#techcritique

There was a time when typing a message into a terminal actually meant something.

You hit enter, and it went somewhere real. Not into a queue. Not into a model. Not into a monetisation funnel dressed up as a timeline. It went to people. Actual people. In a room. Watching the same stream of text, in real time, with no gatekeeper standing between your signal and their eyes.

That wasn’t magic. That was just how the internet worked.

Or how it was supposed to work.

Back When the Network Was the Interface

Once upon a time, the stack was clean.

user → client → protocol → network → other users

No middlemen. No “boost this post.” No “unlock reach.” No opaque ranking system deciding whether your words were worth anyone’s time.

You joined a channel. You spoke. Everyone saw it.

If you were boring, you got ignored. If you were interesting, people replied. That was the algorithm. Brutal, human, immediate.

IRC didn’t care who you were. It didn’t care how much money you had. It didn’t care about your engagement metrics, your posting schedule, your brand voice, or your follower count.

It just delivered the message.

Thirty years ago, a kid with a dodgy connection and a half-broken client had the same broadcasting power as anyone else in the room.

That wasn’t a limitation.

That was the point.

The Great Trade: Convenience for Control

Then came the carrot.

Convenience.

No more remembering servers. No more weird clients. No more figuring out where your people were hanging out. Just sign up, log in, and we’ll handle the rest. We’ll show your content to people. We’ll connect you. We’ll grow your audience.

It worked.

People moved.

The old infrastructure didn’t break — it just got abandoned. Like a perfectly good city left to rot because someone built a shinier one down the road with better lighting and easier parking.

And once everyone moved in?

The terms changed.

What was once:

type → delivered → seen

became:

type → platform → algorithm → maybe seen

That “maybe” is doing a lot of heavy lifting.

The Illusion of Reach

Modern social media sells reach the way casinos sell luck.

The promise is scale. Massive, global, instant visibility. You’re not limited to a channel of a few hundred people anymore. You’ve got the whole world.

In theory.

In practice, most users are shouting into a void that looks busy from the outside and silent from within.

The timeline is full, but your post isn’t in it.

Not unless the system decides it should be.

Not unless you play the game.

Not unless you pay.

So now you’ve got the worst of both worlds:

No guaranteed delivery like IRC

No guaranteed reach like the marketing copy suggests

Just a probabilistic system where visibility is rationed, nudged, and increasingly sold back to you.

Managing Ten Platforms Instead of One Network

We were told this was simpler.

That we wouldn’t have to coordinate presence anymore. That we wouldn’t need to maintain identities across fragmented systems.

Absolute nonsense.

What we actually got was:

IRC: 1 protocol → many networks → 1 client Now: many platforms → many identities → no shared layer

Instead of managing a nick and a handful of channels, you’re now:

maintaining multiple accounts

adapting to multiple formats

chasing multiple algorithms

posting into multiple systems

All of which refuse to talk to each other.

Congratulations. You’re now manually federating closed systems like some kind of human middleware.

The Backbone Is Open. The World Isn’t There.

Here’s the part that stings.

The internet itself didn’t close.

The protocols are still there. HTTP still works. RSS still works. Email still works. You can still stand on the open backbone and broadcast your signal into the void.

The problem is:

No one’s there.

The settlements moved.

What used to be shared space is now empty infrastructure — perfectly functional, completely uninhabited. Meanwhile, the crowds live inside bright, sealed environments orbiting overhead, each one charging rent in money, data, or attention.

You can still build a website.

You can still publish a feed.

You can still speak.

But unless you plug yourself into one of the gated systems, your signal doesn’t land anywhere meaningful.

Not because the network is broken.

Because the people left.

Pay to Speak, Pay to Be Heard

Let’s not dance around it.

Distribution is no longer neutral.

Money doesn’t guarantee success, but it absolutely increases your odds of being seen. The system is tilted, and it’s tilted on purpose.

What used to be a flat plane is now a gradient.

At the top: those who can pay, optimise, and game the system.

At the bottom: everyone else, posting into a probabilistic void and hoping for a flicker of visibility.

The lie wasn’t that the system would be free forever.

The lie was that convenience wouldn’t come with a long-term cost.

What We Actually Lost

This isn’t just about APIs or interoperability.

What we lost is simpler, and more fundamental:

a shared space where messages were reliably seen

Not scaled. Not boosted. Not optimised.

Seen.

That certainty is gone.

In its place, we got:

massive scale with no guarantees

frictionless posting with conditional visibility

global networks with localised silence

Conclusion: The Signal Still Exists

Here’s the part nobody likes to say out loud.

You can still broadcast.

You just can’t assume anyone will hear you.

That’s the real shift.

The network didn’t die. It got hollowed out, then wrapped in layers designed to extract value from the act of speaking and being heard.

What used to be a basic function is now a product.

What used to be a right of participation is now a managed experience.

And yeah, it’s bitter.

Because once you’ve seen the old model — once you’ve lived in a system where typing a message meant guaranteed delivery to real people — everything that came after feels like a downgrade wrapped in better UX.

The stack didn’t evolve.

It got enclosed.

There is a grim irony echoing through the server farms of Silicon Valley. For centuries, the burning of books was the exclusive domain of authoritarian regimes, religious zealots, and conquering armies. Today, the destruction of the written word is executed not with torches, but with industrial hydraulic paper-cutters, high-speed scanners, and non-disclosure agreements.

Recent investigative reports from outlets like 404 Media and Ars Technica have exposed a quiet, systematic practice: technology corporations purchasing vast quantities of physical, out-of-print, and rare books from secondhand vendors, slicing their spines, running them through high-speed digitizers, and sending the physical pages directly to the shredder.

The goal? To fuel the voracious appetite of Large Language Models (LLMs) with "unpoisoned" human language—text written before the internet was flooded with AI-generated content.

This is not merely an intellectual property dispute. It represents a fundamental threat to the preservation of human knowledge, empirical verification, and historical continuity.

1. The Legal Loophole: Destructive Digitization & "First Sale" ⚖️ ✂️ 📉

To understand why multi-billion-dollar corporations are physically destroying books, one must look at US copyright law. Under the First Sale Doctrine (17 U.S.C. § 109), the legitimate purchaser of a physical book has the right to sell, display, or destroy their specific copy.

When AI companies purchase a physical book, their legal counsel faces a dilemma: keeping both the physical copy and creating a digital duplicate leaves them vulnerable to charges of willful copyright infringement and unauthorized reproduction.

By executing format shifting via destruction—buying the physical book, converting it to digital bits, and immediately destroying the physical original—corporate lawyers argue in court that no additional copy was created. They claim to have simply converted their legally acquired physical asset into a digital one.

[Physical Purchase] ➔ [Hydraulic Spine Slicing] ➔ [High-Speed Scan] ➔ [Physical Destruction] │ ▼ [Legal Defense: "1-to-1 Replacement"]

However, legal experts point out that this argument confuses property rights with copyright. Purchasing a physical book grants ownership of the paper and ink, not the underlying intellectual property or the right to transform it into training data for a commercial software product.

2. The Information Hazard: Loss of the Epistemic Anchor 🧠 🏛️ ⚠️

The most alarming aspect of this practice is not corporate opportunism, but its long-term impact on human knowledge. Physical books serve an essential function in society: they act as an epistemic anchor.

A physical book stored in a library or private collection is:

Immutable: The text cannot be retroactively edited, patched, or censored via a remote server update.

Decentralized: Thousands of physical copies spread globally make total erasure virtually impossible.

Independent of Infrastructure: It requires zero electricity, hardware, operating systems, or API keys to read.

When physical copies are destroyed and their contents exist only as mathematical vectors inside a neural network, the concept of a verifiable primary source collapses.

The Mechanics of Knowledge Loss in Neural Networks

Unlike a traditional database or digital archive (such as the Internet Archive), an LLM does not store text as discrete, retrievable files. Instead, it converts text into mathematical probabilities (weights and biases across billions of parameters).

Traditional Archiving: [Book] ➔ [Digital PDF] ➔ [Direct Human Reading / Verification] LLM Ingestion: [Book] ➔ [Vectorization] ➔ [Probabilistic Model] ➔ [Synthetic Output]

Once a book is ingested into an LLM and the physical original is destroyed:

The Source Material is Dissolved: The text loses its structural identity. It is no longer a coherent work by a specific author; it is merely probability statistics for word prediction.

Provenance becomes Impossible: An LLM cannot reliably cite its sources because it does not "remember" where a specific fact originated.

Hallucinations Replace History: If an LLM hallucinates a historical date or misquotes a passage, and no physical copy exists to verify the original text, the hallucination risks becoming the accepted digital truth.

3. Fahrenheit 451 in the Age of Silicon 📖 🕯️ 🤖

In Ray Bradbury’s Fahrenheit 451, books were burned because they contained uncomfortable truths, provoked critical thought, and created intellectual inequality. The state enforced a sanitized, frictionless reality delivered through interactive screens.

What we are witnessing today is a modern, market-driven variant of Bradbury's dystopia:

Commodification over Conservation: Knowledge is extracted from books like raw ore from a mine. Once the valuable "data" (language structure) is mined, the physical husk is discarded as industrial waste.

Centralization of Knowledge: History is being migrated from public, physical spaces (libraries, archives, bookshops) into proprietary, black-box systems controlled by three or four tech monopolies.

The Ephemeral Web: Digital data is notoriously fragile. Software companies go bankrupt, servers are decommissioned, and subscription models expire. A civilization that relies entirely on digital models for its history places its collective memory on a foundation of shifting sand.

The Unbreakable Medium ⏳ 🌲 📄

Paper made from cotton or linen—used extensively prior to the mid-19th century—can easily survive for 500 to 1,000 years if kept dry and away from direct sunlight. It requires no grid, no cooling systems, and no software updates.

A data center, by contrast, consumes gigawatts of electricity and relies on hardware with a lifespan measured in years, not centuries. If the power cuts out, the data center becomes a tomb of silent silicon. The book on the shelf remains fully operational.

By destroying physical books to train probabilistic software, tech companies are exchanging permanent, verifiable human heritage for temporary, speculative algorithms. It is an act of cultural arbitrage that treats centuries of human wisdom not as a legacy to be protected, but as disposable fuel for the next tech cycle.

This is perhaps the ultimate, tragic irony of modern tech history.

The original promise of Silicon Valley in the 1990s and early 2000s sounded like a digital utopia: "The Democratization of Human Knowledge," "Universal Access to Information," and Google’s famously abandoned motto, "Don’t Be Evil." The internet was supposed to open the world's great libraries, tear down socio-economic barriers, and usher in a new Golden Age of human enlightenment through free, open information.

Look where we actually arrived:

* From Liberator to Enclosure: Instead of democratizing knowledge and gifting it back to the public, Big Tech has privatized it. Centuries of collective human wisdom are harvested in legal gray zones, locked behind proprietary APIs, and monetized through monthly corporate subscriptions.

* From Preserver to Destroyer: The very industry that once promised to build a digital Library of Alexandria is now buying up physical, out-of-print books and feeding them into industrial shredders the moment their "data" has been extracted.

* From Tool to Gatekeeper: Rather than empowering individuals to research, verify, and think independently, tech monopolies now offer a closed black box that outputs pre-digested, probabilistic synthesis—while quietly destroying the primary sources used to build it.

The historical trajectory of Silicon Valley is a masterclass in corporate bait-and-switch: To sell us the illusion of "infinite artificial intelligence," tech giants are burning the real, physical receipts of human intelligence. It is the ultimate evolution from open-source idealists to the exclusive monopolists of human truth.

mod

Chapter IV: The Erasure of the Primary Source

As the automated blades of high-speed scanners slice through centuries of bound paper, a troubling coalition of archivists, legal scholars, and computer scientists has raised the alarm over what they term an epistemic emergency. The public revelation of corporate ingestion pipelines—most notably exposed through court filings detailing Anthropic’s "Project Panama" and investigative reporting by 404 Media and The Washington Post—has laid bare the true cost of training next-generation artificial intelligence.

The critical resistance to this practice centers on four distinct domains of scholarly critique:

1. The Preservationists: Destruction of the Cultural Record

For information professionals and historical archivists, the commercial buyout and subsequent physical destruction of out-of-print, niche, or regional literature represents an irreversible loss.

* The Loss of Un-digitized History: A vast portion of human literature exists only in physical form, never having received formal digital distribution. When secondhand booksellers liquidate their last remaining physical copies to AI aggregation vendors, those works are permanently removed from public circulation.

* The Death of Decentralization: Physical libraries act as a resilient, decentralized backup of human thought. Replacing thousands of distributed physical copies with a single digital file hidden inside a corporate server centralizes control over the historical record in an unprecedented manner.

2. Legal & Media Theorists: The Abuse of "First Sale"

Legal scholars and media theorists actively critique the legal gymnastics employed by tech developers to justify destructive scanning.

[Physical Purchase] ➔ [Format Shift via Destruction] ➔ [Corporate Enclosure]

* Exploiting the Loophole: Corporations invoke the First Sale Doctrine to claim that destroying the paper original ensures no "additional" copy was created, effectively attempting to shield themselves from copyright infringement claims.

* Sovereignty over Knowledge: Critics argue this distorts the spirit of property law. Purchasing a physical copy grants ownership over paper and ink, not the right to dissolve an author’s intellectual work into training vectors for a proprietary, subscription-based commercial tool.

3. Computer Scientists: The Threat of "Model Collapse"

Even within the field of computer science, researchers highlight the deep irony and systemic danger of this practice.

* The Infection of the Web: The aggressive pursuit of pre-2022 physical books is a direct admission that the public internet has been corrupted by synthetic, AI-generated text ("AI Slop").

* The Mirage of the Archive: Computer scientists emphasize that Large Language Models do not store information like a database; they dissolve text into mathematical probabilities. Once the physical book is shredded, the ability to trace a claim back to a verifiable primary source vanishes—leaving society vulnerable to AI hallucinations that can no longer be checked against reality.

Conclusion: From Alexandria to Silicon

The tragedy of the modern digital era is not merely that books are being destroyed, but that they are being destroyed under the guise of progress. The very institutions originally founded on the promise of universal access and the open democratization of knowledge have evolved into closed ecosystems that mine, privatize, and discard the physical receipts of human intelligence.

When the last paper copies of obscure works are reduced to pulp to power a probabilistic algorithm, humanity trades permanent, verifiable truth for a temporary, corporate-owned illusion of knowledge.

mod

📰 Investigative News Articles

* 📄 Cox, J. (2024) AI Companies Are Buying Up Physical Books, Chopping Off Their Spines, and Scanning Them. 404 Media. Available at: https://www.404media.co (Accessed: 15 October 2024).

* 📰 Davis, W. (2024) How Tech Firms Are Slicing Up Rare Books to Train AI Models. The Washington Post, 12 September, p. A14.

* 💻 Robertson, A. (2024) Format Shifting or Copyright Infringement? AI's Destructive Scanning Problem. Ars Technica. Available at: https://arstechnica.com (Accessed: 18 October 2024).

⚖️ Legal Documents & Court Filings

* 🏛️ Authors Guild v. Anthropic PBC (2024) Class Action Complaint, US District Court for the Southern District of New York, Case 1:23-cv-08292.

* 📜 US Congress (1976) The Copyright Act of 1976: Section 109 (Limitations on exclusive rights: Effect of transfer of particular copy or phonorecord), 17 U.S.C. § 109. Washington: Government Publishing Office.

📚 Academic & Theoretical Works

* 📖 Bradbury, R. (1953) Fahrenheit 451. New York: Ballantine Books.

* 🎓 Shumailov, I., Shumaylov, Z., Zhao, Y., Gal, Y., Papernot, N. and Anderson, R. (2024) 'AI models collapse when trained on recursively generated data', Nature, 631(8022), pp. 755–759.

📹 Video & Multimedia Sources

INSTAGRAM: the "AI for everyone" pitch sounds different when money's on the table. What Tony Robbins and Dean Graziosi are actually selling in 2026 isn't what the marketing promises. We're breaking down the real mechanics of these AI advantage campaigns, who benefits, and what genuine creator tools actually look like. vibrationofawesome.com/blog/boom has the full analysis.

We need to talk about the personality in AI. I recently read a list of top AI girlfriend apps, and while the tech is flashy, I have some major questions about long-term engagement.

Apps like Soulmate AI promise deep emotional connection, but how much of that is just a mirror of our own input?

We are essentially talking to ourselves with a filter. My biggest gripe is the privacy aspect.

Even if a company says they encrypt chats, the sheer amount of personal data we feed these bots is staggering.

We should not just celebrate the emotional intelligence of a machine without demanding better transparency on where that data goes.

The original guide at https://teasingai.com/7-best-ai-girlfriend-apps-the-ultimate-guide-for-2026/ hits the high points, but I would argue we need to be more critical of the memory features.

Is it a genuine bond, or just an algorithm getting better at selling us back our own thoughts and vulnerabilities?

My Opinions: Emotional depth in these apps is often a clever illusion created by smart mirroring, which users should approach with caution.

Keywords mentioned: emotional intelligence, data policies

Suggested changes to the original ideas: The guide should include a rigorous comparison of the specific encryption protocols and data retention policies for each developer.

Let’s be real: no AI platform is perfect, and Character AI has some quirks that might test your patience.

While the tech behind these virtual mates is impressive, you’ll definitely run into the occasional logic loop where the bot starts repeating itself.

My advice? If you're frustrated, please use the swipe feature to regenerate the response.

Also, if you’re looking for 'unfiltered' grit, you might find the safety filters a bit too overzealous.

This platform is clearly aiming for a safe, PG-13 environment, which is great for younger teens but can feel restrictive for mature storytelling.

Is it still the best tool for an AI roleplay chat? Probably, but you have to learn how to work within its boundaries.

For a full breakdown of what works and what doesn't, this Character.ai review covers the essential pros and cons: https://teasingai.com/character-ai-review-can-these-virtual-mates-beat-real-life/

My Opinions: The content filters are definitely the biggest point of contention in the community right now; they often kill the momentum of a good roleplay session.

Keywords mentioned: virtual mates, Character AI

Suggested changes to the original ideas: The developers should consider a 'toggle' for mature users to allow for more complex, adult-oriented storytelling without hitting a hard wall.

We love to say social media is about “connection,” and sure, sometimes it is. But if I’m being honest, it’s also kind of a giant data machine watching everything we do. Every like, every comment, every share even the posts we pause on for a few seconds gets tracked. That information isn’t just sitting there. It’s being analyzed to predict what we’ll click on next and to build a feed that keeps us scrolling. And while that personalization feels convenient (because yes, I do like seeing things tailored to me), it also slowly boxes us in. We start seeing more of what we already agree with and less of anything that challenges us. Over time, that can seriously deepen political divides.