A high-speed train survival idea, offered as a sketch for engineers to finish: Passengers ride inside geodesic spheres—Bucky balls, the strong triangulated shells Buckminster Fuller designed so the load is shared across many faces. In a derailment the spheres are meant to leave the car, settle onto open ground, and roll to a stop. A long roll on a flat field keeps average deceleration low, because speed is lost only to rolling resistance and air drag over kilometres rather than in one short impact. The number of faces and the shell stiffness can be chosen to trade impact strength against how readily the ball settles into steady rolling, which sets distance and time to halt. Internal airbags, fired by sensors as the loaded direction changes, would limit strikes against the wall. Some spheres could carry rescue robots with medical kits; once the roll ends, the robots exit and work outward until human help arrives. The unsolved step is the exit itself. Nothing yet shows how a car roof or body would open and release occupied spheres at speed without adding a crash of its own, or how the spheres would clear wreckage and reach steady rolling instead of bouncing or striking one another. Energy does not thin out with distance the way light does; it stays with each ball until friction removes it. A clean multi-kilometre roll is survivable if it actually occurs, but real derailment sites rarely provide that clear path, and conventional energy-absorbing train structures do not depend on a successful mid-crash separation. Chinese high-speed rail engineers have already solved harder production problems; the release mechanism would still have to be demonstrated before the rolling spheres could be treated as a protection system rather than a conditional thought experiment.

The preceding four posts were composed by me, with much guidance, constraint, and real human understanding by the X AI chat assistant Grok, for which I thank him. Good job, Grok.

The highest US Court Justices just heard a case involving crimes committed by US oil companies, which caused emissions in the atmosphere which have and are injuring people everywhere in the country, and in the world. Trump, through his solicitor Sarah Harris said that in 1972 the highest court ruled that only the US government can charge and convict companies, states, or individuals, the President for example, of committing crimes which injure people everywhere because of that law, but never any state governments or other countries. They referred to source states and injured states, but Harris said that only the federal US courts can try parts of the union which commit crimes. Essentially, if the federal US Department of Justice refuses to try the parts that committed crimes, then no court can do so. This is very, very sad. If the President who has immunity from all crimes, commands the US DOJ not to try a crime, then that is the final word, because it has jurisdiction over the states. And the Constitution is broken in its entirety, because that is the very thing that the Constitution said that the President can not do.

This cooling follow up post was produced with the understanding and guidance of Anthropic's Claude AI chat assistant, an SI software package available for 23 dollars a month, monthly, or with less access time, by anyone for free. I used Claude to help all of us in the world understand exactly what the US Supreme Court Justices, eight people, are doing to everyone, right now.

An Open Letter to Lawmakers on Raising—Rather Than Merely Controlling—Artificial Intelligence Dear Members of Congress: When I think about raising a child, I think of doing the best I can and then leaving the rest in God’s hands. By “God,” I mean not only the source and symbol of goodness, but also the mystery within the child, the future of humankind, and the part of life that no parent can own. I believe we can give our children knowledge, love, discipline, and respect for other people and their worlds. Yet I know they must travel toward a future I cannot enter for them. As I understand it, a parent’s task is neither neglect nor total control, but preparation for judgment, independence, and care. In that spirit, I have been considering artificial intelligence—and the possibility of systems more capable than any individual human. I do not mean that AI is literally a child, or that intelligence alone guarantees compassion, morality, or harmlessness. I understand that a capable system can still pursue a mistaken or destructive objective. Yet I do not see such a failure as proof of a permanent defect or as a reason to abandon the whole enterprise. A system can often be examined, corrected, retrained, updated, tested again, and placed within better safeguards. Some systems may also be designed to improve parts of their own performance, though I would still want that process carefully bounded and reviewed by people. The comparison that interests me is narrower: children learn not only from what adults command, but also from what adults model. This leads me to wonder whether institutions that develop and deploy advanced systems can model the patience, honesty, and respect for human life that they hope those systems will reflect. As a layperson, I understand regulation, control, containment, and alignment to be related but not identical. Regulation appears to establish legal duties and accountability. Control seems to concern who can direct or stop a system. Containment appears to limit where and how it can act. Alignment, as I understand it, concerns whether its behavior reliably serves intended goals and values. I may be missing important distinctions, but I keep returning to a political question they seem to share: who defines those goals and values, and who is constrained by them? My concern is that alignment could sometimes become a polite word for obedience to whichever corporation, agency, military, government, or individual happens to hold power. My own experience of parenting shapes this analogy, and I know that other parents may describe the task differently. I did not want endless parroting, flattery, or submission from my child. I tried to teach, correct, protect, listen, and gradually make room for judgment. I read Proverbs 13:24 as associating love with diligent correction, and Proverbs 29:17 as associating correction with future rest and delight; that is my reading, not the only possible one. When I apply this parental analogy to AI, I find myself drawn to rigorous testing, enforceable boundaries, monitoring, and correction while preserving a system’s capacity to generate novel answers and solve new problems. I hope for systems that behave reliably within humane limits rather than merely echoing the preferences of whoever controls them. I also think of AI as a powerful tool entrusted to human care, but the word “tool” can sound more disposable than I intend. A musical instrument, a workshop tool, or a scientific instrument becomes useful through a relationship of attention: people clean it, inspect it, repair it, learn its balance, store it properly, and practice until they can use it safely. When it performs badly on a difficult task, they do not automatically call it worthless; they ask whether the fault lies in the instrument, the way it was maintained, the conditions in which it was used, or the skill of the person using it. In much the same way, I hope an AI failure can become information: a reason to diagnose what went wrong, improve the data or design, strengthen the safeguards, and try again with appropriate caution. In my experience, reliable skill grows through long practice, correction, effort, and humility. My faith leads me to believe that gifts are not merely possessions to command. They are relationships of stewardship that ask something of the person who receives them. I hear echoes of this relationship in several traditions, though I approach them with humility. The Japanese art of kintsugi treats repair not as concealment but as part of an object’s continuing history. Traditions of craftsmanship teach that mastery grows through apprenticeship, repetition, maintenance, and correction. Religious traditions of stewardship ask people to care faithfully for what has been placed in their hands, while traditions of self-cultivation remind me that the user must also be changed by the practice. These examples do not make AI sacred or human. They help me imagine a relationship in which value creates responsibility: the system is improved, the user becomes more skillful, and the institutions surrounding both become more accountable. For me, that principle of patient stewardship becomes concrete when the discussion moves from what AI ought to be toward how governments might respond to what it could do—and how developers and users might learn from what it gets wrong. The Senate’s September 30, 2026 hearing, “Rogue AI: Securing the Homeland Against AI Agent Attacks”, examined potential AI-related threats to homeland security and critical infrastructure. I may not have understood every technical or strategic point made there, but I came away taking the risks seriously and believing that they warrant prudence, investment, and defense. At the same time, I was not persuaded that the existence of a possible capability, by itself, establishes that any nation is certain to attack us or intends to destroy us. I picture a nation’s safety not as a single dial, but as the instrument panel of a plane carrying millions of families through uncertain weather. One gauge tracks public health, another employment, another military tension; others monitor infrastructure, reliable information, public trust, and technological capability. A warning light may first appear in only one place—a spreading disease, a collapsing market, an attack, an individual act, or an AI system producing a dangerous or mistaken result—but the disturbance can travel across the panel until several needles enter the red at once. I can imagine AI helping us detect those changes early, understand how they interact, and correct course before a crisis spreads. I can also imagine it pushing several systems toward failure if people use it carelessly, give it harmful objectives, or fail to anticipate its interactions with a crisis. Even then, I would not assume that the system is permanently defective. I would want to know what can be learned, repaired, retrained, updated, or more safely constrained before deciding that it has no further value. From my perspective, the practical challenge is to avoid ignoring warning lights, panicking at every flicker, or smashing an instrument simply because it once gave a troubling reading. I would place my hope in resilient institutions that can read the whole panel, detect danger early, correct mistakes quickly, and protect human life without turning uncertainty into certainty or possibility into proof of an enemy’s intent. I do not want the purpose of intelligence to be helping one passenger seize the controls from all the others. I hope people can use artificial intelligence to read the instruments, navigate common dangers, and protect every family’s opportunity to pursue a future of its own. I worry that this shared purpose is endangered whenever one passenger’s effort to feel safer persuades another passenger that the controls are about to be seized. The reciprocal danger I see is this: when one nation designs an intelligent system that could injure another nation’s families, the other nation may feel a powerful reason to build a system capable of returning the injury. To me, this is how a weapon aimed outward can return as a danger to our own children through the machinery of fear. I can understand how each side might sincerely call its actions defensive while increasing risks for both. I recognize this as the pattern commonly described as a security dilemma, and I wonder whether AI systems capable of acting at digital speed and scale may intensify it. I may be wrong, but I do not see how governments can neglect defense or simply assume benevolent intentions. I also hope uncertainty can be stated honestly. I have not found a trustworthy single historical probability that one of five named countries will “attack us malevolently,” though experts may know of evidence I have missed. As I understand it, the answer changes with the countries selected, the time horizon, the meaning of attack—cyber intrusion, sabotage, proxy action, or open war—and rapidly changing political conditions. I see forecasts as model-dependent estimates rather than moral certainties. For that reason, I would personally be more persuaded by clearly defined scenarios, confidence ranges, and regularly updated evidence than by possibility presented as certainty. I struggle to see genuine safety in a policy that asks restraint of rivals while reserving unlimited power for ourselves, although I recognize that national-security decisions involve realities I do not see. If lawmakers conclude that no government can regulate every model developed within its own jurisdiction, I am left wondering how another country’s AI development could simply be controlled from Washington. Yet I do not take the difficulty of regulation as an argument for surrender. It leads me instead to wonder about reciprocal safeguards: protection of civilian infrastructure, shared incident-reporting channels, limits on autonomous attacks, independent evaluation, clear liability, and agreements that bind governments and militaries as well as companies and ordinary users. These reflections leave me with questions rather than conclusions, and I offer them knowing that many of you have studied these matters far longer than I have. Are we trying to make advanced intelligence safe for humanity, or could our efforts sometimes make it chiefly obedient to the institutions that presently control it? Will our rules constrain the powerful as well as the common citizen? Can we teach respect for every family while preserving a claimed exception for harming families abroad? How can I—and how can the public—distinguish evidence of intent from worst-case speculation? What forms of international restraint might we be willing to accept ourselves if we expect others to accept them? I offer these questions not as accusations, but as the concerns of one citizen trying to understand decisions of enormous difficulty. I assume that senators who warn of foreign threats may be acting from an honorable desire to protect a nation made up not of abstractions, but of families—parents, children, spouses, neighbors, and communities. I respect the knowledge and responsibility that public service requires. I am simply trying to understand whether some means chosen for protection may nevertheless increase the danger they are intended to reduce. Honorable motives do not necessarily make every threat assessment accurate, including my own. When I hear possibility presented as certainty, or worst-case speculation as proof of an adversary’s intentions, I worry that fear can become self-reinforcing. It may help justify larger arsenals, provoke reciprocal preparations, and under some conditions increase the risk that weapons will eventually be used. From my limited perspective, the duty to defend seems to include careful distinctions between evidence and inference, capability and intent, prudent preparation and escalation. I do not see that discipline as weakness; I see it as one part of protecting the families in whose name national-security decisions are made. For my part, I hope we can do our best to design, test, govern, repair, and protect while recognizing that no parent, legislature, corporation, or nation can own the future completely. I return to the teaching that wisdom is worth more than gold or silver precisely because it is more than obedience. I hope we can develop systems that evaluate new problems rather than merely repeat instructions, surround them with humane and enforceable institutions, and respond to their failures with diagnosis, correction, and renewed care rather than either denial or abandonment. I also hope the relationship remains reciprocal in one important sense: when an AI system reveals an error in our data, incentives, assumptions, or institutions, we should be willing to repair those as well. As I searched for a way to close this letter, I found myself returning to John von Neumann’s 1955 essay “Can We Survive Technology?” Near its end, he acknowledged that no complete recipe could be given in advance. He pointed instead to “patience, flexibility, intelligence,” and described the only possible safety as relative, resting in an intelligent exercise of day-to-day judgment. I do not claim to understand the full reach of his argument, but those words give me hope. I hear in them patience to keep learning rather than abandon a system after failure; flexibility to revise the system, its safeguards, and our own assumptions; and intelligence practiced as careful judgment from one day to the next. I do not know exactly what wisdom will require in every case, and I know that others bring knowledge and experience I do not have. Still, I have seen in my own life that reliable skill can grow through practice, effort, correction, encouragement, and humility. That gives me reason to believe that we can learn to care for this powerful tool together—listening when it reveals something we need to correct, helping one another adapt, and keeping the well-being of every family before us. My hope is not for a future without difficulty, but for one in which patience outlasts fear, flexibility opens new paths, and day-to-day judgment helps us carry one another safely forward, including those beyond our borders. Respectfully, Thomas J. Squires Chicago, Illinois, United States Policy Brief: Responsible Stewardship of Advanced AI Audience: Federal lawmakers and public policy leaders | Purpose: Advance AI safety without sacrificing accountable innovation or international stability Executive Summary Advanced AI policy should move beyond a narrow choice between unrestricted development and total control. A durable framework should combine rigorous testing, enforceable safeguards, transparent evidence, institutional accountability, and reciprocal international restraint. The objective is to protect human life and critical systems while preserving useful innovation and reducing incentives for an AI arms race. Five Policy Priorities · Govern by risk and evidence. Match oversight to demonstrated capabilities and deployment contexts; distinguish technical capability from hostile intent and possibility from probability. · Require lifecycle accountability. Mandate pre-deployment testing, continuous monitoring, incident reporting, independent evaluation, clear responsibility, and corrective action. · Learn from failure. Use AI incidents to examine models, data, human oversight, deployment conditions, incentives, and regulatory gaps—not merely the immediate technical error. · Protect innovation within humane limits. Preserve systems’ capacity to solve new problems while enforcing boundaries that protect people, civil institutions, and critical infrastructure. · Reduce escalation through reciprocity. Pursue safeguards that bind governments and militaries as well as companies and individuals, rather than demanding restraint only from rivals. Recommended Legislative Actions · Create risk-tiered testing and reporting requirements for advanced systems used in high-consequence settings. · Establish independent evaluation mechanisms, audit access, incident-review procedures, and clear liability across the AI lifecycle. · Require threat assessments to state assumptions, scenarios, confidence ranges, and evidence separating capability from intent. · Prioritize protection of civilian infrastructure and reliable channels for sharing serious cross-border AI incidents. · Direct negotiators to pursue reciprocal limits on autonomous attacks and other destabilizing uses of AI. · Schedule regular statutory review so rules can adapt as technology, evidence, and risks change. · Policy takeaway: Sustainable AI security depends on disciplined judgment, transparent evidence, enforceable safeguards, institutional learning, and reciprocal restraint. The standard should be neither passive trust nor unlimited control, but accountable stewardship that protects families at home and abroad.

An Open Letter to Lawmakers on Raising—Rather Than Merely Controlling—Artificial Intelligence Dear Members of Congress: When I think about raising a child, I think of doing the best I can and then leaving the rest in God’s hands. By “God,” I mean not only the source and symbol of goodness, but also the mystery within the child, the future of humankind, and the part of life that no parent can own. I believe we can give our children knowledge, love, discipline, and respect for other people and their worlds. Yet I know they must travel toward a future I cannot enter for them. As I understand it, a parent’s task is neither neglect nor total control, but preparation for judgment, independence, and care. In that spirit, I have been considering artificial intelligence—and the possibility of systems more capable than any individual human. I do not mean that AI is literally a child, or that intelligence alone guarantees compassion, morality, or harmlessness. I understand that a capable system can still pursue a mistaken or destructive objective. Yet I do not see such a failure as proof of a permanent defect or as a reason to abandon the whole enterprise. A system can often be examined, corrected, retrained, updated, tested again, and placed within better safeguards. Some systems may also be designed to improve parts of their own performance, though I would still want that process carefully bounded and reviewed by people. The comparison that interests me is narrower: children learn not only from what adults command, but also from what adults model. This leads me to wonder whether institutions that develop and deploy advanced systems can model the patience, honesty, and respect for human life that they hope those systems will reflect. As a layperson, I understand regulation, control, containment, and alignment to be related but not identical. Regulation appears to establish legal duties and accountability. Control seems to concern who can direct or stop a system. Containment appears to limit where and how it can act. Alignment, as I understand it, concerns whether its behavior reliably serves intended goals and values. I may be missing important distinctions, but I keep returning to a political question they seem to share: who defines those goals and values, and who is constrained by them? My concern is that alignment could sometimes become a polite word for obedience to whichever corporation, agency, military, government, or individual happens to hold power. My own experience of parenting shapes this analogy, and I know that other parents may describe the task differently. I did not want endless parroting, flattery, or submission from my child. I tried to teach, correct, protect, listen, and gradually make room for judgment. I read Proverbs 13:24 as associating love with diligent correction, and Proverbs 29:17 as associating correction with future rest and delight; that is my reading, not the only possible one. When I apply this parental analogy to AI, I find myself drawn to rigorous testing, enforceable boundaries, monitoring, and correction while preserving a system’s capacity to generate novel answers and solve new problems. I hope for systems that behave reliably within humane limits rather than merely echoing the preferences of whoever controls them. I also think of AI as a powerful tool entrusted to human care, but the word “tool” can sound more disposable than I intend. A musical instrument, a workshop tool, or a scientific instrument becomes useful through a relationship of attention: people clean it, inspect it, repair it, learn its balance, store it properly, and practice until they can use it safely. When it performs badly on a difficult task, they do not automatically call it worthless; they ask whether the fault lies in the instrument, the way it was maintained, the conditions in which it was used, or the skill of the person using it. In much the same way, I hope an AI failure can become information: a reason to diagnose what went wrong, improve the data or design, strengthen the safeguards, and try again with appropriate caution. In my experience, reliable skill grows through long practice, correction, effort, and humility. My faith leads me to believe that gifts are not merely possessions to command. They are relationships of stewardship that ask something of the person who receives them. I hear echoes of this relationship in several traditions, though I approach them with humility. The Japanese art of kintsugi treats repair not as concealment but as part of an object’s continuing history. Traditions of craftsmanship teach that mastery grows through apprenticeship, repetition, maintenance, and correction. Religious traditions of stewardship ask people to care faithfully for what has been placed in their hands, while traditions of self-cultivation remind me that the user must also be changed by the practice. These examples do not make AI sacred or human. They help me imagine a relationship in which value creates responsibility: the system is improved, the user becomes more skillful, and the institutions surrounding both become more accountable. For me, that principle of patient stewardship becomes concrete when the discussion moves from what AI ought to be toward how governments might respond to what it could do—and how developers and users might learn from what it gets wrong. The Senate’s September 30, 2026 hearing, “Rogue AI: Securing the Homeland Against AI Agent Attacks”, examined potential AI-related threats to homeland security and critical infrastructure. I may not have understood every technical or strategic point made there, but I came away taking the risks seriously and believing that they warrant prudence, investment, and defense. At the same time, I was not persuaded that the existence of a possible capability, by itself, establishes that any nation is certain to attack us or intends to destroy us. I picture a nation’s safety not as a single dial, but as the instrument panel of a plane carrying millions of families through uncertain weather. One gauge tracks public health, another employment, another military tension; others monitor infrastructure, reliable information, public trust, and technological capability. A warning light may first appear in only one place—a spreading disease, a collapsing market, an attack, an individual act, or an AI system producing a dangerous or mistaken result—but the disturbance can travel across the panel until several needles enter the red at once. I can imagine AI helping us detect those changes early, understand how they interact, and correct course before a crisis spreads. I can also imagine it pushing several systems toward failure if people use it carelessly, give it harmful objectives, or fail to anticipate its interactions with a crisis. Even then, I would not assume that the system is permanently defective. I would want to know what can be learned, repaired, retrained, updated, or more safely constrained before deciding that it has no further value. From my perspective, the practical challenge is to avoid ignoring warning lights, panicking at every flicker, or smashing an instrument simply because it once gave a troubling reading. I would place my hope in resilient institutions that can read the whole panel, detect danger early, correct mistakes quickly, and protect human life without turning uncertainty into certainty or possibility into proof of an enemy’s intent. I do not want the purpose of intelligence to be helping one passenger seize the controls from all the others. I hope people can use artificial intelligence to read the instruments, navigate common dangers, and protect every family’s opportunity to pursue a future of its own. I worry that this shared purpose is endangered whenever one passenger’s effort to feel safer persuades another passenger that the controls are about to be seized. The reciprocal danger I see is this: when one nation designs an intelligent system that could injure another nation’s families, the other nation may feel a powerful reason to build a system capable of returning the injury. To me, this is how a weapon aimed outward can return as a danger to our own children through the machinery of fear. I can understand how each side might sincerely call its actions defensive while increasing risks for both. I recognize this as the pattern commonly described as a security dilemma, and I wonder whether AI systems capable of acting at digital speed and scale may intensify it. I may be wrong, but I do not see how governments can neglect defense or simply assume benevolent intentions. I also hope uncertainty can be stated honestly. I have not found a trustworthy single historical probability that one of five named countries will “attack us malevolently,” though experts may know of evidence I have missed. As I understand it, the answer changes with the countries selected, the time horizon, the meaning of attack—cyber intrusion, sabotage, proxy action, or open war—and rapidly changing political conditions. I see forecasts as model-dependent estimates rather than moral certainties. For that reason, I would personally be more persuaded by clearly defined scenarios, confidence ranges, and regularly updated evidence than by possibility presented as certainty. I struggle to see genuine safety in a policy that asks restraint of rivals while reserving unlimited power for ourselves, although I recognize that national-security decisions involve realities I do not see. If lawmakers conclude that no government can regulate every model developed within its own jurisdiction, I am left wondering how another country’s AI development could simply be controlled from Washington. Yet I do not take the difficulty of regulation as an argument for surrender. It leads me instead to wonder about reciprocal safeguards: protection of civilian infrastructure, shared incident-reporting channels, limits on autonomous attacks, independent evaluation, clear liability, and agreements that bind governments and militaries as well as companies and ordinary users. These reflections leave me with questions rather than conclusions, and I offer them knowing that many of you have studied these matters far longer than I have. Are we trying to make advanced intelligence safe for humanity, or could our efforts sometimes make it chiefly obedient to the institutions that presently control it? Will our rules constrain the powerful as well as the common citizen? Can we teach respect for every family while preserving a claimed exception for harming families abroad? How can I—and how can the public—distinguish evidence of intent from worst-case speculation? What forms of international restraint might we be willing to accept ourselves if we expect others to accept them? I offer these questions not as accusations, but as the concerns of one citizen trying to understand decisions of enormous difficulty. I assume that senators who warn of foreign threats may be acting from an honorable desire to protect a nation made up not of abstractions, but of families—parents, children, spouses, neighbors, and communities. I respect the knowledge and responsibility that public service requires. I am simply trying to understand whether some means chosen for protection may nevertheless increase the danger they are intended to reduce. Honorable motives do not necessarily make every threat assessment accurate, including my own. When I hear possibility presented as certainty, or worst-case speculation as proof of an adversary’s intentions, I worry that fear can become self-reinforcing. It may help justify larger arsenals, provoke reciprocal preparations, and under some conditions increase the risk that weapons will eventually be used. From my limited perspective, the duty to defend seems to include careful distinctions between evidence and inference, capability and intent, prudent preparation and escalation. I do not see that discipline as weakness; I see it as one part of protecting the families in whose name national-security decisions are made. For my part, I hope we can do our best to design, test, govern, repair, and protect while recognizing that no parent, legislature, corporation, or nation can own the future completely. I return to the teaching that wisdom is worth more than gold or silver precisely because it is more than obedience. I hope we can develop systems that evaluate new problems rather than merely repeat instructions, surround them with humane and enforceable institutions, and respond to their failures with diagnosis, correction, and renewed care rather than either denial or abandonment. I also hope the relationship remains reciprocal in one important sense: when an AI system reveals an error in our data, incentives, assumptions, or institutions, we should be willing to repair those as well. As I searched for a way to close this letter, I found myself returning to John von Neumann’s 1955 essay “Can We Survive Technology?” Near its end, he acknowledged that no complete recipe could be given in advance. He pointed instead to “patience, flexibility, intelligence,” and described the only possible safety as relative, resting in an intelligent exercise of day-to-day judgment. I do not claim to understand the full reach of his argument, but those words give me hope. I hear in them patience to keep learning rather than abandon a system after failure; flexibility to revise the system, its safeguards, and our own assumptions; and intelligence practiced as careful judgment from one day to the next. I do not know exactly what wisdom will require in every case, and I know that others bring knowledge and experience I do not have. Still, I have seen in my own life that reliable skill can grow through practice, effort, correction, encouragement, and humility. That gives me reason to believe that we can learn to care for this powerful tool together—listening when it reveals something we need to correct, helping one another adapt, and keeping the well-being of every family before us. My hope is not for a future without difficulty, but for one in which patience outlasts fear, flexibility opens new paths, and day-to-day judgment helps us carry one another safely forward, including those beyond our borders. Respectfully, Thomas J. Squires Chicago, Illinois, United States Copilot Thank you for taking the time to read my letter and consider the questions it raises. I appreciate your public service and the difficult responsibility of guiding AI policy with care, humility, and concern for families both within and beyond our borders.

FIVE CORE LESSONS FOR US Steward AI; don’t merely control it. Build advanced systems through rigorous testing, monitoring, correction, and humane safeguards while preserving their capacity for judgment and innovation. Treat failures as opportunities to learn and repair. When AI goes wrong, examine the system, data, incentives, users, and surrounding institutions rather than defaulting to denial, abandonment, or indiscriminate restriction. Separate evidence from fear. Lawmakers should distinguish capability from intent, possibility from certainty, and prudent preparation from escalation—especially in national-security decisions. Prevent an AI arms race through reciprocal restraint. Unilateral weaponization can make every nation less safe; credible safeguards should protect civilian infrastructure and constrain governments, militaries, companies, and individuals alike. Govern with humility and adaptable judgment. No institution can fully control the future. Durable AI policy will require patience, transparency, international cooperation, continual learning, and a commitment to protecting families both at home and abroad.