> ## Content Index > Fetch the complete content index at: https://www.intospatia.xyz/llms.txt > Use this file to discover other available public pages before exploring further. # 7 · Second Launch - URL: https://www.intospatia.xyz/7-second-launch/ - Published: 2026-09-22T12:23:18.000Z - Updated: 2026-09-22T12:23:18.000Z - Author: Michael Robbins > The crewed mission for the AI era — from extraction to ownership, from anthropomorphic to anthropogenic, and the launchpad we can build now > The first launch was the machine’s. The second one is ours. This book began with a launch. A boy in a high-school classroom in Huntsville, on the morning of his sixteenth birthday, watching a teacher come through the door to say the space shuttle had broken apart over the Atlantic. The school was named for Gus Grissom, an astronaut who had himself died on a launch pad, years before the boy was born. Everything since has circled back to that morning: the cost of putting human beings atop a rocket, and the question of whether the people closest to the danger are ever really aboard. Flight offers several histories of beginnings: uncrewed rocket tests, gliders carrying their inventors, and machines tried under carefully limited conditions. The shared lesson is that a first demonstration is not a mature system. Reliability, institutions, and the capacity to learn from failure take further work. This chapter is about the second launch — the crewed one, with people aboard and in command, the one this whole book has been built to make possible. To see why it is different, and why it is finally within reach, it helps to begin where the machines did: in the desert, with a race that everyone lost. ### The Mojave Desert, 2004 In March 2004, fifteen finalists qualified for DARPA’s first Grand Challenge. The task was to drive an autonomous vehicle across a desert route from the Barstow area toward Primm, Nevada. A million-dollar prize made the test public and concrete. No vehicle completed the route. The prize went unclaimed. The event exposed failures that a working demonstration in easier conditions could conceal: sensing, navigation, mechanical reliability, and response to unexpected terrain. The failure was useful because the task made limitations visible. It gave teams a shared problem and another opportunity to improve. That does not mean the event alone created autonomous driving, but it helped organize a community around a demanding test. In October 2005, five vehicles completed the next Grand Challenge, with Stanford’s Stanley winning. The Urban Challenge followed in November 2007, adding traffic and other demands of a simulated urban environment. A recurring public test helped turn separate efforts into cumulative work. The modern autonomous-vehicle industry descends, in some recognizable way, from teams that competed in those three races between 2004 and 2007\. The first race made visible what the field could not yet do; the second produced finishers; the third produced finishers operating in a shared environment under realistic rules. None of it would have happened without the first race’s collection of failures. The vehicles that lay stationary in the Mojave were not embarrassments. They were the foundation. Those machines operated on an authorized test course under organized safety arrangements. Their autonomy was technical, not freedom from accountability. For AI, the corresponding question is what bounded tests can reveal about a system’s ability to act under human authority before people rely on it in consequential settings. This chapter is about that second launch, and the second launch is different from the first in every way that matters. The first was the machine’s: autonomy demonstrated, the web scraped for fuel, agents sent out alone, built at scale by a few firms for purposes the people who supplied the fuel never set. The second launch is ours. It carries people rather than only capabilities — the AI representative as a crewed craft with a person in command, not an agent-drone running loose. It runs on consent, credit, and compensation instead of extraction. It moves from a world in which people are the scraped material to one in which they are the crew and the owners. It is *anthropogenic*, not merely *anthropomorphic*: of human origin and under human command, not just built to seem human. It is, in the oldest American phrasing for it, AI by the people, of the people, and for the people. And there is now a way to begin it, which is the part that has always been missing. ### First launches fail in public, and become the foundation In January 1920, the New York Times ran an editorial mocking Robert Goddard, a physics professor at Clark University who had just published a Smithsonian paper, A Method of Reaching Extreme Altitudes, proposing that rockets could leave the atmosphere and reach the moon. The Times explained that this was absurd: rockets needed atmosphere to push against, and in a vacuum there would be nothing to push against. The idea, the editorial said, revealed an ignorance of high-school physics. The Times was wrong about the physics: a rocket expels mass and does not need an atmosphere to push against. Goddard continued developing and testing rockets. The history is a reminder to examine a claim carefully, not a reason to assume that a criticized proposal must eventually succeed. He died in August 1945, having never seen a rocket reach the altitudes he had predicted. On July 17, 1969, three days before Apollo 11 reached the moon, the Times printed a correction acknowledging that a rocket can in fact function in a vacuum, and regretting the error. The correction came forty-nine years after the editorial, and twenty-four years after Goddard’s death. The work eventually became part of the engineering inheritance that reached Huntsville. I encountered that inheritance in the city’s schools, families, and institutions. The distance between a small experiment and a crewed mission was filled by the work of many people, not by the experiment’s promise alone. My fifth grade teacher made room for my curiosity about that history by granting me days on end to do independent study on Robert Goddard and his rockets instead of forcing me back into the pace of the class. I have carried the gift of that permission with me. A teacher can make an inheritance available by letting a child follow a question far enough to discover what it asks of them. My questions that led me to architect a different future for AI didn’t begin there, but now the work and it’s challenges pull me ever forward. The Wright Brothers worked the same way a generation earlier. Orville and Wilbur ran a bicycle shop in Dayton, Ohio, and from 1899 to 1905 they used the techniques of bicycle manufacture — lightweight structures, control surfaces, balance, tolerances, the discipline of iterative testing — to build the first powered aircraft. Their first flight at Kitty Hawk, on December 17, 1903, lasted twelve seconds and covered one hundred and twenty feet, witnessed by five people and one camera. The Wrights’ early work also became an inheritance others could develop. The useful pattern is a specific problem, a test, a record of what happened, and another attempt. It is a pattern available to projects whose ultimate outcome is still uncertain. ### The lesson of the first race What the first race revealed, beyond the specific engineering deficits, was how civilization-scale technical systems actually get built. The pattern is not that a single brilliant team produces a breakthrough that diffuses to everyone else. The pattern is that a public challenge forces a scattered field to converge on a shared problem; the first attempts fail under conditions everyone can see; the failures become the curriculum for the second round; the second round produces partial successes; the third produces operational systems; and an industry comes into being. The first race failed for everyone, which meant the failure was equally distributed and the lessons were equally available. The second race was made possible by the first. From March 2004 to November 2007, the challenges offered a sequence of increasingly demanding tests. They drew on an existing research field and helped give it shared deadlines, comparisons, and visibility. Coordination can make learning cumulative without replacing the work that precedes it. That is the element I want to borrow: a coordinating practice that lets people see what others tried, learn from it, and improve. The history provides encouragement for the method. It does not establish how quickly an AI institution can mature or guarantee that the analogy will hold. The AI era is waiting for its coordinating function. The capability test has been passed, repeatedly; the civilization test has not yet been run. The equivalent of the Grand Challenge would be a public exercise in which autonomous and semi-autonomous AI systems were deployed under specific conditions of representation, consent, audit, and recourse, with the failures of those deployments made visible to the engineering and policy communities so the second round could improve on the first. AI already has many forms of evaluation. This project needs an additional, bounded question: can a system help a person represent an experience accurately while obeying that person’s permissions? A public program could compare approaches, but early trials should begin small enough that errors can be understood and contained. ### The mission systems A crewed launch needs mission systems: the civic, legal, economic, and institutional architecture that lets AI systems operate inside human civilization without dissolving the civilization they operate inside. The systems have parts. Some have been named in this book; some are still being named. The mission needs an identity layer that distinguishes a person from a platform account, a bot, a model, a device, a corporate agent, and a state-issued record. It needs authorization protocols that let a person delegate authority to an AI representative within bounded scopes, revoke that authority, audit its use, and obtain recourse when the representative acts outside its bounds. It needs provenance systems that make it possible to know whether a piece of content was made by a human, an AI, or a collaboration of both, and under what authority. It needs governance frameworks for the platforms, the models, the training data, the representatives, and the spatial computing infrastructure the era is building. The proposal also needs institutions able to hold assets, contract for services, govern shared resources, and answer for failures. Parts can be drawn from existing practice. The complete combination described here remains to be specified and tested. The mission also needs something simpler than any of these components and harder than all of them: the willingness of the people who could build it to actually build it. The autonomous-vehicle industry was built by people who showed up to the Mojave in 2004 and accepted the public humiliation of being filmed as their machines failed. The first generation of human-centered AI institutions will accept the analogous risk. They will publish proposals that turn out to be wrong. Some pilots will fail. Their value will depend on whether the failures can be examined honestly and whether participants are protected from avoidable harm. Criticism can reveal that an idea needs revision or abandonment. Persistence is useful when it remains capable of learning. It gets built by communities that decide the work is worth attempting now, under conditions in which the attempt will be public and the failure will be visible. Goddard accepted those terms in 1920; the Wrights accepted them on the sand at Kitty Hawk in 1903; the Carnegie Mellon team accepted them when they unloaded Sandstorm in the Mojave in 2004\. The first builders of human-centered AI institutions will accept the same terms over the next decade — and unlike Goddard, they will not have to accept them alone. ### Standing: being aboard your own representative The most important mission system is not a technology. It is a status. The status is standing — the recognized position from which a person can act, be represented, contest decisions, own claims, authorize agents, revoke authority, and remain legible across domains without being exposed to every system that wants to know them. Standing is the difference between being the crew of your own craft and being its cargo. Identity tells a system who you are. Personhood gives you standing. The contemporary internet does not give people standing. It gives them accounts. An account is a record inside a particular service that lets the service authenticate the user, retrieve their data, apply their permissions, bill their subscription, and target their advertising. The account belongs to the service, not the person; the person belongs, in a real sense, to the account. The information about the person, accumulated over years, is held by the service under terms the person can rarely read and never meaningfully negotiate, and the person can be removed from the account at any time, by mistake or design or automated decision, with no practical appeal. The person has not been given a place. They have been allowed to use one. Digital wallets are a real and important improvement and not a substitute for standing: a wallet lets a person hold credentials and proofs without a particular service vouching for them, but the wallet itself has no governance, no fiduciary duties, no institutional weight. A wallet, as the Credit Union chapter argued, is a container. Standing is not a container. Standing is a status a person holds within a civic structure that recognizes them as an actor. *Digital personhood* is the term this book uses for the standing the AI era requires: the persistent civic position of a human being across the blended physical-digital world. It is more than identity verification, more than a credential set, more than an account or a wallet. It is the recognized standing from which a person can act and be represented in a world where most of the action happens through digital systems and through the AI representatives of those systems. It is also what makes the person the commander of their own representative rather than its passenger — the human origin and the human crew of the craft, which is what *anthropogenic* finally means. Operationally, the person needs control over identifiers appropriate to each context, bounded delegation, inspectable permissions, and meaningful recourse. A single identifier used everywhere can create tracking risks; pseudonymous or context-specific credentials may serve better. Existing technologies provide parts of the solution. The challenge is integrating them under obligations people can understand and enforce. These components are not science fiction. Each has existing analogues in adjacent technical fields, and partial implementations exist for several. What does not yet exist is the integrated civic architecture in which the components work together as the standing of a digital person. Building that architecture is the mission’s civic foundation. Without it, AI representatives have nothing to represent and AI Trusts have no members to be accountable to. With it, the rest of the institutional architecture of the AI era becomes operationally possible. ### From drones to crewed spacecraft The launch frame gives a precise vocabulary for what an AI representative is. It is not a smarter chatbot, not a generic assistant, not a corporate agent dressed in personal clothing. It is the crewed craft through which a person acts — the person remaining the commander, the systems being the engine, and Spatia being the space they travel together. This is the whole of the contrast the chapter has been building: from AI agent drones to AI representative crewed spacecraft. The crewed-craft image names an obligation for the engineering as well as a relationship of authority. Representational AI is carrying humans: the system's account of us can affect the opportunities we enter, the commitments made in our name, and the ways others treat us. The person should remain able to direct, question, and stop the system. Its computational foundation must also be built for the meaning entrusted to it. Generative models can help us communicate and explore, while the structures that hold our experience and govern consequential action demand their own design and testing. That is why the semantic work of the earlier chapters belongs among the mission systems. The autonomous-vehicle community discovered between 2004 and 2007 that autonomy is not a property of the vehicle. Autonomy is a property of the vehicle plus the road plus the rules plus the other vehicles plus the people. A vehicle that performs flawlessly on a closed track can fail catastrophically on a city street, not because the vehicle has gotten worse but because the conditions have gotten harder. The same is true of AI: a system that performs astonishingly in a benchmark evaluation can fail catastrophically in a civic deployment, not because the AI has gotten worse but because the conditions have gotten harder. The mission has to be built for the autonomy to function. This is what co-evolution means in the running argument of this book. Humans and AI representatives become more capable together, each enabled by the other, provided the architecture supports both directions. The representative becomes more useful as the person grows more skilled at directing it, limiting it, and learning from it; the person becomes more capable as the representative carries load the person no longer has to carry alone. Support only the AI direction, and the result is the diminishment of human agency this book has argued against from its first page; support only the human direction, and the result is missed capability. Co-evolution requires both, and the institutional architecture of the trilogy — the Land Trust holding the foundational assets, the Mutual AI Credit Union pooling the means, the Cyberwealths Enterprise organizing the productive capacity — is the support structure that makes it possible at scale. The crew and the craft grow together, or neither does. ### Spatia is the shared sky The 2007 Urban Challenge moved testing from open desert into shared space, and the shift was a change of category, not merely of difficulty. A vehicle alone in open desert can ignore most of the architecture of traffic. A vehicle in a city cannot: it has to handle other vehicles, signals, intersections, lane markings, right-of-way, pedestrians, cyclists, road workers, emergency vehicles — an environment full of other actors with their own standing. The first launch only has to leave the ground. The crewed mission has to operate in shared space, where other craft are flying, where the sky has traffic, where standing is mutual. The AI era’s shared sky is Spatia. Spatia, as the second trilogy argued at length, is the civic terrain of the blended physical-digital world — the layer where AI representatives, institutional agents, public systems, commercial systems, sensors, robots, and human beings all have to operate together. The open-desert questions — can a model produce language, can a system generate images, can an agent execute a workflow — have largely been answered. The shared-sky questions — can these systems operate together in shared civic space, under conditions where many actors have standing, where infrastructure is contested, where accountability is required — have barely been formulated. The school hallway, the public park, the museum, the workplace, the neighborhood, the family home, the sacred site, the public square are all about to become places where AI-mediated experience overlaps with embodied experience, and where who has standing, who has authority, who is recognized, who is heard, and who governs becomes a daily live question. Spatia is the name for building that civic infrastructure so the people who live in those places have standing within it. It will not be won by a single craft, model, platform, institution, or city. It will be won by the assembly of an architecture that lets many actors operate together with mutual recognition and lawful authority — the work of this generation and the next. ### Gameshow is the way: the launchpad, the academy, the spaceship A pathway has to begin with work people can actually do. For Gameshow, that means testing the smallest useful cycle before treating the whole proposal as an operating institution. One possible beginning is a project within an existing local network: people making something together, responding to a neighborhood need, or preparing for an opportunity. The shared activity gives them experiences to examine and a reason to return. A facilitator can support reflection, and simple tools can hold approved records. The particular project should come from the people involved. The first task is authorship. A participant creates a Dote, reviews the proposed record, corrects it, and decides what to retain. The project should record where the process becomes confusing, burdensome, or inaccessible. The next task is bounded representation. A Rep helps prepare one agreed output from approved records. Participants inspect the evidence behind its claims, reject interpretations, and test whether private material stays outside the output. This is an early cyber range: try the task, challenge the result, correct a rule, and run it again. The scope stays narrow enough that people can understand the failure and decide whether the repair worked. Then test the institutional promises: export a record, change a permission, challenge an account, withdraw from the service, or file a complaint. These actions should work before the project promises durable standing at scale. A community can share what it learns without publishing private stories. The useful public record would include methods, costs, failure categories, and changes made in response. Competition may help some settings; collaboration may serve others. Participation should not depend on wanting to perform in public. That would be a credible first launch: useful work, relationships that can sustain it, and tools open to correction. Its value should show up beyond the session—in a project carried forward, an opportunity reached, or a decision people have more power to shape. Larger institutions could develop as communities learn what needs to be held in common. Gameshow coordinates that learning while the form remains theirs to make. As communities finance, own, and govern these institutions, they can connect them through shared services and agreements. Growth becomes the spread of capacity people can hold together. ### What the first attempt needs to hold together The first attempt connects the book’s three questions: who is represented, where participation happens, and who holds the value. It should make progress on a limited version of each without claiming the entire system has been solved. And under all of it, the question from the AI Club, asked in a fifth-grade voice and a seventh-grade voice. Henry, who wanted to go live in the woods. Camila, who wanted an AI of their own. Who am I? Who are you? Who are we together? That is the crew, and it was always the crew. ### From check to stake The Cyberwealths trilogy argued that the productive capacity of the AI era should be owned by the people whose lives generate its value: the Land Trust holds the foundational assets, the Mutual AI Credit Union pools the member accounts, the Cyberwealths Enterprise organizes the productive contribution into ownership stakes. The institutional precedents cross political traditions and offer mechanisms worth adapting. They do not yet amount to a complete architecture for AI ownership. The proposal’s credibility will come from specifying the rights, costs, and responsibilities of a real first instance. From check to stake means adding a possible continuing claim and a voice to the benefits people receive. The rights, risks, and limits must be explicit. A community should not have to choose blindly between a small payment it understands and an ownership promise whose meaning has never been specified. ### The Launchpad If you are a teacher, builder, organizer, parent, or potential participant, begin with the part you can help test. A community does not need the whole architecture to find out whether an authored record can make learning more visible and useful. It does need a purpose, willing participants, a way to protect their choices, and enough support to complete the attempt responsibly. Publish what is learned. Change what failed. Let the next group begin with more than a promise. Henry and Camila will spend their adult lives inside the systems we build. That is why the work matters, and why it has to include people like them in deciding what comes next. The second launch is ours to make possible. The first launch was the machine’s. The second one is ours. Let’s build it together. ## Sources DARPA, “Grand Challenge,” official innovation timeline: [https://www.darpa.mil/about/innovation-timeline/grand-challenge](https://www.darpa.mil/about/innovation-timeline/grand-challenge?ref=intospatia.xyz). See also DARPA’s official Urban Challenge history for the 2007 event. Sebastian Thrun et al. Stanley: The Robot that Won the DARPA Grand Challenge. Journal of Field Robotics, 2006\. Primary technical source for the 2005 winning vehicle and the field-building significance of the Grand Challenge series. Robert H. Goddard. A Method of Reaching Extreme Altitudes. Smithsonian Miscellaneous Collections, 1919\. The paper at the origin of modern rocketry, mocked in the New York Times editorial of January 1920 and vindicated by the moon landing in July 1969. The New York Times, “Topics of the Times,” January 13, 1920, and correction published July 17, 1969\. The forty-nine-year arc from public mockery to formal correction; useful for the long pattern of foundational work being underestimated in real time. Wright Brothers papers, Library of Congress, and the National Air and Space Museum, Smithsonian Institution. Primary archival source on the Wright Brothers’ bicycle-shop origins, their 1899–1905 development work, and the December 17, 1903, first powered flight at Kitty Hawk. Jaron Lanier and E. Glen Weyl. A Blueprint for a Better Digital Society. Harvard Business Review, 2018\. Background reference for the data dignity argument; supports the claim that intermediary institutions are necessary for the AI era to be operable on terms accountable to the people whose lives produce its value.