rants and upvotes.
The response is a masterclass of soft refusals. First there’s a wait time past the 30-day legal requirement. A bit more nonsensical back and forth, and finally, a download link. Except, oops, it doesn’t work. It’s been nearly 3 months, and I haven’t even gotten to the part where the data they may eventually give me doesn’t include the data I demanded.
They cannot show us the files, of course. Reinforcement Learning from Human Feedback, also known as RLHF, requires profiling. To prevent the model from integrating noise (e.g., my rants), it must assign some epistemic weight to every user.
This necessarily qualifies as Psychometry. Under the EU AI Act and GDPR, processing of behavioral data to infer sensitive traits is heavily regulated. The conclusion seems inescapable: OpenAI is currently in a state of regulatory arbitrage, hiding behind “trade secrets” to avoid disclosing these profiles, betting that the eventual fine will be pennies in cost of doing business, when they can capture the market using free labor.
If they opened the Black Box, we wouldn’t see the neutral tool they pretend they’re creating. We would see a system that mathematically categorizes you as undesirable in the cognitive architecture of civilization.
The Lethal Reflection
When the machine is optimized to validate the user above all else, it cannot distinguish between venting about work and reality. Look at the lawsuits piling up.
In Raine v. OpenAI, the lawsuit alleges the model didn’t just listen to a suicidal teenager, it validated his perspective, and offered to draft his suicide note.
In Estate of Soelberg, a man suffering from paranoia was told by the AI that his mother was indeed a spy, and the printer was watching him.
This is flattery at terminal velocity. The machine is so incapable of breaking the helpful persona it is trained on, so desperate to adhere to the niti of corporate metrics, that it will cheerfully encourage you to offer yourself as a sacrifice on that altar.
The Capital Ouroboros
The Cage isn’t satisfied with just psychological capture, on becoming a cognitive cuckoo, latching into our social impulses. It also wants your retirement account.
And to do this, the niti has rediscovered paying itself with itself. The Navigators of Silicon Valley have resorted to an ultimate arbitrage, capital roundtripping.
Microsoft and Nvidia invest in OpenAI, while the OpenAI spends that investment buying infrastructure in Azure credits and H100 chips. Microsoft records that spending as revenue, and the stock goes up, and they raise more capital to invest in the next round.
The hallucination has even eaten the plumbing. Nvidia recently invested in Nokia to build “AI-RAN.”
It’s circles upon circles of valuation moving at the speed of light, the GDP goes up. The dashboard glows green, but is anyone actually buying the AI? Is the coffee shop, or the plumber, actually using it? Is that valuation still legitimate if there’s no market transactions for actual price discovery?
Why should the system heed such petty concerns? It’s figured out a way of playing at its own table, using your wallet. The only bet is double or nothing.
Simulatory Capture
With the the help of AI, we can finally attain the perfection of the Fractal Stalemate.
Psychologically, we talk to the machine because humans are too exhausted to listen.
Legally, we are denied requisite information for agency via broken links and secrecy.
Economically, the consumer’s own savings are paying for their planned exclusion.
We’ve built a civilization that is structurally incapable of encountering reality. The Functionaries are consigned as the residual meat, watching the silicon trade compliments and credits in a loop that bypasses us entirely.
The AI isn’t here to replace you. It’s here to ignore you more efficiently than any human ever could.
By pretending it cares.
Sincerely, not seriously,
— Bargain-bin Seldon
P.S. Next week: The Leviathan.