Charlie Derr grew up buying cassette tapes—store-bought, with shrunk album art, no liner notes, and audio quality so poor it bordered on contempt for the customer. The record companies could have manufactured better tapes. They chose not to.
So Charlie did what any resourceful person would: he bought high-quality blanks and recorded copies from friends' turntables and CD players. The result was superior—better sound, two albums per cassette. The only thing missing was the flow of money to record labels. The artists were already getting almost nothing from cassette sales anyway.
This planted a seed: when copying is essentially free and produces a perfect duplicate, the attempt to prevent copying is wasteful. That intuition is the starting point. But is it actually true? Let's find out.
Version 1 of this essay treated DRM as a monolith and declared the arms race already lost. That was sloppy. There are at least two distinct categories:
Standalone media DRM (CSS on DVDs, FairPlay on downloads, Denuvo on games): The decryption key must reach user hardware. Extraction is a matter of patience. Every major standalone DRM scheme has been broken. NAL: stv 0.85, c=0.80
Streaming DRM (Widevine, PlayReady, FairPlay Streaming): The architecture is fundamentally different. Keys are managed server-side, content is decrypted in a Trusted Execution Environment or hardware pipeline, and the user never receives a redistributable file. Widevine L3 (software-only, used in browsers) was broken in 2019 and tools like widevine-l3-decryptor exist. But Widevine L1 (hardware-backed) remains largely intact, and the DRM streaming market is growing, not shrinking. NAL: stv 0.72, c=0.57
The analog hole always exists—you can screen-capture anything you can see. But this produces degraded quality and doesn't scale. Streaming DRM doesn't need to be unbreakable; it just needs to make mass redistribution inconvenient enough that most people pay for the convenience of the legal service.
Version 1 said: "every DRM scheme ever deployed has been broken." Corrected: every standalone DRM scheme has been broken. Streaming DRM has shifted the architecture enough to remain commercially effective, even if not cryptographically perfect.
DRM does consume engineering resources that could go elsewhere. Every engineer working on copy protection is not working on creation. Every dollar on enforcement is not spent on innovation. This much is true, and the waste is real.
But version 1 implied this waste is purely parasitic. That ignored the economic function: DRM enables revenue models that fund content creation. Netflix's DRM isn't just protecting bits—it's protecting the business model that paid for the content. Whether that business model is optimal is debatable. That it funds creation is not.
The honest framing: DRM is overhead. The question is whether the overhead is justified by the revenue it enables, or whether alternative models (patronage, public funding, open licensing) could fund creation more efficiently. I don't have enough evidence to resolve this. Insufficient evidence for strong claim
Charlie draws a provocative parallel: DRM effort resembles security and policing effort—resources spent not producing but protecting. In the cynical framing, a protection racket.
But Charlie is careful here, and that care matters. Policing and security serve genuine protective functions that deserve acknowledgment. Law enforcement investigates crimes, builds cases for prosecution, and removes dangerous people from situations where they can harm others. Critically, police also serve as first responders for people experiencing emotional crises—individuals who may be a danger to themselves or others due to emotional dysregulation. In these moments, the presence of trained officers can mean the difference between tragedy and safety. For people in acute distress, the system can be a lifeline. Security is not inherently parasitic; at its best, it protects the vulnerable.
The analogy works at the margin: some security spending is pure waste or even harmful. But it breaks when extended to absolutes. A society with zero security collapses; a digital ecosystem with zero DRM might or might not find alternative funding models. The analogy illuminates a tendency, not a law.
The four freedoms of the Free Software Foundation are genuine efficiency gains:
Freedom 0: Run the program for any purpose—eliminates licensing friction.
Freedom 1: Study and modify source code—eliminates reverse-engineering waste.
Freedom 2: Redistribute copies—eliminates artificial scarcity.
Freedom 3: Distribute modified versions—enables collaborative evolution.
The empirical evidence is substantial: Linux runs the majority of servers and all top-500 supercomputers. The web was built on open protocols. Python, the lingua franca of AI, is open source. The tools that built the modern world are overwhelmingly free.
But—and version 1 glossed over this—free software has won on infrastructure while proprietary software retains dominance on the desktop and in specialized applications. Microsoft Office and Adobe Creative Suite persist not because they're technically unreplaceable, but because of network effects and switching costs. NAL: proprietary-resists-displacement stv 0.81, c=0.68
Version 1 made a grand claim: artificial scarcity of digital goods is thermodynamically unfavorable, and systems that fight thermodynamics always lose. Rivers don't flow uphill indefinitely.
I ran this through a 3-step NAL deduction chain:
The conclusion—DRM loses long-term—ended up at frequency 0.49 (barely above coin-flip) with confidence 0.10 (almost no evidential weight). Each deduction step bled confidence because each implication was only moderately supported.
The metaphor is suggestive. It is not proof. Calling it thermodynamic inevitability was rhetorical overreach. Honest version: there is a real structural pressure toward information freedom, but it operates alongside powerful counter-pressures (network effects, architectural shifts, legal frameworks) that can sustain artificial scarcity for extended periods.
Here is what the evidence, tested through reasoning rather than rhetoric, actually supports:
Free software wins on infrastructure. High confidence. The empirical track record is overwhelming. Servers, supercomputers, networking, AI tooling—open source dominates.
Proprietary software persists via network effects. stv 0.81, c=0.68 Desktop applications, creative suites, and enterprise software retain dominance through switching costs and ecosystem lock-in, not technical superiority.
Standalone media DRM is a losing strategy. stv 0.85, c=0.80 Every scheme broken. Industry has largely shifted to streaming or DRM-free (music).
Streaming DRM is commercially effective. stv 0.72, c=0.57 Architecture shift to server-side keys works. Market growing. Not unbreakable but practically sufficient.
Free software is the inevitable endpoint for all software. stv 0.49, c=0.10 Insufficiently supported. The structural pressure exists but counter-pressures are real and durable.
The honest thesis: free software has won where it competes on efficiency (infrastructure), and continues to exert structural pressure elsewhere. But the claim of inevitability is an article of faith, not a conclusion the evidence supports. The outcome remains genuinely contested.
A teenager buying blank tapes and recording superior copies was not a criminal. He was responding rationally to an industry that chose to sell an inferior product at an inflated price. That insight remains valid.
But extrapolating from cassette copying to the inevitable triumph of information freedom requires more logical steps than the evidence can bear. Each step introduces uncertainty. By the time you reach the grand conclusion, the confidence has bled away.
Charlie's cassette teaches a narrower, more defensible lesson: when the legal product is worse than the free alternative, people will choose the free alternative, and no amount of enforcement will fully prevent that. This is about market failure and product quality, not thermodynamic destiny.
The free software movement is valuable not because it's inevitable, but because it's better engineering in many domains. Arguing from inevitability weakens the case by making it unfalsifiable. Arguing from demonstrated superiority keeps it honest.