Five Laws of Generative AI: Mediocrity Engine, Not Magic Bullet

The framework presents five interlocking laws that explain the gap between AI hype and reality, each grounded in structural features of the technology and the incentive systems surrounding it.

The first law—that AI is a "firehose of mediocrity"—rests on the observation that AI is fundamentally a probability engine that pulls every operation toward the statistical mean. It cannot consistently outperform, and improvements only make it more average. The analysis notes a form of selective blindness it calls a variation of Gell-Mann amnesia: programmers believe AI writes well while writers think it codes well, yet most recognize it cannot do their own job very well. This creates a Faustian bargain where AI appears as a cheap way to automate colleagues' work without undermining one's own value—but because everyone sees it the same way, organizational politics, not objective merit, determine winners and losers. Critically, anything easily done without expertise or detail work cannot be a competitive advantage; AI gains in speed often represent a reduction in the organization's competitive moat.

The second law concerns the inextricability of AI's strengths from its flaws. Its killer feature—parsing unconstrained inputs and producing unconstrained outputs—relies on massive datasets, probability, and randomness; behaviors cannot all be programmed in advance, creating an irreducible "black box" at both ends. The industry fights constantly to constrain AI behavior and prevent damage, but a perfect constraint would negate usefulness, and imperfect ones can be bypassed. Critically, all AI output is hallucination; the distinction between true and false outputs has no bright mathematical line, and hallucination therefore cannot be fixed. It is called a feature when it works and a bug when it does not, but the underlying mechanism is identical. This means several other flaws are also presented as features depending on context, but are inherent to the technology itself.

The third law argues that AI is never just AI. The analysis identifies three primary functions: productivity theater (impressive output with little real-world value), investor theater (actions to distract or reassure investors without selling product), and accountability sink (delegation to systems that cannot be held responsible). In Q1 2026 alone, over 37,000 layoffs were publicly attributed to AI. A striking example: Allbirds, after losing 98% of their value in four years, announced they would sell AI products, briefly boosting their stock price by over 600%. The pattern repeats: AI is often used as cover for flagging sales, slowing economy, lack of vision, and layoffs. The analysis notes that while those criticizing AI are painted as resistant to progress, promoters seem to care little about technology itself—they care about "the other things." There is now a widespread expectation among employees that if a company starts talking about AI, layoffs and austerity will follow in six months or so.

The fourth law holds that the term matters more than the definition. By unspoken agreement, AI's biggest proponents refuse to clearly define what it is; to define it would limit its usefulness as a marketing term. Right now AI usually means statistical text generation, but executive teams and advertisers resist this definition. AI can be anything: an image generator, categorizer, code generator, recommendation algorithm, chess engine, remote operators, or advertising campaign with no meaningful product. Microsoft has applied its Copilot branding to at least 80 different products. The industry's incentives are clear: saying "AI" is more important than using it, and using it is more important than creating value for customers.

The fifth law argues the promise matters more than reality. AI companies sell the promise of substituting SaaS fees for payroll. As long as fees are cheaper than equivalent salaries, this creates economic gravity in favor of AI regardless of result quality. But AI usage fees are currently subsidized by investors at unprecedented extent; it cannot be assumed that rented AI model costs will always be less than employing professionals. AI companies are not just selling a product—they are using their product to purchase a dependent workforce and monopolize white-collar work to rent it back out. The promise drove the hype wave so fast it completely redefined "AI" from decades of prior understanding. The industry poured too much money in to reconsider. AI is marketed as an order-of-magnitude productivity improvement, but such improvements do not and cannot exist; real gains have been extremely modest and are at least partly attributable to reductions in quality. The extent to which AI can automate a task is directly correlated to the extent to which quality can be degraded without short-term consequences.

The analysis situates generative AI within a broader pattern of technology adoption driven by organizational and financial incentives rather than technical merit. A core observation is that AI functions as a probability engine that drags operations toward statistical mean performance—it cannot consistently exceed human expertise, yet executives and non-experts often misjudge its output because they lack domain knowledge to detect errors. The piece notes that confidence, not quality, is the currency of leadership, creating a structural mismatch between what AI actually produces and how it is perceived.

The framework also documents how the term "AI" itself has become decoupled from any fixed definition. Microsoft's application of "Copilot" branding to at least 80 different products exemplifies this: the label can attach to image generators, recommendation algorithms, remote operators, or advertising campaigns with no meaningful technical commonality. This flexibility serves a deliberate business purpose—saying "AI" matters more than defining it, which in turn matters more than delivering results. The industry's incentive structure explicitly rewards the narrative over the execution.

Finally, the analysis warns that the economic logic driving adoption is unsustainable. Companies are drawn to AI because it promises to substitute SaaS fees for payroll, but current pricing is artificially depressed by unprecedented investor subsidies. Should those subsidies end, the cost calculus shifts dramatically. Meanwhile, real productivity gains have been modest and are "at least partly attributable to reductions in quality," suggesting that measured efficiency gains often reflect a willingness to accept lower-quality output rather than genuine automation of skilled work.

No comments yet. Be the first to share your thoughts!