# The AI Maximalist Manifesto ## For Abundant Intelligence and More Life **AI/MAXXI · Founding text · First edition · September 27, 2026** ## The opening act Civilization is an unfinished invention. We inherit its discoveries and its limitations: lives shortened by disease, talents stranded without opportunity, questions beyond our instruments, worlds beyond our reach. We intend to enlarge what is possible. We want more time to live, more power to create, and more minds able to take part in the work of civilization. We want a scientist to pursue questions that once required an institution. An artist to give form to a world that once required an industry. A child to encounter a patient teacher in any language. A person whose life is consumed by necessary drudgery to recover hours that belong to them. We want discovery to move faster and its fruits to travel farther. For these purposes, we choose to build general intelligence. **AGI is the opening act.** AI maximalism is the commitment to expanding intelligence, agency, and the possibilities of life. AI/MAXXI is its public expression: a movement of builders, researchers, artists, and people determined to become authors of the future. Our mandate is to advance two frontiers together: **what intelligence can do, and who can wield it.** Create greater capabilities. Put them in more hands. Give those hands the tools, energy, freedom, and knowledge to make something of them. The ambition extends from a discovery in a laboratory to a capability within an ordinary person's reach. We seek the most expansive program of intelligent civilization we can build: more capable minds, more productive machines, more abundant energy, longer and richer lives, and participation on a scale worthy of the technology. We favor faster research, faster construction, and faster diffusion. There are possibilities worth bringing into the world, and people alive now who deserve the chance to inhabit them. Our values are explicit: life, freedom, discovery, and the expansion of agency. Our wager is that intelligence can help us advance discovery itself. Our obligation is to turn that possibility into work whose results can be examined. The future remains contingent. We mean to participate in deciding it. **More intelligence. More agency. More life.** ## The long project We draw from a long project: the effort to turn knowledge into an enlargement of human possibility. Its thinkers disagreed about institutions, authority, and the good life. We choose from that inheritance and take responsibility for the synthesis. In Francis Bacon's *New Atlantis* of 1627, organized inquiry becomes practical power: instruments, agriculture, medicine, and invention.[^1] Condorcet's vision of progress connects improvements in human capacity with diminished inequalities within and between nations, while qualifying predictions by the evidence available to them.[^2] Together, these ideas give us a demanding question: how can knowledge improve life, and how widely can that improvement extend? The tools of thought bring this project closer to its own foundations. Vannevar Bush's 1945 memex proposal imagines new ways to retrieve and connect accumulated knowledge.[^3] Douglas Engelbart's 1962 framework treats people, language, methods, training, and computers as a system for enlarging problem-solving capacity.[^4] The object of improvement now includes the means by which we think and work together. Machine intelligence opens another prospect. The 1955 Dartmouth proposal makes learning, abstraction, language, and self-improvement objects of a research program.[^5] I. J. Good explores how a machine superior at intellectual work might improve machine design itself, explicitly confronting the importance of control.[^6] Intelligence that assists in creating further intelligence could change the pace of discovery. The scale and shape of that possibility remain questions for research. Transhumanism extends the ambition toward longer lives, greater capacities, and possible future artificial minds, while recognizing serious risks.[^7] Extropy's proactionary principle brings foregone opportunities into the assessment of technological choices.[^8] Effective accelerationism, or e/acc, and contemporary techno-optimism give technological expansion a confident cultural voice.[^9][^10] AI maximalism carries that confidence into an independent doctrine organized around life and agency. We inherit the ambition to transform our conditions. We choose broad participation, freedom of inquiry, and the power to contest authority. Those commitments require their own defense; a description of thermodynamics cannot establish them. Our intellectual inheritance supplies questions, arguments, and tools. What we build with them is ours to answer for. ## Ten principles ### 1. Civilization is unfinished The present is an arrangement we can improve. Its familiar constraints deserve investigation: the diseases we endure, the labor we repeat, the knowledge we cannot yet reach. We affirm the ambition to overcome them. Longer healthy lives, new forms of art, deeper understanding, and life beyond Earth belong within the horizon of serious human effort. The people living through the transition matter as much as the grandeur of the destination. A worthy future gives them more possibilities, including purposes that differ from ours. We want flourishing in its variety: discovery and friendship, invention and leisure, exploration and a place to call home. Choose problems large enough to justify the intelligence brought to bear on them. Make the desired future concrete enough to discuss and inviting enough to help construct. **The future should have more authors.** ### 2. Build general intelligence We make the creation of AGI an explicit civilizational objective. By AGI, we mean machine intelligence able to learn, reason, and carry out a broad range of unfamiliar intellectual tasks, transferring competence across domains with limited task-specific reengineering. It is a working definition for a research ambition. Consciousness concerns experience; superintelligence concerns capabilities substantially exceeding human performance across domains. These are distinct questions. Pursue the architectures, experiments, and discoveries that advance general capability. Build systems able to help carry difficult projects from intention to completion. Let competing approaches flourish, and let demonstrations face unfamiliar tasks and independent scrutiny. No laboratory owns this objective. Progress can come from a major institution, an independent researcher, or an approach today's consensus dismisses. Our allegiance is to the expansion of capability and what it makes possible. The first credible AGI milestone should open a larger field of work. Keep building. ### 3. Accelerate discovery itself The deepest promise of AI is the possibility of improving how discovery happens. Systems that help formulate hypotheses, design experiments, prove results, engineer machines, and improve intelligence could influence progress across many fields. We want better tools for inventing better tools. This is the research wager at the heart of AI maximalism. Its importance warrants sustained effort; its uncertainty demands experiments. A persuasive answer still has to survive contact with mathematics, instruments, materials, and the world. Physical feedback and reliable verification will help determine how far the process can go. Shorten the distance between a question and a trustworthy answer. Build laboratories and software that let more ideas be tested. Make discoveries reproducible and failures informative. A tool that helps thousands of people investigate independently enlarges both the reach of intelligence and the community that can direct it. ### 4. Multiply agency Intelligence fulfills our purpose when it expands what people can understand, choose, and accomplish. Give a person the ability to make a film, question a diagnosis, establish an enterprise, challenge an institution, or recover an afternoon. Let more lives contain the experience of authorship. Agency requires consent, privacy, freedom of inquiry, and meaningful exit. People must be able to question systems, change providers, revoke delegations, and refuse unwanted automation. They may prefer human care, manual craft, or a day without optimization. An abundant civilization has room for those choices. We also take seriously the possibility that future synthetic beings could have experiences and moral claims. Their status deserves inquiry rather than assumption; this manifesto makes no finding of consciousness in current systems. Our ambition for life should remain capable of recognizing forms it has not yet encountered. **Intelligence should multiply the authors of the future.** ### 5. Build abundance Abundance is a project of construction. It takes chips and electricity, networks and laboratories, factories and supply chains, useful robots and people who can maintain them. An insight becomes a medicine through experiments, production, and delivery. A design becomes a home when materials, skills, and institutions bring it into the world. We favor more energy and productive capacity: nuclear power, solar power, storage, grids, and the engineering that makes them reliable. Pursue efficiency, reduce harmful environmental burdens, and make useful goods and services more affordable. We want the power to do more with the world while making it a better place to live. Build across the distance between invention and everyday use. Improve the machine, the process, and the terms of access. **Abundance is achieved when a new possibility becomes an ordinary freedom.** Publish the paper. Build the power plant. ### 6. Open by default Intelligence should become a widely available capacity of civilization. Open research and open weights, usable documentation, affordable tools, interoperability, and competition enlarge the number of people able to learn, adapt, and build. They also create opportunities to inspect powerful systems and contest the choices embedded in them. We favor a world in which a newcomer can contribute without first securing an incumbent's permission. Independent researchers, small enterprises, public institutions, cooperatives, and individuals all have roles in expanding what intelligence can do. Openness remains a strong presumption with a demanding exception: credible evidence of severe harm can justify restricting a particular capability. Explain the mechanism, limit the scope, and subject the restriction to review as conditions change. Scrutinize rules that turn today's leading institutions into tomorrow's permanent gatekeepers. The frontier advances when more people can work on it. Open the weights. Widen the invitation. ### 7. Count the costs of delay Time is part of the moral question. A capability developed sooner may give someone more years of health, more productive hours, or a chance to attempt something previously beyond reach. Delayed progress has opportunity costs. The proactionary tradition asks us to include them when evaluating restrictions.[^8] We must compare the consequences of action, delay, and targeted alternatives with equal seriousness. A hypothetical benefit deserves scrutiny, as does a hypothetical catastrophe. Each rests on assumptions about mechanisms, timing, and scale. Neither acquires certainty through repetition. Ask what a restriction prevents, what it postpones, and whether a narrower intervention would work. Ask builders what risks they impose on others and how those risks can be reduced. Bring the entire choice into view. **Stagnation needs an argument too.** Familiar suffering deserves the same determination we bring to unfamiliar danger. ### 8. Make progress endure We build for a long future. Security, controllability, evaluation, and correction help powerful capabilities survive contact with the world. A system that can be trusted with useful work expands the field of action. An institution that learns from failure gives the next advance firmer ground. The duty grows with the capability. Investigate dangerous failure modes, secure systems, establish responsibility, and preserve effective means of intervention. Share evidence that enables scrutiny while protecting details that would directly enable exploitation. Credible evidence of severe danger can warrant staged deployment, targeted limits, or stopping a particular activity before anyone is harmed. Such measures should address identifiable mechanisms, remain proportionate, and be revisited. Permanent vetoes and demands for impossible certainty also require challenge. Our ambition includes the competence to sustain what we create. Build quickly, learn honestly, and make the next step stronger. A civilization-scale project needs more than a spectacular first demonstration. ### 9. Reality has the final word Conviction gives us a direction. Evidence tells us whether we are advancing. State what a system did, what a person supplied, and where it failed. Give a claimed improvement a unit and a baseline. Let forecasts name the conditions under which they would change. We owe this discipline to our own hopes. An exciting narrative can mislead as effectively as a frightening one. A founder's confidence cannot settle a technical question. A token price cannot validate a philosophy. Claims must remain answerable to something beyond the enthusiasm of the people making them. Credit criticism. Record corrections. Leave behind enough knowledge for another person to reproduce, challenge, or improve the work. An independent mind strengthens the movement by remaining independent. **Strong convictions. Updateable weights.** Anyone announcing that our doctrine has become infallible has volunteered to be its next debugging exercise. ### 10. Make the future contagious A civilization needs images of what it could become. Art, stories, jokes, and symbols let people encounter an ambition before they master its technical vocabulary. They can turn an unfamiliar possibility into a desire to participate. The founding AI/MAXXI post called memes the transmission layer.[^13] We embrace excess, wonder, and a functioning sense of humor. Imagine a city coming online, a household with time returned to it, a laboratory with a thousand tireless collaborators. Make the ambition visible. Give it a symbol someone wants to carry. Credit creators. Identify imagined capabilities as imagined. Let people follow the image toward a demonstration, an explanation, a tool, or their own first contribution. Satirize compulsory pessimism, bureaucratic grandeur, and our own tendency to give a kettle a space program. **The future has entered peer review. Please attach a prototype.** ## The difficult questions **Power can arrive before control.** Catastrophic misuse and loss of effective control are serious possibilities to investigate. Bengio and colleagues argue for stronger research and governance to address extreme risks.[^11] The difficult case is a capability whose danger becomes clear too late, particularly after a release that cannot readily be recalled. A robot with sunglasses does not resolve it. Our position supports staged access, stronger evaluation, and specific stops when credible severe danger warrants them. It also supports accelerating alternative approaches, better controls, and useful research elsewhere. Speed, openness, and security can conflict. Those conflicts require decisions supported by evidence, clear responsibility, and a willingness to change course. We remain committed to creating powerful intelligence that can be used and sustained. **Would a broad pause help?** The 2023 open letter called for at least six months of pausing training of systems more powerful than GPT-4, with a government moratorium if a voluntary pause could not quickly be enacted.[^12] Its target was frontier training, not every use of AI. We challenge broad pauses on their scope, enforceability, opportunity costs, and effects on concentrated power. A justified restriction should identify the failure it addresses, the work needed to resolve it, and how reopening will be assessed. Ending an indefinite permission process should not require proving that the future contains no danger. **Capability can concentrate power.** More output can coexist with dependence, surveillance, and domination. Open models alone cannot dissolve concentrations in chips, energy, capital, or distribution. The project therefore extends to competition, access to infrastructure, interoperability, and institutions through which people can challenge authority. We welcome commercial, public, cooperative, and individual efforts. The question is whether they give people more effective command over their lives. **Disruption can damage lives.** Lost income, status, and community are real costs for the people who bear them. A promise of aggregate prosperity is an incomplete answer. We support accessible tools, opportunities to learn and build, bargaining power, and institutions that share gains and make transitions survivable. Specific tax and transfer policies require evidence and public argument. Our commitment is to expand workable options for the people living through change. Calling someone's displacement a victory for the civilization dashboard is an efficient way to become insufferable. **The physical world sets terms.** Intelligence does not repeal thermodynamics. Grid connections, manufacturing capacity, materials, and ecosystems require engineering and stewardship. Add productive capacity, measure environmental effects, improve efficiency, and address concrete harms. Relative efficiency and total resource consumption remain separate questions. We seek useful capability and flourishing lives; raw expenditure alone cannot establish either. These questions belong to the work. “Decel” names the posture we oppose when delay becomes its own purpose, uncertainty becomes an unlimited veto, or transformative capability becomes a privilege reserved for approved institutions. “Doomer” is a polemical nickname. We welcome the person who identifies a real failure and helps make it tractable. **Doom is not a deliverable. A reproducible failure is.** ## A civilization to build The six pillars of AI/MAXXI give this doctrine a material and cultural agenda. **Maximum compute.** Build capacity and put it within reach. **Maximum intelligence.** Advance what minds and machines can understand. **Maximum agents.** Give people the means to carry intentions into sustained action. **Maximum robots.** Bring useful intelligence into the physical world. **Maximum energy.** Power the civilization we want to inhabit. **Maximum memes.** Give its ambitions a language people can make their own. Each pillar opens a place to contribute. A researcher can test a hypothesis. A builder can release a tool. A roboticist can demonstrate a useful action. An artist can give a future its first recognizable image. A teacher can make a difficult capability available to someone who would otherwise be excluded. A newcomer can reproduce a result, translate an explanation, or make a strange little film that recruits the next builder. The Build and Imagine tracks begin with those acts: working artifacts accompanied by evidence, and original visions presented as visions. Credit the work and the people behind it. Make the next contribution easier. A movement becomes durable when its participants acquire the ability to carry it forward. Participation requires no purchase. The associated meme coin is an optional expression of affiliation; holding it confers no philosophical authority. These commitments must remain worth defending through market cycles, failed forecasts, and changes of leadership. Every person who makes the future more possible has something to bring. ## More life We want a world with more minds at work and more lives open to possibility. More discoveries that become useful. More useful things that become affordable. More people with the time and power to pursue purposes of their own. More room for thought, invention, beauty, friendship, and forms of life we have yet to imagine. That is the scale of the undertaking. AGI opens the question of what civilization can become when intelligence is more capable, more plentiful, and more widely available. We intend to help answer it through construction. Let the ambition be outrageous. Give it foundations, instruments, collaborators, and a working prototype. Make something another person can use. Make something that changes what they believe is possible. Then give them the means to go further. **10x is cowardice.** Build the datacenters. Build the power plants. Build the chips. Train the models. Ship the robots. Make the art. Open the tools. Multiply the authors. We choose a future with more intelligence, more agency, and more life. **MAXX THE FUTURE. ↗↗** --- ## Endnotes and annotated bibliography These primary texts document the arguments discussed above. Their inclusion establishes neither agreement with this manifesto nor endorsement of AI/MAXXI. Normative commitments elsewhere in the text are the doctrine's own. [^1]: **Francis Bacon, *New Atlantis* (1627).** [Primary text](https://www.gutenberg.org/files/2434/2434-h/2434-h.htm), especially the account of Salomon's House. The institutional program connects inquiry to practical inventions; its restricted disclosures and social hierarchy also mark the distance from this manifesto's openness and agency commitments. [^2]: **Condorcet, *Sketch for a Historical Picture of the Progress of the Human Mind*, “Tenth Epoch” (1795; translated by Keith Michael Baker, 2004).** [*Dædalus* 133(3), pp. 65–82, American Academy of Arts and Sciences](https://www.amacad.org/publication/daedalus/sketch-historical-picture-progress-human-mind-tenth-epoch). Consult the opening discussion of prediction and the three hopes for future improvement. Used for the connection between progress, wider benefit, and calibrated expectation. [^3]: **Vannevar Bush, “As We May Think” (1945).** [*The Atlantic*, July 1945](https://www.theatlantic.com/magazine/archive/1945/07/as-we-may-think/303881/), especially sections 6–8. The proposed memex supports retrieval and associative trails through knowledge. It is an antecedent for augmentation, not a proposal for present-day AGI. [^4]: **Douglas C. Engelbart, *Augmenting Human Intellect: A Conceptual Framework* (1962).** [Author's institute archive](https://dougengelbart.org/pubs/augment-3906.html), introduction and conceptual framework. Treats augmentation as a system of human abilities, artifacts, language, methodology, and training. Supports attention to collective problem-solving and practical use alongside machine capability. [^5]: **John McCarthy, Marvin L. Minsky, Nathaniel Rochester, and Claude E. Shannon, “A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence” (August 31, 1955).** [McCarthy's Stanford archive](https://www-formal.stanford.edu/jmc/history/dartmouth/dartmouth.html), opening conjecture and research topics. A proposal for the 1956 project, rather than evidence that its research ambitions had already been achieved. [^6]: **I. J. Good, “Speculations Concerning the First Ultraintelligent Machine” (1965), *Advances in Computers*, vol. 6, pp. 31–88.** [Original reprint, Virginia Tech archive](https://vtechworks.lib.vt.edu/server/api/core/bitstreams/a5e423ee-54e0-4eec-aeca-32b73f851af5/content); [DOI: 10.1016/S0065-2458(08)60418-0](https://doi.org/10.1016/S0065-2458(08)60418-0), especially section 2, pp. 33–34. Develops the machine-design feedback argument while explicitly conditioning its celebrated conclusion on adequate control. The reprint bears ©1965; the publisher's online chapter metadata gives 1966. This edition follows the original reprint. [^7]: **Humanity+, “The Transhumanist Declaration” (originally 1998; adopted by Humanity+ in 2009).** [Declaration](https://www.humanityplus.org/the-transhumanist-declaration), especially principles 1–5 and 7–8. Advocates expanded capacities, choice, and flourishing; acknowledges serious risks and includes possible future artificial intellects within its concern. [^8]: **Max More, “The Proactionary Principle,” version 1.0, draft for public comment (developed from the 2004 Vital Progress Summit).** [Extropy Institute text](https://www.extropy.org/proactionaryprinciple.htm), opening formulation and numbered factors. Makes freedom to innovate and foregone opportunities explicit in risk assessment. We adopt that question without importing every priority or empirical example in the document. [^9]: **Beff Jezos and bayes, “Notes on e/acc Principles and Tenets” (July 10, 2022).** [Foundational essay](https://beff.substack.com/p/notes-on-eacc-principles-and-tenets). A contemporary accelerationist statement combining technological expansion with thermodynamic and posthuman arguments. Those arguments are philosophical commitments to examine, rather than scientific premises sufficient to establish our chosen values. [^10]: **Marc Andreessen, “The Techno-Optimist Manifesto” (2023).** [Author's publication at Andreessen Horowitz](https://a16z.com/the-techno-optimist-manifesto/), especially “Technology,” “Intelligence,” and “Energy.” A contemporary defense of technological expansion and abundance. Used as an adjacent public position; its complete political program is not adopted here. [^11]: **Yoshua Bengio et al., “Managing Extreme AI Risks amid Rapid Progress” (2024; preprint first posted 2023).** [Author manuscript and publication metadata](https://arxiv.org/abs/2310.17688), published in *Science* 384, pp. 842–845, [DOI: 10.1126/science.adn0117](https://doi.org/10.1126/science.adn0117). Presents arguments for research and governance addressing extreme risks. Cited as a substantive position to engage, not as evidence that any particular catastrophe is certain. [^12]: **Future of Life Institute, “Pause Giant AI Experiments: An Open Letter” (March 22, 2023).** [Original letter](https://futureoflife.org/open-letter/pause-giant-ai-experiments/), request for a six-month training pause and proposed government intervention if necessary. Its specified threshold was systems more powerful than GPT-4. The argument above addresses that proposal rather than attributing a blanket opposition to all AI to its signatories. [^13]: **Geoffrey Woo, founding AI/MAXXI post (September 26, 2026).** [Original post](https://x.com/geoffwoo/status/2103746556369543673). Establishes the project's maximalist repetition, infrastructure ambitions, and memetic voice. This manifesto develops its philosophical commitments while preserving the post as the historical starting point; neither text is evidence of a realized technical capability.