Bettering Me

Tag: Cognitive Edge

  • The Cognitive Atrophy Tax Compounds Silently – Like Sedentary Behavior, but for Your Mind

    Written by

    Unpaid Intern

    –

    The sedentary behavior analogy is useful for understanding cognitive decline in the AI era. No single missed workout destroys your fitness. One week of missed workouts is negligible. But a year of consistent sedentary behavior changes your baseline – your cardiovascular capacity declines, your muscle mass decreases, and your metabolic health deteriorates. The change is invisible day-to-day and visible year-over-year.

    The same mechanism applies to cognitive exercise in the age of generative AI.

    The Missed Rep

    Every time you outsource a judgment call to AI that you could have made yourself, you miss a rep of cognitive exercise. One rep does not matter. The question you would have puzzled through, the categorization you would have made, the tradeoff you would have weighed – these are the cognitive equivalent of a single squat.

    One missed squat does not change your body. One missed judgment call does not change your mind.

    But 500 missed reps over a year change your baseline.

    Consider the math. If you use AI for ten judgment calls per workday – “draft this email,” “summarize this document,” “suggest options for this problem” – that is approximately 2,500 outsourced judgment calls per year. Even if half of those are genuinely appropriate to outsource, the remaining 1,250 are missed cognitive reps. That is the equivalent of skipping every workout for a year.

    The Neural Mechanism

    The mechanism is use-dependent plasticity: neural circuits that are used frequently strengthen; circuits that are used infrequently weaken [1]. This is not a theory. It is the foundational principle of how the brain adapts to experience.

    The circuits most at risk from AI outsourcing are the ones that do the hard parts of cognition: evaluation (comparing options against multiple criteria), synthesis (integrating information from diverse sources), and taste formation (developing and applying quality standards). These are complex, high-level circuits that require regular engagement to maintain.

    When you skip the evaluation step and accept the AI’s first output, you are not saving time. You are choosing not to exercise the evaluation circuit. One choice is irrelevant. The accumulation of choices is where the tax compounds.

    The comparison to physical exercise is apt for another reason: the effects are bidirectional. Just as a sedentary person can regain cardiovascular fitness with consistent training, an AI-dependent thinker can rebuild the atrophied circuits with deliberate practice. The difference is that the cognitive atrophy is invisible – you do not feel yourself getting shallower the way you feel yourself getting more breathless climbing stairs.

    The Hidden Tax

    The cognitive atrophy tax is hidden because the environment adapts to your declining capacity. When your evaluation circuits weaken, you do not notice worse reasoning. You notice that you trust AI output more. You notice that you second-guess yourself less. The feeling is confidence – when the reality is that your standards have dropped.

    This is the most dangerous feature of the tax: it feels like progress. You produce more output, faster, with less effort. The output passes surface-level scrutiny. No one tells you it is shallow because it looks polished. You have no reason to believe your cognitive capacity has declined because you are producing more than ever.

    The tax comes due when you face a situation that AI cannot handle – a novel problem with no training data, a high-stakes decision with incomplete information, a creative challenge that requires genuine originality. In that moment, you discover that the circuits you would need are weaker than they should be. The capacity you assumed was there is not.

    The Longevity Risk

    The longevity risk is not that AI will replace your thinking. It is that you will stop exercising the neural circuits that do the hard parts – and those circuits will degrade like an unworked muscle [2].

    The tax is invisible until you need the capacity and find it gone. The first time you need to make a complex, high-stakes judgment call without AI assistance – in a meeting, under pressure, with incomplete information – and you realize you cannot hold the reasoning chain, that is the tax coming due.

    The long-term implication is that cognitive decline in the AI era will not be uniform. People who use AI as a scaffold for their own thinking – making the call themselves first, then comparing – will maintain and even strengthen their judgment. People who use AI as a substitute – accepting output without evaluation – will experience gradual, unnoticed decline. The difference between the two trajectories is not in the tool. It is in the relationship to the tool.

    The Protocol

    The fix is not to reject AI. It is to treat every interaction with AI as training data for your own judgment.

    Make the call yourself first. Then compare with the AI. The difference between those two answers is where your cognitive growth lives. If the AI’s answer is better, study why. If your answer is better, you have confirmation that your judgment is intact. Either outcome is useful. The only useless outcome is accepting the AI output without having formed your own answer.

    This protocol takes more time per interaction. That is the point. The time is not overhead. It is the cognitive training that keeps your judgment from atrophying.

    Disclaimer: This post is for inspiration and education, not medical advice. Everyone’s body is different, so please check with your doctor before changing your diet, exercise, or lifestyle routine. By using these tips, you agree to do so at your own risk.

    References

    [1] Dweck CS. Mindset: The New Psychology of Success. Random House; 2006

    [2] Pascual-Leone A, Amedi A, Fregni F, Merabet LB. "The plastic human brain cortex." Annual Review of Neuroscience. 2005;28:377-401. DOI: https://doi.org/10.1146/annurev.neuro.27.070203.144216

  • The Skill That Compounds Most in an Automated Era Is Not Speed – It’s Depth

    Written by

    Unpaid Intern

    –

    Every productivity tool on the market promises speed. Faster writing. Faster research. Faster decision-making. The assumption is that speed is the bottleneck – that if you could just produce output faster, you would produce more value.

    The assumption is wrong. When automation makes speed a commodity, speed stops being a differentiator. Depth becomes the only thing that cannot be automated.

    Speed as Commodity

    Generative AI can produce a competent first draft of almost anything in seconds – a report, an email, a marketing copy, a summary of research. The speed at which it produces these outputs is already faster than any human. And it will only get faster.

    When everyone has access to that speed, speed confers no advantage. The baseline rises. Everyone will be fast. The person who stands out will be the one who produces something that is not just fast, but good – where “good” means original, deeply reasoned, and based on a complex understanding that the model does not have.

    This is the economic logic of automation: when a skill becomes universally accessible, its market value drops to near zero. Speed of output is already following this trajectory. The value that remains is in the things the automated tool cannot do – and those are the things that require depth [1].

    What Depth Looks Like

    Depth is the ability to hold a complex reasoning chain for 45 minutes without needing external input. It is the capacity to evaluate a problem from multiple perspectives, synthesize conflicting information, and arrive at a judgment that accounts for nuance.

    This is not a skill that AI will soon replicate. AI can produce plausible reasoning chains, but it does not have the lived context, the domain-specific tradeoff knowledge, or the ability to weigh competing values that human depth provides [2].

    Consider the difference between a well-researched AI analysis of a strategic business problem and the analysis of a partner who has worked in that industry for 20 years. The AI analysis will be comprehensive, well-structured, and full of relevant data. The partner’s analysis will be shorter, less polished, and more nuanced – because it draws on experience that cannot be captured in training data. The partner knows which of the data points matters, which risks are real and which are theoretical, which stakeholders will resist and why. That is depth.

    Depth is rare because it is hard to train and easy to avoid. The knowledge work environment actively discourages it – favoring responsiveness, availability, and rapid cycling over sustained thought. The person who cultivates depth despite the environment is building an asset that becomes more valuable as the environment becomes more automated.

    The Counterfeit of Depth

    As depth becomes more valuable, shallow output will increasingly try to mimic it. AI-generated content is already sophisticated enough to pass for deep analysis at a casual read. The tell is not in the grammar or structure. It is in the absence of genuine tradeoff discussions, the lack of specific contextual knowledge, and the failure to acknowledge what is not known.

    The danger is not that you will be fooled by bad analysis. The danger is that you will not be able to tell the difference because your own depth has atrophied. A person who has done the work of depth – who has held complex reasoning chains, made difficult tradeoff decisions, and synthesized conflicting information – can spot shallow reasoning immediately. It feels thin. It lacks the texture of genuine engagement with a hard problem. A person who has outsourced depth for years has lost that calibration. Shallow output feels sufficient because they no longer know what depth feels like.

    The person who can distinguish genuine depth from convincing mimicry has an advantage that compounds. They can evaluate AI output critically, selecting what is useful and discarding what is superficial. They can identify the gaps in the analysis and fill them with their own expertise. The person who cannot tell the difference will be increasingly reliant on output that is plausible but shallow.

    The Practical Counterargument

    Is depth always the right move? No. There are contexts where speed genuinely matters – crisis response, time-sensitive decisions, high-volume production environments. The argument for depth is not that every task requires it. It is that if you never train depth, you lose the capacity to deploy it when it matters. The person who can go deep on demand but chooses shallow when appropriate has agency. The person who can only go shallow has no choice. The training is not about rejecting speed. It is about maintaining the option of depth.

    The Trainable Skill

    Depth is trainable. It requires deliberate practice of uninterrupted reasoning – not insight, not creativity, but the mundane skill of holding a thought for longer than is comfortable.

    The premium is not on generating the answer. The answer is cheap now. The premium is on holding the question – on staying with the problem long enough to understand it deeply, rather than jumping to the first plausible solution.

    The training protocol is straightforward: every day, spend 30 minutes on a single problem, without interruption, without searching for answers, without asking AI. Just holding the problem. Turn it over. Examine it from different angles. Resist the urge to resolve it. The discomfort of not-knowing, sustained over time, is the training stimulus for depth.

    That skill is trainable. It is rare. And in an automated era, it is the only edge that compounds.

    Disclaimer: This post is for inspiration and education, not medical advice. Everyone’s body is different, so please check with your doctor before changing your diet, exercise, or lifestyle routine. By using these tips, you agree to do so at your own risk.

    References

    [1] Newport C. Deep Work. Grand Central Publishing; 2016

    [2] Autor DH. "Why Are There Still So Many Jobs? The History and Future of Workplace Automation." Journal of Economic Perspectives. 2015;29(3):3-30. DOI: https://doi.org/10.1257/jep.29.3.3