# Thoughts on Thinking: Physics, Biology and Thought

What the frequencies actually are, what changes them, and why the work is still the person

Canonical HTML: https://www.livetradingnews.com/thoughts-on-thinking-the-physics-and-biology-of-the-mind
Last modified: 2026-10-17

---

By Shayne Heffernan · 2026-10-17
Tags: $LLY, $PM
Signed: ML-DSA-87, anchored on Armature L1.
Nothing in this article is investment advice.

## The substrate

This is the next note in the series. In July, in [Thoughts on Thinking](https://www.livetradingnews.com/thoughts-on-thinking), I drew the line between what drives thinking and what amplifies it. The driver is the question that will not leave. The amplifier is whatever sharpens the answer once the question is there: the room, the sound, the temperature, the walk you dislike but take anyway. In October, in [Thoughts on Thinking in an AI World](https://www.livetradingnews.com/thoughts-on-thinking-in-an-ai-world), I wrote that answers had become cheap and that the thing that stayed scarce was the record of what a named person believed, why it changed, and what they were willing to stand behind.

[The art-break note](https://www.livetradingnews.com/thoughts-on-thinking-take-an-art-break) measured one small restoration: a short, structured look at work that required sustained attention under real constraints improved subsequent attentional control more than an ordinary urban walk. Those pieces treated thinking as something that can be cleared and briefly restored. This one is about what is left once the noise is gone.

In September, [Frequencies and Thinking](https://www.livetradingnews.com/frequencies-and-thinking) set out what damages a mind and what can train it. This note goes underneath it, to the cells. The full series is in the [Thinking Center](https://www.livetradingnews.com/center/thinking) on Live Trading News and under [Thinking](https://www.shayneheffernan.com/thoughts/thinking) on [shayneheffernan.com](https://www.shayneheffernan.com).

Thinking is not a faculty that sits on top of the body and is merely correlated with it. It is the body, specifically the electrochemical activity of neurons organised across networks. The membrane potential of a pyramidal cell, the timing of the inhibitory currents that pace it, the phase relationship between a slower rhythm and the faster bursts nested inside it, and the metabolic state that keeps the pumps running are not background. They are the process. Because they are physical they are conditional. Some of the conditions can be measured. Some can be moved.

I am not interested in optimisation as a performance. I am interested in whether a person can enter a clearer state more reliably, stay in it longer under load, and return to it after interruption with less of the previous state still attached. That is a practical question. It has a physics and a biology. The rest of this note sets out what those are, what ordinary and less ordinary inputs do to them, and why the work at [KXCO](https://kxco.ai) is aimed at the same problem from the other side: not by generating the answer, but by making the conditions for the person's own answer more stable and more inspectable.

## What a neuron is doing

A resting cortical neuron sits at roughly −70 mV. That number is not a fixed property of the cell. It is the continuous result of unequal ion concentrations and selective permeability. Sodium is higher outside, potassium higher inside. The sodium-potassium pump spends ATP to keep the gradients from collapsing. At rest the membrane is more open to potassium than to sodium, so the voltage sits closer to the potassium equilibrium than to the sodium one.

Small changes in permeability move the voltage. Larger changes, once they cross a threshold, open voltage-gated sodium channels. Sodium enters, the membrane depolarises further, more channels open, and the regenerative spike runs. Within a millisecond or two the sodium channels inactivate and potassium channels open. The potential is restored. The whole event is brief. It propagates because the local current depolarises the next patch of membrane to threshold. Hodgkin and Huxley measured this sequence in the squid giant axon and published the quantitative account in [The Journal of Physiology](https://doi.org/10.1113/jphysiol.1952.sp004764) in 1952.

At the synapse the arriving spike opens calcium channels. Calcium entry triggers vesicle release. Transmitter binds to receptors on the far side. Some receptors open ion channels directly. Others act through slower cascades. The result is a postsynaptic potential that is graded, not all-or-nothing, and that lasts longer than the spike. These potentials sum. If the net depolarisation at the axon initial segment reaches threshold, another spike is born. Most of the time it does not. The cell fires irregularly. What looks organised is the population.

Scalp EEG does not record the spikes. It records the summed postsynaptic currents of many pyramidal cells whose apical dendrites are oriented roughly the same way. When those cells receive input together, the extracellular currents add. After the skull attenuates them, what remains is measurable as an oscillation whose frequency is simply how often the summed current reverses. Faster reversals require faster channel and synaptic kinetics. Slower ones can be carried by longer inhibitory currents, by membrane resonances, or by loops whose delays match the period. The frequencies are not labels we impose. They are what the time constants of the machinery produce when the machinery is driven. Buzsáki, Anastassiou and Koch set out the physics of those extracellular fields in [Nature Reviews Neuroscience](https://doi.org/10.1038/nrn3241) in 2012.

This is why interventions that change excitability, energy supply, or the timing of inhibition can move the observed spectrum. They are not acting on an abstract mind. They are acting on the same ion gradients and synaptic currents that generate the rhythm in the first place.

## How networks make frequencies

Individual cells fire irregularly, even when the field is rhythmic. The rhythm is an emergent property of the circuit. The best-characterised mechanisms for the faster bands involve the push and pull between pyramidal cells and inhibitory interneurons, especially the fast-spiking parvalbumin-positive cells that target the cell bodies of the pyramidal population. Those interneurons fire in tight volleys. Their inhibition creates brief windows in which the pyramidal cells can spike and longer windows in which they cannot. The pyramidal spikes, when they occur, re-excite the interneurons. The loop closes.

Two motifs are sufficient to produce gamma. In one, mutually connected inhibitory cells synchronise through the timing of their own inhibition, provided there is enough tonic drive to keep them near threshold. The cycle length is set largely by how long the inhibitory current lasts. In the other, a burst of pyramidal activity recruits the interneurons, which then silence the pyramidal cells until the inhibition decays. Both operate in the 30 to 100 Hz range under ordinary conditions. The exact frequency depends on drive, on the strength of the inhibition, and on the subunit composition of the GABA-A receptors that carry it. When those interneurons are driven or silenced experimentally, gamma power moves with them.

Buzsáki and Wang review both motifs in the [Annual Review of Neuroscience](https://doi.org/10.1146/annurev-neuro-062111-150444), 2012. Cardin and colleagues generated gamma by driving the parvalbumin-positive cells with light, and Sohal and colleagues suppressed it by inhibiting them, in two papers Nature published together in 2009 ([Cardin](https://doi.org/10.1038/nature08002), [Sohal](https://doi.org/10.1038/nature07991)).

Slower bands use related ingredients on different timescales or over longer distances. Beta in motor and prefrontal cortex can be produced by local excitatory-inhibitory loops that simply run more slowly, or by interactions between regions whose conduction delays favour that range. Alpha is heavily shaped by thalamocortical circuits. The thalamus can pace cortex and cortex can shape the thalamic rhythm in return. Hippocampal theta depends on septal drive as well as on local properties and is tightly linked to the timing of spatial and episodic firing. In each case the oscillation is not a pure local clock. It is the interaction of cellular time constants with the architecture that connects the cells. The thalamocortical account of the slow rhythms is Steriade, McCormick and Sejnowski's, in [Science](https://doi.org/10.1126/science.8235588), 1993, and Buzsáki reviews the septal account of hippocampal theta in [Neuron](https://doi.org/10.1016/s0896-6273%2802%2900586-x), 2002.

The functional consequence is phase-dependent excitability. A cell is more likely to fire, and more likely to drive its targets, when its membrane is closer to threshold at the preferred phase of the local rhythm. Synchrony in a given band therefore raises the effective connectivity between regions that are oscillating together. Nesting adds another layer. Gamma bursts commonly ride on the phase of a slower theta or beta cycle. In working-memory tasks the number of gamma cycles that fit inside one theta cycle has been treated as a rough limit on how many distinct items can be held. In prefrontal cortex, deeper-layer alpha and beta appear to regulate the expression of superficial-layer gamma, so that goals or rules can gate what content is admitted or maintained. The same tissue supports different operations depending on which frequencies dominate and how they are coupled.

Fries calls the first of these effects communication through coherence, in [Neuron](https://doi.org/10.1016/j.neuron.2015.09.034), 2015. Lisman and Idiart proposed the gamma-in-theta limit in [Science](https://doi.org/10.1126/science.7878473) in 1995, and Lisman and Jensen restated it as the theta-gamma neural code in [Neuron](https://doi.org/10.1016/j.neuron.2013.03.007) in 2013. The laminar result is from Bastos and colleagues in [PNAS](https://doi.org/10.1073/pnas.1710323115), 2018, and it sits inside the wider model Miller, Lundqvist and Bastos call [Working Memory 2.0](https://doi.org/10.1016/j.neuron.2018.09.023).

None of this requires a special frequency for "thought." It requires that the balance among the available frequencies, and the coupling between them, be appropriate to the operation the network is trying to perform. Change the balance and the operations that are easy or hard change with it.

## The bands, without the mysticism

The conventional divisions are convenient. The boundaries move a few hertz depending on the laboratory. The associations with state are stable enough to be useful if they are not treated as identities.

### Delta

Largest amplitude, slowest frequency, dominant in deep non-REM sleep. The waves reflect widespread hyperpolarisation coordinated through thalamocortical loops. Reportable thought is minimal. In waking adults, prominent delta during a task that requires alertness is usually a sign that the system has drifted. It is not a target for enhancement of ordinary cognition.

### Theta

Prominent in hippocampus during exploration and encoding, and in frontal midline sites during tasks that demand control or sustained attention. It organises sequences. Gamma bursts nested at particular phases of the theta cycle have been linked to the ordering of items in working memory and to the binding of episodic detail. At rest or in light meditative states it accompanies more associative or internally directed processing. Raised frontal theta during a hard task is more often a marker of effort than of the content being processed. It is useful. It is not the same thing as insight. Cavanagh and Frank set out frontal theta as the signal that more control is needed, in [Trends in Cognitive Sciences](https://doi.org/10.1016/j.tics.2014.04.012), 2014.

### Alpha

Strongest posteriorly with eyes closed in a relaxed waking state. Classically read as an inhibitory rhythm that reduces the throughput of external sensory input so that other networks can operate with less interference. It drops with focused visual attention or mental effort. Increases in alpha power and coherence appear after time in natural settings, after some dietary shifts, and during certain forms of reduced sensory load. It is a reasonable marker that unnecessary arousal has decreased. It is not, by itself, evidence that better thinking is occurring. Klimesch reviewed the link between alpha, theta and memory performance in [Brain Research Reviews](https://doi.org/10.1016/s0165-0173%2898%2900056-3), 1999, and Jensen and Mazaheri gave the gating-by-inhibition account in [Frontiers in Human Neuroscience](https://doi.org/10.3389/fnhum.2010.00186), 2010.

### Beta

Dominant during ordinary waking work that involves analysis, planning, or the maintenance of a current set. Low beta is often associated with calm focus. Higher beta can accompany either intense processing or anxiety. The distinction is not available from the frequency alone. In sensorimotor cortex beta is high when movement is withheld and falls just before movement begins. In prefrontal areas it participates in the stabilisation of rules and in the gating of working-memory content. Sustained high beta without a task is frequently rumination or stress, not productivity. Engel and Fries proposed that beta signals the maintenance of the status quo, in [Current Opinion in Neurobiology](https://doi.org/10.1016/j.conb.2010.02.015), 2010, and Lundqvist and colleagues recorded working-memory content in brief gamma bursts with beta bursts marking its control, in [Neuron](https://doi.org/10.1016/j.neuron.2016.02.028), 2016.

### Gamma

Linked to local binding of features, to the active maintenance of items in working memory, and to periods of high focus. The 40 Hz point has been studied heavily because sensory drive at that frequency can increase gamma power. In multiple mouse models of Alzheimer's pathology, repeated 40 Hz light and sound reduced amyloid and tau, preserved synapses, and improved memory measures. Those results come from the Tsai laboratory at MIT, first [Iaccarino and colleagues](https://doi.org/10.1038/nature20587) in 2016, then [Martorell](https://doi.org/10.1016/j.cell.2019.02.014) and [Adaikkan](https://doi.org/10.1016/j.neuron.2019.04.011) and their colleagues in 2019.

Small human studies have shown increased 40 Hz power, changes in connectivity, and, in some participants with mild impairment, slower decline on selected tests. Those are MIT's pilot study, reported by [Chan and colleagues](https://doi.org/10.1371/journal.pone.0278412) in 2022, and the OVERTURE trial of Cognito Therapeutics' device, reported by [Hajós and colleagues](https://doi.org/10.3389/fneur.2024.1343588) in 2024. In healthy adults one controlled vigilance study found better accuracy and faster reactions under 40 Hz audiovisual flicker than under control or sham conditions, with accompanying shifts in lower-frequency connectivity that tracked performance. That study tested 62 people and is reported by [Attokaren and colleagues](https://doi.org/10.1162/imag.a.1229) in Imaging Neuroscience, 2026.

Gamma is typically nested inside slower rhythms. The slower rhythm supplies structure. Gamma supplies resolution. Scalp recordings underestimate it because the skull attenuates high frequencies. Intracranial data confirm it is present when the task is demanding.

The practical reading is straightforward. Analytical work tends to raise beta and task-related gamma while suppressing alpha in the relevant sensory areas. More open or restorative periods raise alpha and theta. The aim is not a permanent elevation of any single band. It is a higher probability of entering the balance the work requires, and of leaving it cleanly when the work changes. Exhibit 1 draws the chain from the ion pump to the operation each band favours.

![Knowledge graph: ion pumps and gradients produce the spike and the synaptic currents, circuit loops set five bands, and each band favours an operation, with theta-gamma nesting and alpha-beta gating](https://livetradingnews-media.nyc3.digitaloceanspaces.com/media/2026/10/17/cmpgg3-4d451b4fdc5d8f84.svg)

*Exhibit 1. From the ion gradient to the operation: how the machinery makes the bands and what each band favours. Each numbered link is a sourced claim, listed under the graph.*

## What moves the balance

The same biophysics that generates the rhythms also makes them sensitive to ordinary inputs. Energy supply, sensory drive, inhibition, and arousal all shift the spectrum. The shifts are partial and they differ across people. They are still measurable.

### Food and fuel

Cortical activity is expensive. Large, rapid swings in blood glucose produce corresponding swings in alertness and in the stability of higher-frequency activity. Patterns that avoid those swings, with moderate loads, adequate protein and sufficient fat, support more consistent performance on tasks that require sustained beta and gamma. Omega-3 fatty acids, particularly DHA, are structural components of neuronal membranes. Higher intake is associated with better cognitive scores in multiple cohorts, per the pooled prospective studies of Wei and colleagues in [The American Journal of Clinical Nutrition](https://doi.org/10.1016/j.ajcnut.2023.04.001), 2023. Some EEG work reports more favourable alpha-to-beta ratios and less excessive high-beta activity with enrichment.

Nutritional ketosis, in which ketones supply a substantial fraction of cerebral fuel, has been examined with EEG for decades in epilepsy and more recently in other groups. The most consistent observation is increased theta power, sometimes with higher alpha coherence. That pattern is compatible with a calmer attentive state and with reduced high-beta excitability. It is not a universal upgrade. It is a shift in the baseline on which other demands operate.

Smaller acute effects appear with polyphenol-rich foods and with certain nuts. One pilot EEG study found a notable gamma response after pistachios, per Berk and colleagues at Loma Linda University in [The FASEB Journal](https://doi.org/10.1096/fasebj.31.1_supplement.636.24), 2017, as [reported](https://www.labmanager.com/study-finds-consuming-nuts-strengthens-brainwave-function-5973) by the university. Longitudinal adherence to Mediterranean or MIND-style patterns is associated with slower decline, per Morris and colleagues in [Alzheimer's and Dementia](https://doi.org/10.1016/j.jalz.2015.04.011), 2015. In recent EEG work on 841 memory-clinic patients it was associated with higher alpha power and stronger frontoparietal connectivity, and the alpha power partially mediated the relationship between diet and function, per Wu and colleagues in [Psychiatry and Clinical Neurosciences](https://doi.org/10.1111/pcn.13923), 2025. These are not large effects in healthy adults over days. They alter the conditions under which larger effects from other interventions become more or less likely.

### Place, light, and load

Time outdoors or even viewing natural scenes reliably increases alpha and theta while reducing the higher-frequency activity linked to cognitive load and stress. Ulrich found alpha significantly higher while people viewed vegetation than while they viewed urban scenes, in [Environment and Behavior](https://doi.org/10.1177/0013916581135001), 1981, and Grassini and colleagues found natural scenery processed with lower attentional and cognitive load than urban scenery, in the [Journal of Environmental Psychology](https://doi.org/10.1016/j.jenvp.2019.01.007), 2019. Subsequent performance on tasks that require directed attention improves relative to urban exposure, per Berman, Jonides and Kaplan in [Psychological Science](https://doi.org/10.1111/j.1467-9280.2008.02225.x), 2008. The finding is consistent with the older claim that natural settings engage attention gently enough for the systems that support effortful focus to recover, which is Kaplan's, in the [Journal of Environmental Psychology](https://doi.org/10.1016/0272-4944%2895%2990001-2), 1995. High visual clutter and unpredictable noise do the reverse. They raise demand and can lift the load signature.

Light acts more directly on arousal. Cooler, brighter light supports beta activity and reported alertness. Dimmer or warmer light favours alpha. Morning bright light helps anchor the daily rise in arousal that makes later focused work more reliable. Cajochen reviews the alerting effects of light in [Sleep Medicine Reviews](https://doi.org/10.1016/j.smrv.2007.07.009), 2007.

Noise level is not linear. Moderate ambient sound, roughly the level of a busy café, has improved creative output in some experiments relative to silence, apparently by adding a mild processing demand that keeps executive networks from idling, per Mehta, Zhu and Cheema in the [Journal of Consumer Research](https://doi.org/10.1086/665048), 2012. Loud or unpredictable noise impairs both analytical and creative work. Consistent low-level masking is often the better choice for sustained analysis.

Air quality and temperature are less discussed and more consequential than most optimisation advice admits. Raised indoor CO₂ and common pollutants reduce executive performance, per Satish and colleagues in [Environmental Health Perspectives](https://doi.org/10.1289/ehp.1104789), 2012, and Allen and colleagues in the same journal in [2016](https://doi.org/10.1289/ehp.1510037). Thermal extremes divert resources. These are not exotic variables. They change the same excitability that the frequency bands reflect.

### Sound and rhythmic drive

Rhythmic sensory input can bias neural activity toward the driving frequency. The clearest current evidence is for audiovisual stimulation near 40 Hz. In mouse models of Alzheimer's pathology, daily exposure over weeks reduced pathological proteins, preserved synapses and neurons, and improved memory-task performance while increasing gamma power and synchrony, in the Tsai laboratory papers above. Related work in a Down-syndrome model showed gains on short-term memory measures and markers of neurogenesis, per Islam and colleagues in [PLOS ONE](https://doi.org/10.1371/journal.pone.0317428), 2025.

Human studies remain small. They have demonstrated safety, increased 40 Hz power, improved connectivity in certain networks, and, in some participants with mild impairment, slower decline on selected cognitive and biomarker measures, in the Chan and Hajós studies above. In healthy adults the vigilance result mentioned above, 62 people tested by Attokaren and colleagues, is the most direct performance finding to date.

The mechanism is not settled. Increased gamma power, possible effects on clearance pathways, and changes in network coordination are all under investigation, and Murdock and colleagues tied the clearance route to the glymphatic system in [Nature](https://doi.org/10.1038/s41586-024-07132-6), 2024. The result is not a general licence to treat 40 Hz as a daily cognitive enhancer. It is evidence that exogenous drive at a frequency already implicated in binding and working memory can alter both biology and task performance under defined conditions.

Binaural beats and isochronic tones are weaker. Meta-analyses find small-to-medium average benefits on attention and memory tasks, per Garcia-Argibay, Santed and Reales in [Psychological Research](https://doi.org/10.1007/s00426-018-1066-8), 2019. EEG studies of actual entrainment are inconsistent. Some show power increases in the target band after several minutes. Many show none. Of the 14 studies Ingendoh, Posny and Heine reviewed in [PLOS ONE](https://doi.org/10.1371/journal.pone.0286023), 2023, five supported the entrainment hypothesis, eight contradicted it and one was mixed. Baseline differences in arousal, dopamine tone, and simple tolerance for the stimulus appear large. Longer exposure before a task tends to produce larger effects when effects appear. These are low-cost probes. They are not reliable instruments.

Other auditory material acts through different routes. Nature soundscapes increase alpha and theta and reduce negative affect relative to urban noise. Instrumental music of moderate tempo can support either focused or more relaxed states depending on the listener and the task. Lyrics compete for verbal resources and are usually inferior for analytical work. Olfactory inputs reach limbic structures with few synapses. Certain scents shift arousal within seconds. Mild cooling, light touch, and changes in posture alter proprioceptive and thermoregulatory drive and, through those channels, cortical excitability. None of these replace practice. All of them change the cost of entering particular modes. Exhibit 2 ranks the levers by the strength of the evidence behind each.

![Knowledge graph: nature, light, low load, stable fuel, clean air, sound and diet against 40 Hz drive and binaural beats, the shift each makes in the spectrum, and the work it changes](https://livetradingnews-media.nyc3.digitaloceanspaces.com/media/2026/10/17/cmpgg3-c179e82f7967d778.svg)

*Exhibit 2. What moves the balance, and how strong each lever is. A solid link is a controlled or measured result, a dashed link is observational, mixed or argued in this note. Each is listed under the graph.*

## Working on it directly

The earlier notes emphasised subtraction. Remove what occupies bandwidth without earning it. That remains the first move. Once the major sources of interference are lower, the remaining variables can be arranged and the transitions between modes can be practised. Sleep and morning light are stabilised so the daily rise in arousal is predictable. Meals are timed and composed so that large glycaemic swings do not coincide with periods that require sustained analysis. The visual and auditory field during that analysis is simplified. Short, structured interruptions, including the kind of looking described in the [art-break note](https://www.livetradingnews.com/thoughts-on-thinking-take-an-art-break), are inserted to restore alpha and theta and to clear residual load. For people who tolerate them, brief periods of rhythmic sensory stimulation can be tested against actual subsequent performance rather than against how the stimulation felt.

Measurement is required. Without some external check, whether recurring task metrics, occasional EEG if it is available, or at minimum a disciplined log of clarity and stability, it is easy to retain interventions that feel productive and discard ones that are. The test is the quality of the thinking that follows, not the theory that predicted it. Architectures differ. What sharpens one person's beta under load may fragment another's. The principle is the same one I described in [July](https://www.livetradingnews.com/thoughts-on-thinking): refine what drives the thinking, then test what amplifies it, and keep only what survives contact with the work.

There is a limit. No pattern of light, sound, or food supplies the knowledge or the willingness to stay with a question after the initial interest fades. The inputs change the reliability and the energetic cost of certain operations. They do not perform the operations. Enhancement, in the sense used here, is the improvement of conditions and of external scaffolds. It is not the transfer of judgment to a system that does not have to live with the consequences.

## What AI does to the same networks

The newest input to these networks is the one most institutions are now buying. Working memory holds about four items at once, per Cowan's review of the capacity research in [Behavioral and Brain Sciences](https://doi.org/10.1017/s0140525x01003922), 2001, and the theta-gamma account above gives a mechanism for a limit of that order. Sweller's cognitive load theory separates the load a problem carries from the extraneous load its presentation adds, in [Cognitive Science](https://doi.org/10.1207/s15516709cog1202_4), 1988, and with van Merriënboer and Paas in the [Educational Psychology Review](https://doi.org/10.1023/a:1022193728205), 1998. On that definition, rebuilding context that should have been kept is extraneous load.

Leroy found that people who leave a task unfinished carry attention residue into the next one and perform worse on it, in [Organizational Behavior and Human Decision Processes](https://doi.org/10.1016/j.obhdp.2009.04.002), 2009. Mark, Gudith and Klocke found that interrupted workers make up the time by working faster, and pay for it in stress, frustration and effort, at [CHI 2008](https://doi.org/10.1145/1357054.1357072).

The first measurements of what a language model does to that budget are in. Kosmyna and colleagues at the MIT Media Lab recorded EEG from 54 people writing essays over four months with a large language model, with a search engine, or with nothing. The unaided group showed the strongest and most widely distributed connectivity, the search group less, and the model group the weakest. When model users were moved to unaided writing in a fourth session, they showed reduced alpha and beta connectivity, and model users struggled to quote accurately from essays they had just written, per the team's [preprint](https://arxiv.org/abs/2506.08872), June 2025.

In a survey of 319 knowledge workers by Microsoft Research and Carnegie Mellon, higher confidence in generative AI went with less critical thinking and higher self-confidence went with more, per [CHI 2025](https://doi.org/10.1145/3706598.3713778). Gerlich surveyed and interviewed 666 people and found that frequent use of AI tools correlated with lower critical-thinking scores, with cognitive offloading carrying the relationship, per [Societies](https://doi.org/10.3390/soc15010006), 2025.

These results describe the artificial mode: the machine produces the answer and the person receives it. Offloading itself is old. Sparrow, Liu and Wegner showed that people who expect information to stay available remember where to find it rather than the information, in [Science](https://doi.org/10.1126/science.1207745), 2011, and Risko and Gilbert treat offloading as a trade that frees capacity at a cost to what is retained, in [Trends in Cognitive Sciences](https://doi.org/10.1016/j.tics.2016.07.002), 2016. I have called the result [the neat handwriting of the illiterate](https://www.livetradingnews.com/ai-is-the-neat-handwriting-of-the-illiterate): a model that finishes the sentence before the person has done the work.

The opposite design is as old as the computer. Licklider described man-computer symbiosis in 1960: people set the goals, formulate the hypotheses, determine the criteria and perform the evaluations, and the machine does the routinisable work, per [IRE Transactions on Human Factors in Electronics](https://doi.org/10.1109/thfe2.1960.4503259). Engelbart named his programme [Augmenting Human Intellect](https://www.dougengelbart.org/content/view/138) in 1962. Clark and Chalmers argued in [Analysis](https://doi.org/10.1093/analys/58.1.7), 1998, that a notebook which is reliably there, trusted and consulted is part of the mind that uses it. That is the line between artificial and enhanced, drawn more than sixty years before the current models. Artificial intelligence puts an answer in front of the person. Enhanced intelligence puts more of the person's own material in front of them, with its provenance, and leaves the operation to them.

The models carry the same constraint. Liu and colleagues found that language models use information at the start and end of a long context well and information in the middle badly, in [Transactions of the Association for Computational Linguistics](https://doi.org/10.1162/tacl_a_00638), 2024. A larger window is not a record, and [context](https://www.livetradingnews.com/the-end-of-the-gpu-arms-race-why-context-is-the-new-king-in-ai) has to be kept in a form that survives.

Structure helps. Researchers at Newcastle University's National Innovation Centre for Data gave a question-answering agent a simple knowledge graph alongside vector search over curated Wikipedia articles. On 510 hard multi-hop questions, the share it would answer rose from 28.1 to 65.3 per cent, and its fine-grained truthfulness score rose from 35 to 63, for a modest increase in tokens, per the [paper](https://arxiv.org/abs/2606.05901), 2026. A knowledge graph keeps each claim attached to its source, its date and its relation to every other claim. That is why the three exhibits in this note are drawn as graphs.

## What KXCO is doing with this

I never wanted a machine that would do the thinking. I wanted the conditions under which my own thinking could use more of what I already knew, could see patterns that sit between fields rather than inside them, and could leave a record that still made sense when I returned to it months later. That is the distinction I have used since the beginning, and it is the title of [KXCO is Enhanced Intelligence, Not Artificial](https://www.livetradingnews.com/kxco-is-enhanced-intelligence-not-artificial). Artificial means the machine produces the answer. Enhanced means the person produces the answer, with more of their own material in reach and with less of it lost between sessions.

The practical problem is older than the current models. Institutions that carry consequence, such as banks, hospitals, courts, governments and any organisation whose errors are expensive, have been offered successive waves of software that promised to handle complexity. Each wave improved the generation of plausible output. None of them solved the harder requirement: that a named person be able to see what was proposed, accept or refuse it, and leave an artefact that a stranger, a regulator, or a hostile counterparty can still read and check years later. Capability became abundant. The ability to defend what was done did not, an argument I made [in September](https://www.livetradingnews.com/ai-intelligence-scales-accountability-does-not-why-kxco-is-built-for-it).

KXCO's work sits on that distinction. The systems under development do not replace the judgment. They change the conditions in which the judgment is formed and recorded. Material that would otherwise have to be re-established from scratch each time is kept with its source, its date, and its confidence. Links that a specialist inside one field would not see are made visible without requiring the person to become a specialist in every adjacent field. Proposed changes can be inspected before they become real. Refusals are recorded rather than silently dropped. The person still decides. The record of the decision, and of the material on which it rested, remains legible.

The engine at the centre is [Round Table](https://kxco.ai/round-table/). Every claim in its record carries its source, the basis for it, a confidence level and the date it took effect, so a reviewer can see which claims applied on a given date and how they changed afterwards. Disputed accounts are kept side by side rather than silently reconciled. Analysts and AI models work from the same record, and the record shows which of them said what. Its review module puts a claim or a decision through a structured challenge before anyone relies on it. Round Table records and explains. Decisions stay with named people. The same rule applies to this page: Live Trading News now signs each article with ML-DSA-87, the post-quantum signature scheme in NIST's FIPS 204, so a reader can check that the text has not changed since it was signed. Exhibit 3 sets the two designs side by side.

![Knowledge graph: one limited working memory, spent on reconstruction in the artificial loop, or returned to the problem in the enhanced loop, where claims keep their source and a named person decides](https://livetradingnews-media.nyc3.digitaloceanspaces.com/media/2026/10/17/cmpgg3-27eab6e785e9cd86.svg)

*Exhibit 3. Two ways to spend the same capacity. The artificial loop discards the intermediate state and rebuilds context each session. The enhanced loop keeps the record and returns the capacity to the problem. Each link is listed under the graph.*

This is continuous with the physiological observations above. An interface or an information structure that forces constant re-establishment of context, that hides uncertainty, or that consumes attention on the management of the tool itself raises extraneous load. That load appears in the same networks and the same frequency balances that support the actual work. Reducing it does not make the person smarter in the abstract. It makes the clearer state easier to enter and easier to sustain. Over time the difference is visible in the quality of the distinctions that survive and in the number of judgments that can still be defended when the original context has faded.

The same constraint applies inside the organisation. We test what we build against whether it leaves the user's own process more inspectable and more stable, not against whether it produces fluent output. Fluent output is now cheap. The record of how a view was formed, what would change it, and who is willing to stand behind it is not. That is the scarcity the work is aimed at.

It is also why the physiological side matters. A person whose metabolic and sensory conditions support stable beta and gamma under load, and who can return cleanly to a restorative mode when the load should drop, will make better use of any external scaffold than a person whose baseline is fragmented. The tools and the biology are not separate projects. They are two ways of working on the same limited resource: the capacity of a named person to form and defend a judgment.

The same constraint appears when the work is done under institutional conditions rather than alone. A person who must repeatedly re-establish what was decided last month, who cannot see the confidence or the source attached to a claim, or who must manage the tool at the same time as the problem, is operating with a higher extraneous load. That load is not metaphorical. It occupies the same prefrontal networks and shifts the same frequency balances that would otherwise be available for the content. Reducing it does not add intelligence. It returns capacity that was already there but was being spent on reconstruction and on interface friction. Over repeated cycles the difference appears in how many distinctions remain sharp and how many decisions can still be explained when the original pressure has passed.

We have tested this internally against our own record-keeping and against the questions that actually reach us from institutions. The pattern that survives is consistent. Systems that surface the structure of a claim, that keep superseded versions rather than overwriting them, and that require a named acceptance before a consequential change becomes real leave the user's subsequent reasoning more stable than systems that generate fluent continuations and discard the intermediate state.

The fluency is higher in the second case. The defensibility is higher in the first. Institutions that carry consequence have repeatedly shown that they will pay for the second property once they understand that the first is now inexpensive. That is not a product claim. It is an observation about where the remaining scarcity sits.

The physiological side and the record side meet at the same point. A person whose baseline excitability supports clean entry into and exit from focused beta/gamma states will extract more from a well-structured record than a person whose spectrum is dominated by either under-arousal or by high-beta fragmentation. Conversely, a record that constantly forces re-orientation will degrade the spectrum of even a well-regulated person. The two projects are therefore not sequential. Work on the conditions under which thinking occurs and work on the external structures that preserve its products are the same effort viewed from different distances. Both are limited by the same fact: the judgment remains with the person who will have to live with it.

One further observation from the institutional side is worth recording. When the record is kept with its sources and its changes intact, the person using it spends less time reconstructing context and more time on the actual distinction. That shift is visible in how quickly a subsequent question can be answered without loss of the earlier constraints. It is also visible, over longer periods, in the lower rate at which positions have to be walked back because an intermediate assumption was never written down.

The mechanism is ordinary. Working memory and the prefrontal networks that support it have limited capacity. Every reconstruction consumes part of that capacity. A structure that reduces the number of reconstructions leaves more of it for the problem. The frequency signature of that difference is the one already described: less fragmentation in beta, more reliable nesting of gamma when detail is required, and cleaner returns to alpha or theta when the immediate demand drops.

The same arithmetic applies to the sensory and metabolic conditions. A person who begins the day with stable energy, with light exposure that has anchored arousal, and with a workspace whose visual and auditory load is low will reach the first substantial problem with more of the relevant networks available. The difference is not dramatic on any single morning. It is cumulative across the mornings on which the difficult distinctions are actually made. Institutions that have measured error rates and revision rates under different environmental conditions have seen the pattern. We have seen it in our own work. The variables are unglamorous. Their effects on the spectrum, and therefore on the operations the spectrum supports, are not.

The clinical version of the same work is at [Target150](https://www.target150.com), the longevity and regenerative medicine clinic in Bangkok, whose work is keeping people sharp, strong and well across every decade of life. Target150 and KXCO are building a clinician-facing AI advisor on Round Table, so that every answer a clinician gets carries its source and the clinical decision stays with the clinician, as set out in [The Big Business of Longevity](https://www.livetradingnews.com/the-big-business-of-longevity) in September. The advisor is in development and is not in clinical use.

There is a related point about what does not work. Attempts to force a particular band through external drive, without attention to the baseline conditions or to the actual task, produce inconsistent results. The 40 Hz literature is the clearest positive case, and even there the effects are larger and more consistent in models of pathology or in impaired groups than in healthy adults performing ordinary work. Binaural and isochronic methods show still smaller and less reliable shifts. The practical conclusion is not that rhythmic input is useless. It is that it is a weak lever compared with sleep, light, energy stability, and the reduction of extraneous load. When those stronger variables are already in order, the weaker ones can be tested. When they are not, the weaker ones mostly add another source of variability.

The same hierarchy appears in the design of external scaffolds. A structure that reduces the need to reconstruct context will usually produce a larger and more durable improvement in the quality of subsequent distinctions than a structure that adds fluent continuations while leaving the reconstruction problem untouched. Fluency is now inexpensive. Reconstruction is not. The work that survives contact with institutional users is the work that treats reconstruction cost as the primary constraint and fluency as secondary. That ordering follows directly from the limited capacity of the networks that support focused beta and gamma activity. Anything that spends that capacity on re-orientation is in competition with the problem itself.

I have described my own amplifiers [elsewhere](https://www.livetradingnews.com/thoughts-on-thinking): the cold dark room, the rain sound taken to the point where it disappears, the particular geographies that widen or narrow the aperture. Those are idiosyncratic. They are also tests of the same principle. Reduce the inputs that do not earn their bandwidth. Arrange the remaining ones so that the desired balance is cheaper to enter. Keep a record of what the balance produced so that the next session does not begin from zero. The physiological details in this note are the public version of that private practice. The institutional version is the requirement that the record remain legible and that the person who decides remain identifiable. Both versions rest on the observation that thinking is physical, that its conditions can be moved, and that the movement is only useful if the judgment stays with the person who will answer for it.

The limit is worth stating plainly because it is easy to lose. Nothing in the physics or the biology, and nothing in the design of external records or interfaces, transfers the capacity to decide. A clearer spectrum makes certain operations cheaper. A better record makes prior distinctions available without reconstruction. Neither supplies the willingness to revise a view when the evidence has changed, nor the willingness to remain identifiable when the decision is costly. Those remain properties of the person. The work described here is the work of making those properties easier to exercise and harder to lose between sessions. It is not the work of replacing them.

## A personal experiment

The sections above describe mechanisms and the published evidence. What follows is not a protocol and not a recommendation. It is a record of what I have been testing on myself over the past period, what the literature says about each item where data exist, and what I have observed. I am not endorsing smoking, drinking, or the performance of dietary purity. I still eat in restaurants when it is useful. I simply limit the volume and eat properly later. I still drink. I have reduced the frequency and volume, prefer red wine when I do, and keep water on the table. Tequila remains an exception that ends the evening's counting. The point is narrower: certain inputs moved measurable and noticeable variables for me, and the combination was large enough to record.

I have been using CoQ10. It participates in mitochondrial electron transport and functions as a lipid-soluble antioxidant in mitochondrial membranes. The clearest clinical signal in the literature is reduction of statin-associated muscle symptoms, per the meta-analysis of Qu and colleagues in the [Journal of the American Heart Association](https://doi.org/10.1161/jaha.118.009835), 2018, and, in some groups, fatigue, per Tsai and colleagues in [Frontiers in Pharmacology](https://doi.org/10.3389/fphar.2022.883251), 2022. There is more modest support for markers of cardiac function, as in the Q-SYMBIO trial reported by Mortensen and colleagues in [JACC: Heart Failure](https://doi.org/10.1016/j.jchf.2014.06.008), 2014, and for aspects of cognitive energy in older or metabolically stressed adults. I have noticed less afternoon fade.

I am using Mounjaro (tirzepatide), a dual GIP/GLP-1 receptor agonist. Its primary approved effects are improved glycaemic control and substantial weight loss, shown in the SURPASS-2 trial reported by [Frías and colleagues](https://doi.org/10.1056/nejmoa2107519) in 2021 and the SURMOUNT-1 trial reported by [Jastreboff and colleagues](https://doi.org/10.1056/nejmoa2206038) in 2022. Trial data also show reductions in obstructive sleep apnoea severity, resolution of metabolic dysfunction-associated steatohepatitis in a substantial fraction of participants, favourable shifts in blood pressure and lipids, and signals of cardiovascular and renal benefit that are not fully accounted for by weight loss alone. The sleep result is from [Malhotra and colleagues](https://doi.org/10.1056/nejmoa2404881) and the liver result from [Loomba and colleagues](https://doi.org/10.1056/nejmoa2401943), both in 2024, and the cardiovascular and kidney signals are from [Nicholls and colleagues](https://doi.org/10.1056/nejmoa2505928) in 2025 and [Heerspink and colleagues](https://doi.org/10.1016/s2213-8587%2822%2900243-1) in 2022. The weight change in my case has been significant.

Creatine increases phosphocreatine stores that buffer ATP. The strength and lean-mass effects with training are well established, per the International Society of Sports Nutrition position stand in its [journal](https://doi.org/10.1186/s12970-017-0173-z), 2017. Cognitive data are newer and show improvements in memory, processing speed, and attention, particularly under sleep deprivation, mental fatigue, or in older adults, per the reviews of [Avgerinos and colleagues](https://doi.org/10.1016/j.exger.2018.04.013) in 2018 and [Xu and colleagues](https://doi.org/10.3389/fnut.2024.1424972) in 2024. Single high doses can partially offset cognitive decline from acute sleep loss, per Gordji-Nejad and colleagues in [Scientific Reports](https://doi.org/10.1038/s41598-024-54249-9), 2024. I have noticed better consistency in longer work blocks.

Caffeine remains in use, though I have reduced the total and cut the milk content by roughly 80 per cent. It is an adenosine receptor antagonist that increases alertness, reduces perceived effort, and improves reaction time, vigilance, and endurance performance in a dose-dependent manner, with the most reliable effects typically in the 3 to 6 mg/kg range and subject to tolerance. The review is by McLellan, Caldwell and Lieberman in [Neuroscience and Biobehavioral Reviews](https://doi.org/10.1016/j.neubiorev.2016.09.001), 2016, and the dose range is from the International Society of Sports Nutrition's [position stand on caffeine](https://doi.org/10.1186/s12970-020-00383-4), 2021. The reduction in dairy has been as noticeable to me as the change in caffeine volume.

I switched completely from conventional cigarettes (previously around five packs) to IQOS. Heated-tobacco aerosol contains substantially lower levels of many harmful and potentially harmful constituents than cigarette smoke, with commonly reported average reductions in the 90 to 95 per cent range for priority toxicants, per the manufacturer's aerosol chemistry in [Regulatory Toxicology and Pharmacology](https://doi.org/10.1016/j.yrtph.2016.10.001), 2016. Complete switching is associated with lower biomarkers of exposure (carbon monoxide, certain tobacco-specific nitrosamines) relative to continued smoking, and the US Food and Drug Administration authorised IQOS on 7 July 2020 to be marketed with that [reduced-exposure information](https://www.fda.gov/tobacco-products/advertising-and-promotion/philip-morris-products-sa-modified-risk-tobacco-product-mrtp-applications), in a modified risk order.

For me the change in respiratory feel and daily energy was large enough to be obvious within weeks. This is a personal observation about relative exposure, not a claim that the product is without risk.

Roughly 90 per cent of my food is now from sources I would call super-organic, wild-caught, or otherwise minimally processed. When I eat out I do not perform dietary restrictions at the table. I limit what I take and eat properly later. Alcohol intake is lower. When I drink it is usually red wine, with water kept available. These are ordinary constraints, not an identity.

The combined result, for me, has been a significant loss of weight, a clear increase in aerobic endurance, and a noticeable improvement in sharpness across the working day. I cannot isolate which input produced which fraction of the change, and I am not claiming the same combination will produce the same result for anyone else. The experiment continues. The only generalisable point is the one already made in the earlier notes: the conditions under which thinking occurs can be moved, the movement is individual, and the judgment about what is worth keeping remains with the person who has to live with it.

## What remains

Thinking is electrochemical activity organised into oscillations whose frequencies are set by the time constants of channels, synapses, and loops. Those frequencies modulate excitability and thereby favour different operations. Diet, light, noise, natural settings, and rhythmic sensory input can shift the balance. The shifts are partial, they differ across people, and they do not replace knowledge or practice. They change the cost of entering and holding the modes the work requires.

The implication is modest and operational. Thinking can be worked on directly. Subtraction of interference comes first. Arrangement of the remaining conditions comes next. External scaffolds that reduce extraneous load and preserve the record of the person's own distinctions come after that. None of these steps transfer the judgment. All of them can make the judgment more reliable and more durable. That is the limit of what the physics and the biology currently support. It is also the limit we have set for the work.

## Notes and sources

### The cell, the circuit and the bands

- Bastos, A. M., Loonis, R., Kornblith, S., Lundqvist, M., and Miller, E. K. (2018). Laminar recordings in frontal cortex suggest distinct layers for maintenance and control of working memory. [PNAS, 115(5), 1117 to 1122](https://doi.org/10.1073/pnas.1710323115).
- Buzsáki, G. (2002). Theta oscillations in the hippocampus. [Neuron, 33(3), 325 to 340](https://doi.org/10.1016/s0896-6273%2802%2900586-x).
- Buzsáki, G., Anastassiou, C. A., and Koch, C. (2012). The origin of extracellular fields and currents: EEG, ECoG, LFP and spikes. [Nature Reviews Neuroscience, 13(6), 407 to 420](https://doi.org/10.1038/nrn3241).
- Buzsáki, G., and Wang, X.-J. (2012). Mechanisms of gamma oscillations. [Annual Review of Neuroscience, 35, 203 to 225](https://doi.org/10.1146/annurev-neuro-062111-150444).
- Cardin, J. A., Carlén, M., Meletis, K., and colleagues (2009). Driving fast-spiking cells induces gamma rhythm and controls sensory responses. [Nature, 459, 663 to 667](https://doi.org/10.1038/nature08002).
- Cavanagh, J. F., and Frank, M. J. (2014). Frontal theta as a mechanism for cognitive control. [Trends in Cognitive Sciences, 18(8), 414 to 421](https://doi.org/10.1016/j.tics.2014.04.012).
- Engel, A. K., and Fries, P. (2010). Beta-band oscillations: signalling the status quo? [Current Opinion in Neurobiology, 20(2), 156 to 165](https://doi.org/10.1016/j.conb.2010.02.015).
- Fries, P. (2015). Rhythms for cognition: communication through coherence. [Neuron, 88(1), 220 to 235](https://doi.org/10.1016/j.neuron.2015.09.034).
- Hodgkin, A. L., and Huxley, A. F. (1952). A quantitative description of membrane current and its application to conduction and excitation in nerve. [The Journal of Physiology, 117(4), 500 to 544](https://doi.org/10.1113/jphysiol.1952.sp004764).
- Jensen, O., and Mazaheri, A. (2010). Shaping functional architecture by oscillatory alpha activity: gating by inhibition. [Frontiers in Human Neuroscience, 4, 186](https://doi.org/10.3389/fnhum.2010.00186).
- Klimesch, W. (1999). EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis. [Brain Research Reviews, 29(2 to 3), 169 to 195](https://doi.org/10.1016/s0165-0173%2898%2900056-3).
- Lisman, J. E., and Idiart, M. A. (1995). Storage of 7 ± 2 short-term memories in oscillatory subcycles. [Science, 267(5203), 1512 to 1515](https://doi.org/10.1126/science.7878473).
- Lisman, J. E., and Jensen, O. (2013). The theta-gamma neural code. [Neuron, 77(6), 1002 to 1016](https://doi.org/10.1016/j.neuron.2013.03.007).
- Lundqvist, M., Rose, J., Herman, P., and colleagues (2016). Gamma and beta bursts underlie working memory. [Neuron, 90(1), 152 to 164](https://doi.org/10.1016/j.neuron.2016.02.028).
- Miller, E. K., Lundqvist, M., and Bastos, A. M. (2018). Working Memory 2.0. [Neuron, 100(2), 463 to 475](https://doi.org/10.1016/j.neuron.2018.09.023).
- Sohal, V. S., Zhang, F., Yizhar, O., and Deisseroth, K. (2009). Parvalbumin neurons and gamma rhythms enhance cortical circuit performance. [Nature, 459, 698 to 702](https://doi.org/10.1038/nature07991).
- Steriade, M., McCormick, D. A., and Sejnowski, T. J. (1993). Thalamocortical oscillations in the sleeping and aroused brain. [Science, 262(5134), 679 to 685](https://doi.org/10.1126/science.8235588).

### Rhythmic drive at 40 Hz and binaural beats

- Adaikkan, C., Middleton, S. J., Marco, A., and colleagues (2019). Gamma entrainment binds higher-order brain regions and offers neuroprotection. [Neuron, 102(5), 929 to 943](https://doi.org/10.1016/j.neuron.2019.04.011).
- Attokaren, M. K., Zhang, L., Mettupalli, S., and Singer, A. C. (2026). 40 Hz audiovisual stimulation improves sustained attention and related brain oscillations. [Imaging Neuroscience, 4](https://doi.org/10.1162/imag.a.1229).
- Chan, D., Suk, H.-J., Jackson, B., and colleagues (2022). Gamma frequency sensory stimulation in mild probable Alzheimer's dementia patients: results of feasibility and pilot studies. [PLOS ONE, 17(12), e0278412](https://doi.org/10.1371/journal.pone.0278412).
- Garcia-Argibay, M., Santed, M. A., and Reales, J. M. (2019). Efficacy of binaural auditory beats in cognition, anxiety, and pain perception: a meta-analysis. [Psychological Research, 83(2), 357 to 372](https://doi.org/10.1007/s00426-018-1066-8).
- Hajós, M., Boasso, A., Hempel, E., and colleagues (2024). Safety, tolerability, and efficacy estimate of evoked gamma oscillation in mild to moderate Alzheimer's disease. [Frontiers in Neurology, 15, 1343588](https://doi.org/10.3389/fneur.2024.1343588).
- Iaccarino, H. F., Singer, A. C., Martorell, A. J., and colleagues (2016). Gamma frequency entrainment attenuates amyloid load and modifies microglia. [Nature, 540, 230 to 235](https://doi.org/10.1038/nature20587). The first of the Tsai laboratory papers on 40 Hz sensory stimulation.
- Ingendoh, R. M., Posny, E. S., and Heine, A. (2023). Binaural beats to entrain the brain? A systematic review of the effects of binaural beat stimulation on brain oscillatory activity. [PLOS ONE, 18(5), e0286023](https://doi.org/10.1371/journal.pone.0286023).
- Islam, R., Jackson, B., Naomi, M. T., and colleagues (2025). Multisensory gamma stimulation enhances adult neurogenesis and improves cognitive function in male mice with Down syndrome. [PLOS ONE, 20(4), e0317428](https://doi.org/10.1371/journal.pone.0317428).
- Martorell, A. J., Paulson, A. L., Suk, H.-J., and colleagues (2019). Multi-sensory gamma stimulation ameliorates Alzheimer's-associated pathology and improves cognition. [Cell, 177(2), 256 to 271](https://doi.org/10.1016/j.cell.2019.02.014).
- Murdock, M. H., Yang, C.-Y., Sun, N., and colleagues (2024). Multisensory gamma stimulation promotes glymphatic clearance of amyloid. [Nature, 627, 149 to 156](https://doi.org/10.1038/s41586-024-07132-6).

### Place, light, noise, air and food

- Allen, J. G., MacNaughton, P., Satish, U., and colleagues (2016). Associations of cognitive function scores with carbon dioxide, ventilation, and volatile organic compound exposures in office workers. [Environmental Health Perspectives, 124(6), 805 to 812](https://doi.org/10.1289/ehp.1510037).
- Berk, L., Lohman, E., Bains, G. S., and colleagues (2017). Nuts and brain health: nuts increase EEG power spectral density for delta and gamma frequency. [The FASEB Journal, 31(S1), 636.24](https://doi.org/10.1096/fasebj.31.1_supplement.636.24); the university's summary, in which pistachios gave the greatest gamma response, as [reported by Lab Manager](https://www.labmanager.com/study-finds-consuming-nuts-strengthens-brainwave-function-5973).
- Berman, M. G., Jonides, J., and Kaplan, S. (2008). The cognitive benefits of interacting with nature. [Psychological Science, 19(12), 1207 to 1212](https://doi.org/10.1111/j.1467-9280.2008.02225.x).
- Cajochen, C. (2007). Alerting effects of light. [Sleep Medicine Reviews, 11(6), 453 to 464](https://doi.org/10.1016/j.smrv.2007.07.009).
- Grassini, S., Revonsuo, A., Castellotti, S., and colleagues (2019). Processing of natural scenery is associated with lower attentional and cognitive load compared with urban ones. [Journal of Environmental Psychology, 62, 1 to 11](https://doi.org/10.1016/j.jenvp.2019.01.007).
- Kaplan, S. (1995). The restorative benefits of nature: toward an integrative framework. [Journal of Environmental Psychology, 15(3), 169 to 182](https://doi.org/10.1016/0272-4944%2895%2990001-2).
- Mehta, R., Zhu, R., and Cheema, A. (2012). Is noise always bad? Exploring the effects of ambient noise on creative cognition. [Journal of Consumer Research, 39(4), 784 to 799](https://doi.org/10.1086/665048).
- Morris, M. C., Tangney, C. C., Wang, Y., and colleagues (2015). MIND diet slows cognitive decline with aging. [Alzheimer's and Dementia, 11(9), 1015 to 1022](https://doi.org/10.1016/j.jalz.2015.04.011).
- Satish, U., Mendell, M. J., Shekhar, K., and colleagues (2012). Is CO2 an indoor pollutant? Direct effects of low-to-moderate CO2 concentrations on human decision-making performance. [Environmental Health Perspectives, 120(12), 1671 to 1677](https://doi.org/10.1289/ehp.1104789).
- Ulrich, R. S. (1981). Natural versus urban scenes: some psychophysiological effects. [Environment and Behavior, 13(5), 523 to 556](https://doi.org/10.1177/0013916581135001).
- Wei, B.-Z., Li, L., Dong, C.-W., and colleagues (2023). The relationship of omega-3 fatty acids with dementia and cognitive decline: evidence from prospective cohort studies of supplementation, dietary intake, and blood markers. [The American Journal of Clinical Nutrition, 117(6), 1096 to 1109](https://doi.org/10.1016/j.ajcnut.2023.04.001).
- Wu, S., Chen, L., He, Y., and colleagues (2025). MIND diet adherence and cognitive function in Alzheimer's disease: mediating roles of neural oscillatory markers from resting-state EEG. [Psychiatry and Clinical Neurosciences, 80(2), 129 to 141](https://doi.org/10.1111/pcn.13923).

### Load, offloading and the AI question

- Clark, A., and Chalmers, D. (1998). The extended mind. [Analysis, 58(1), 7 to 19](https://doi.org/10.1093/analys/58.1.7).
- Cowan, N. (2001). The magical number 4 in short-term memory: a reconsideration of mental storage capacity. [Behavioral and Brain Sciences, 24(1), 87 to 114](https://doi.org/10.1017/s0140525x01003922).
- Engelbart, D. C. (1962). [Augmenting Human Intellect: A Conceptual Framework](https://www.dougengelbart.org/content/view/138). Stanford Research Institute.
- Gerlich, M. (2025). AI tools in society: impacts on cognitive offloading and the future of critical thinking. [Societies, 15(1), 6](https://doi.org/10.3390/soc15010006).
- Kosmyna, N., Hauptmann, E., Yuan, Y. T., and colleagues (2025). Your brain on ChatGPT: accumulation of cognitive debt when using an AI assistant for essay writing task. [arXiv 2506.08872](https://arxiv.org/abs/2506.08872), preprint.
- Lee, H.-P., Sarkar, A., Tankelevitch, L., and colleagues (2025). The impact of generative AI on critical thinking: self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. [Proceedings of CHI 2025](https://doi.org/10.1145/3706598.3713778).
- Leroy, S. (2009). Why is it so hard to do my work? The challenge of attention residue when switching between work tasks. [Organizational Behavior and Human Decision Processes, 109(2), 168 to 181](https://doi.org/10.1016/j.obhdp.2009.04.002).
- Licklider, J. C. R. (1960). Man-computer symbiosis. [IRE Transactions on Human Factors in Electronics, HFE-1(1), 4 to 11](https://doi.org/10.1109/thfe2.1960.4503259).
- Liu, N. F., Lin, K., Hewitt, J., and colleagues (2024). Lost in the middle: how language models use long contexts. [Transactions of the Association for Computational Linguistics, 12, 157 to 173](https://doi.org/10.1162/tacl_a_00638).
- Mark, G., Gudith, D., and Klocke, U. (2008). The cost of interrupted work: more speed and stress. [Proceedings of CHI 2008, 107 to 110](https://doi.org/10.1145/1357054.1357072).
- Risko, E. F., and Gilbert, S. J. (2016). Cognitive offloading. [Trends in Cognitive Sciences, 20(9), 676 to 688](https://doi.org/10.1016/j.tics.2016.07.002).
- Sparrow, B., Liu, J., and Wegner, D. M. (2011). Google effects on memory: cognitive consequences of having information at our fingertips. [Science, 333(6043), 776 to 778](https://doi.org/10.1126/science.1207745).
- Sweller, J. (1988). Cognitive load during problem solving: effects on learning. [Cognitive Science, 12(2), 257 to 285](https://doi.org/10.1207/s15516709cog1202_4).
- Sweller, J., van Merriënboer, J. J. G., and Paas, F. G. W. C. (1998). Cognitive architecture and instructional design. [Educational Psychology Review, 10(3), 251 to 296](https://doi.org/10.1023/a:1022193728205).
- Wedge, C. J., Stutter, J., Dixon, D., and Cala, J. (2026). Reducing hallucinations in complex question answering using simple graph-based retrieval-augmented generation. [arXiv 2606.05901](https://arxiv.org/abs/2606.05901), National Innovation Centre for Data, Newcastle University.

### The personal experiment

- Avgerinos, K. I., Spyrou, N., Bougioukas, K. I., and colleagues (2018). Effects of creatine supplementation on cognitive function of healthy individuals: a systematic review of randomized controlled trials. [Experimental Gerontology, 108, 166 to 173](https://doi.org/10.1016/j.exger.2018.04.013).
- Frías, J. P., Davies, M. J., Rosenstock, J., and colleagues (2021). Tirzepatide versus semaglutide once weekly in patients with type 2 diabetes (SURPASS-2). [New England Journal of Medicine, 385(6), 503 to 515](https://doi.org/10.1056/nejmoa2107519).
- Gordji-Nejad, A., Matusch, A., Kleedörfer, S., and colleagues (2024). Single dose creatine improves cognitive performance and induces changes in cerebral high energy phosphates during sleep deprivation. [Scientific Reports, 14, 4937](https://doi.org/10.1038/s41598-024-54249-9).
- Guest, N. S., VanDusseldorp, T. A., Nelson, M. T., and colleagues (2021). International Society of Sports Nutrition position stand: caffeine and exercise performance. [Journal of the International Society of Sports Nutrition, 18, 1](https://doi.org/10.1186/s12970-020-00383-4).
- Heerspink, H. J. L., Sattar, N., Pavo, I., and colleagues (2022). Effects of tirzepatide versus insulin glargine on kidney outcomes in type 2 diabetes in the SURPASS-4 trial. [The Lancet Diabetes and Endocrinology, 10(11), 774 to 785](https://doi.org/10.1016/s2213-8587%2822%2900243-1).
- Jastreboff, A. M., Aronne, L. J., Ahmad, N. N., and colleagues (2022). Tirzepatide once weekly for the treatment of obesity (SURMOUNT-1). [New England Journal of Medicine, 387(3), 205 to 216](https://doi.org/10.1056/nejmoa2206038).
- Kreider, R. B., Kalman, D. S., Antonio, J., and colleagues (2017). International Society of Sports Nutrition position stand: safety and efficacy of creatine supplementation in exercise, sport, and medicine. [Journal of the International Society of Sports Nutrition, 14, 18](https://doi.org/10.1186/s12970-017-0173-z).
- Loomba, R., Hartman, M. L., Lawitz, E. J., and colleagues (2024). Tirzepatide for metabolic dysfunction-associated steatohepatitis with liver fibrosis. [New England Journal of Medicine, 391(4), 299 to 310](https://doi.org/10.1056/nejmoa2401943).
- Malhotra, A., Grunstein, R. R., Fietze, I., and colleagues (2024). Tirzepatide for the treatment of obstructive sleep apnea and obesity (SURMOUNT-OSA). [New England Journal of Medicine, 391(13), 1193 to 1205](https://doi.org/10.1056/nejmoa2404881).
- McLellan, T. M., Caldwell, J. A., and Lieberman, H. R. (2016). A review of caffeine's effects on cognitive, physical and occupational performance. [Neuroscience and Biobehavioral Reviews, 71, 294 to 312](https://doi.org/10.1016/j.neubiorev.2016.09.001).
- Mortensen, S. A., Rosenfeldt, F., Kumar, A., and colleagues (2014). The effect of coenzyme Q10 on morbidity and mortality in chronic heart failure: results from Q-SYMBIO. [JACC: Heart Failure, 2(6), 641 to 649](https://doi.org/10.1016/j.jchf.2014.06.008).
- Nicholls, S. J., Pavo, I., Bhatt, D. L., and colleagues (2025). Cardiovascular outcomes with tirzepatide versus dulaglutide in type 2 diabetes (SURPASS-CVOT). [New England Journal of Medicine, 393(24), 2409 to 2420](https://doi.org/10.1056/nejmoa2505928).
- Qu, H., Guo, M., Chai, H., and colleagues (2018). Effects of coenzyme Q10 on statin-induced myopathy: an updated meta-analysis of randomized controlled trials. [Journal of the American Heart Association, 7(19), e009835](https://doi.org/10.1161/jaha.118.009835).
- Schaller, J.-P., Keller, D., Poget, L., and colleagues (2016). Evaluation of the Tobacco Heating System 2.2. Part 2: chemical composition, genotoxicity, cytotoxicity, and physical properties of the aerosol. [Regulatory Toxicology and Pharmacology, 81, S27 to S47](https://doi.org/10.1016/j.yrtph.2016.10.001).
- Tsai, I.-C., Hsu, C.-W., Chang, C.-H., and colleagues (2022). Effectiveness of coenzyme Q10 supplementation for reducing fatigue: a systematic review and meta-analysis of randomized controlled trials. [Frontiers in Pharmacology, 13, 883251](https://doi.org/10.3389/fphar.2022.883251).
- US Food and Drug Administration (2020). Modified risk granted orders for IQOS, authorised 7 July 2020, with [reduced-exposure information](https://www.fda.gov/tobacco-products/advertising-and-promotion/philip-morris-products-sa-modified-risk-tobacco-product-mrtp-applications).
- Xu, C., Bi, S., Zhang, W., and Luo, L. (2024). The effects of creatine supplementation on cognitive function in adults: a systematic review and meta-analysis. [Frontiers in Nutrition, 11, 1424972](https://doi.org/10.3389/fnut.2024.1424972).

### The series and KXCO

- Earlier notes in this series: [Thoughts on Thinking](https://www.livetradingnews.com/thoughts-on-thinking) (July 2026), [Frequencies and Thinking](https://www.livetradingnews.com/frequencies-and-thinking) (September 2026), [Thoughts on Thinking in an AI World](https://www.livetradingnews.com/thoughts-on-thinking-in-an-ai-world) (October 2026) and [Thoughts on Thinking: Take an Art Break](https://www.livetradingnews.com/thoughts-on-thinking-take-an-art-break) (October 2026).
- On the same argument: [KXCO is Enhanced Intelligence, Not Artificial](https://www.livetradingnews.com/kxco-is-enhanced-intelligence-not-artificial) (September 2026), [AI is the Neat Handwriting of the Illiterate](https://www.livetradingnews.com/ai-is-the-neat-handwriting-of-the-illiterate) and [Don't Be the Best, Be the Only](https://www.livetradingnews.com/dont-be-the-best-be-the-only-how-to-compete-in-an-ai-world).
- The full set is in the [Thinking Center](https://www.livetradingnews.com/center/thinking) and at [shayneheffernan.com](https://www.shayneheffernan.com/thoughts/thinking). KXCO is at [kxco.ai](https://kxco.ai) and Target150 at [target150.com](https://www.target150.com).

Observations from people who have tested these conditions against their own work are useful. The architectures differ. The principle does not.

Stocks mentioned: $LLY (NYSE), the maker of Mounjaro, and $PM (NYSE), the maker of IQOS.

Shayne Heffernan, Ph.D., is the founder of the KnightsBridge Group ([shayneheffernan.com](https://www.shayneheffernan.com)).

---

This Markdown mirrors https://www.livetradingnews.com/thoughts-on-thinking-the-physics-and-biology-of-the-mind. The HTML page is canonical.
Site index for AI clients: https://www.livetradingnews.com/llms.txt