The shift from passive surveillance to active capacity marks a fundamental inversion of power. When the cost of delivering the effects falls dramatically relative to the cost of defending against them, traditional balances dissolve. Declining marginal costs turn modest investments into instruments of outsized impact.
Some definitions are necessary. Those familiar with the lingo can skip the next seven paragraphs.
Leverage magnifies the results of input, effort, or capital. Compounding is when returns on an asset or consistent small improvements or degradations build on themselves over time, creating a non-linear curve. Warren Buffett rightly calls compounding one of the most powerful forces in the universe.
A non-linear curve is a line on a graph that does not form a straight path. In a linear curve (oxymoronic terminology at first blush, but bear with me), a change in one variable causes a constant, predictable change in another (e.g., working two hours versus one hour doubles your pay). In a non-linear relationship, the rate of change shifts, meaning the output is disproportionate to the input (think overtime pay).
Characteristics include:
Changing slopes: bends, flattens, or spikes at different points.
Exponential shifts: structures where small early changes produce massive subsequent results.
Unpredictable scale: small input tweaks result in no change at all, or a massive, explosive jump in output.
It is easy to presume that the non-linear curve is exponential, upward, to the right, and that it sits above the x-axis. All are misconceptions. Exponential includes downward and below the x-axis. And neither needs be to the right. New York University professor Nassim Taleb proves most helpful in his Incerto works, particularly Antifragility.
Additional definitions help regarding convex and concave non-linear curves:
Convex (antifragile in the graph below): limited, capped downsides but massive, open-ended upsides. More gain than pain.
Concave (fragile in the graph below): very little upside, but massive, catastrophic downsides. More pain than gain.
Figure 1 – with permission from Stoic Agilist.
Convexity and concavity are inversions of one another. I am taking time here as they relate to the subsequent parts of the essay and inform applications (equations) that were requested by readers of The death of thought and the mirage of expertise.
Back to modest investments into instruments with outsized impact. Low-cost autonomous systems achieve results that once required industrial arsenals. The asymmetry is staggering.
As fellow BIG Media contributor Grant Wilde documented earlier this year, military drone production has scaled into the millions per year; cumulative monthly launches in recent conflicts exceed tens of thousands. One-way attack drones costing thousands of dollars now force the expenditure of interceptors costing multiple millions. Cost ratios of 1,000:1 or higher are no longer theoretical. That ratio stands to widen. Attack drones now inhabit the seas, with Anduril’s Dive-LD apparently captured by Iran. Are cheap robot troops and police forces far off?
This same inversion operates at the level of cognition. Tools that make thinking cheap also reduce the practice that builds cognitive endurance. That practice is best honed with friction, difficulty, and failure because the brain adapts to the environment that it repeatedly inhabits. An environment of constant novelty, rapid switching, and immediate answers trains the opposite of sustained attention and intolerance for confusion. Such an environment induces frustration and impatience. Reading, mathematics, and science all require this same, demanding,
uncompromising, underlying machinery. Equation 1 looks like this:
Modern digital environments increasingly train the reverse loop. Equation 2 looks like this:
(The preceding equations are adapted from SightBringer.)
After the AI tech tool is removed, performance often falls. Cheap power, cheap production, and cheap cognition share runaway dynamics (leverage, compounding, and non-linearity) once a critical threshold is crossed.
Power becomes hyper-concentrated. A small number of operators, or automated loops, can hold keystroke-level influence over outcomes that affect thousands. Recent rogue agent discoveries (or active releases depending on your proclivity for accidental/deliberate) attest to out-sized returns or impacts. Targeting platforms compress human decision points and accelerate the loop, causing disorientation. Once closed, the marginal cost of additional cycles falls while the cost of rebuilding resilient, human-heavy alternatives rises. The same asymmetrical pattern appears in the informational domain: securing narrative control or regulatory framing is relatively cheap; reversing institutional inertia is expensive. And largely inaccessible.
Infrastructure politics reveal a further asymmetry. Data centres face widespread local opposition, yet polling shows the dominant drivers are conventional concerns (environment, costs, locality) rather than sophisticated arguments about cognitive or existential risk. AI-specific framing adds only a modest penalty. Public attention remains fixed on proximate costs while the deeper inversion continues. If data-centre headlines do not reflect the ground truth, will data-centre policies be any better?
Societies treat (are sold) these developments as technical upgrades. Efficiency narratives predominate. Everyday populations and local governance show limited capacity for sustained resistance. Fatigue and the incremental character of the changes favour accommodation. The sleepwalking continues because each individual step appears rational when viewed in isolation. Each step unrecognizable in a journey toward a destination bereft of perspective.
I am returning to SightBringer for the next equation, which informs his lament for the increase in teenage dimness (The Economist) in reading, mathematics and science in my second essay, The dehumanization engine. Equation 3 claims a reflexive civilizational pattern of cognitive outsourcing:
Progressive, cognitive outsourcing removes the exact cognitive operations that used to develop intelligence in the first place.
Accommodation is the norm, and the AI industrial complex knows it (check the names of models for nods to myths and heroes). The paddlers, purveyors, and supervisors of cognitive outsourcing know leverage, know compounding, and know non-linearity. They practise and protect them ruthlessly, daily. Those among us taking a heroic stand face a triad of Hydra, Gorgon, and Cerberus.
I am invoking a recent series of parallel scaling of irreversible decisions. Medical assistance in dying (MAiD) converts rare, exceptional choices into increasing procedural volumes. Official Canadian data shows continuing provisions under streamlined pathways where fewer gatekeepers exercise precautionary judgment under the full weight of the stakes, or what are better called incentives. Once normalized, expansion follows the path of least resistance. Lower ohms make for more current. If that series interests you, From Rodriguez to routine, where does it end, and Financial reckonings provide cautions on fence dismantling and boundary softness.
SightBringer’s loop benefits from similar asymmetry. It does not require overwhelming force when small, smart applications suffice. Little by little, the old rules erode in favour of new ones that favour speed, scale, and detachment. Move fast and break things, with little consideration to what is being broken – only that it breaks, and quickly. This is no dramatic apocalypse; it is the steady, incremental democratization of ruin, made possible by trading resilience for cheaper capability at every step. When dopamine morphs into concave pathology, entropy is easier to spawn.
This decay includes the capability to think without the tool – birthing more dependency.
A and B: a practical frame
Let A stand for the default path: allow the inversion to swamp you. Accept cheaper tools as pure gain, outsource the cognitive operations that used to build endurance, treat profiles and scores as adequate substitutes for persons, and watch the cost of reversal climb while the cost of further optimization falls. A is the path of least immediate resistance. Many people, perhaps most, are already on it, whether they recognize or admit it.
Let B be the set of deliberate counter-moves. Three concrete examples:
- Reintroduce friction on purpose. Schedule regular periods of unaided problem solving, long-form reading, or manual calculation. Pen/pencil and paper are markedly better than fingers on keyboards. The machinery of sustained attention atrophies without use; it can be rebuilt with deliberate practice, but, as with all atrophy, rehab is a bitch.
- Refuse the easy proxy. When a system offers a score, a risk category, or a generated summary, demand the underlying evidence or reconstruct the reasoning yourself before accepting it. This is slower, but it is the only way to keep discernment alive.
- Protect the attentional environment of the next generation. Limit the stimulus-reaction-novelty loop in the years when working memory and resistance to distraction are still being wired. The data already shows the cost of failing to do so.
None of these restores the old equilibrium. Nor do they guarantee victory. But they do restore a measure of agency, for without agency you kneel.
One reader of the first essay, The death of thought and the mirage of expertise, asked what can be done. In addition to the preceding three points, here is my attempt at the anti-pattern (or antidote) civilizational Equation 4 that SightBringer did not, ahem, bring:
This is the precise counter-sequence to SightBringer’s cognitive outsourcing loop. Where the original chain treats every reduction in friction as progress, thereby eroding the very faculties the tools were meant to augment, the antidote equation begins by re-inserting friction on the front end intentionally. The rest of the causal chain follows from that key first step: endurance recovers, dependence loosens, native capacity returns, and tools can finally be used without progressive self-diminishment.
My attempt is the operational form of the three concrete moves already listed under path B, and a Chestertonian decision to stop mistaking friction removal for advancement. Antifragility cannot be promised, but some means for resilience vs concavity can be provided that one can both leverage and compound as countermeasures. The equation is in your hands, practice makes it stronger.
Admittedly, my humble equation faces the heroic triad test. Pressing the easy button is B.F. Skinner’s box. Do you think the tech masterminds have not operationalized and distributed the operant conditioning of caged rats?Chesterton’s fences were dismantled because their purpose was forgotten. Reassembly begins with the decision to stop treating every reduction in friction as progress. [Note that some tech leaders refuse their own tools for their own children.]
At the same time, new fences for innovations, which Chesterton may not have contemplated, will need to be erected and new bridges constructed to overcome both the disorientation and the chasm from the present to impossibilities that AI will enable humanity to make real.[1]
References
[1] Disorientation is an inevitable outcome and a great risk, in the form of collateral damage for which we are very early on the cohort curve. Will digital natives yield to AI natives? I am neither advocating Luddism nor technophilia on the matter of AI disorientation. I am merely confronting its inevitability its value. Disorientation can be a worthy, necessary phase preceding epistemic leaps: where the gaps between the present and the potential escape thought, understanding, language, and meaning. I recommend this post (and others) by media theorist and philosopher Leon Tsvasman on the matter. AI permits not only the novel and the orthogonal, but the unprecedented, even impossible. This perspective will contribute to the final two essays in the series.
(Richard LeBlanc – BIG Media Ltd., 2026)