{ "archive_path": "archive/1781518242.322736", "base_url": "matthewbutterick.com/extinction-level-capitalism.html", "basename": "extinction-level-capitalism.html", "bookmarked_date": "2026-06-15 10:10", "canonical": { "archive_org_path": "https://web.archive.org/web/matthewbutterick.com/extinction-level-capitalism.html", "dom_path": "output.html", "favicon_path": "favicon.ico", "git_path": "git/", "google_favicon_path": "https://www.google.com/s2/favicons?domain=matthewbutterick.com", "headers_path": "headers.json", "htmltotext_path": "htmltotext.txt", "index_path": "index.html", "media_path": "media/", "mercury_path": "mercury/content.html", "pdf_path": "output.pdf", "readability_path": "readability/content.html", "screenshot_path": "screenshot.png", "singlefile_path": "singlefile.html", "warc_path": "warc/", "wget_path": null }, "domain": "matthewbutterick.com", "downloaded_at": "2026-06-15T10:10:50.102484+00:00", "downloaded_datestr": "2026-06-15 10:10", "extension": "html", "hash": "TWR3DRDZJBVK3D8RSWHK", "history": { "archive_org": [ { "cmd": [ "/usr/bin/curl", "--silent", "--location", "--compressed", "--proxy", "socks5://tor-socks-proxy:9150", "--head", "--max-time", "60", "--user-agent", "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/128.0.0.0 Safari/537.36 ArchiveBox/{VERSION} (+https://github.com/ArchiveBox/ArchiveBox/)", "https://web.archive.org/save/https://matthewbutterick.com/extinction-level-capitalism.html" ], "cmd_version": "8.10.1", "end_ts": "2026-06-15T10:12:41.945013+00:00", "index_texts": null, "output": "https://web.archive.org/web/20260615101221/https://matthewbutterick.com/extinction-level-capitalism.html", "pwd": "/data/archive/1781518242.322736", "schema": "ArchiveResult", "start_ts": "2026-06-15T10:11:45.842835+00:00", "status": "succeeded" } ], "dom": [ { "cmd": [ "/usr/bin/chromium-browser", "--proxy-server=socks5://tor-socks-proxy:9150", "--disable-features=DarkMode", "--run-all-compositor-stages-before-draw", "--hide-scrollbars", "--autoplay-policy=no-user-gesture-required", "--no-first-run", "--use-fake-ui-for-media-stream", "--use-fake-device-for-media-stream", "--simulate-outdated-no-au='Tue, 31 Dec 2099 23:59:59 GMT'", "--headless=new", "--no-sandbox", "--no-zygote", "--disable-dev-shm-usage", "--disable-software-rasterizer", "--disable-sync", "--window-size=1440,2000", "--user-agent=Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/128.0.0.0 Safari/537.36 ArchiveBox/{VERSION} (+https://github.com/ArchiveBox/ArchiveBox/)", "--user-data-dir=/data/personas/Default/chrome_profile", "--profile-directory=Default", "--dump-dom", "https://matthewbutterick.com/extinction-level-capitalism.html" ], "cmd_version": "131.0.6778", "end_ts": "2026-06-15T10:11:18.849312+00:00", "index_texts": null, "output": "output.html", "pwd": "/data/archive/1781518242.322736", "schema": "ArchiveResult", "start_ts": "2026-06-15T10:11:04.273212+00:00", "status": "succeeded" } ], "favicon": [ { "cmd": [ "/usr/bin/curl", "--silent", "--location", "--compressed", "--proxy", "socks5://tor-socks-proxy:9150", "--max-time", "60", "--output", "favicon.ico", "--user-agent", "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/128.0.0.0 Safari/537.36 ArchiveBox/{VERSION} (+https://github.com/ArchiveBox/ArchiveBox/)", "https://www.google.com/s2/favicons?domain=matthewbutterick.com" ], "cmd_version": "8.10.1", "end_ts": "2026-06-15T10:10:54.184422+00:00", "index_texts": null, "output": "favicon.ico", "pwd": "/data/archive/1781518242.322736", "schema": "ArchiveResult", "start_ts": "2026-06-15T10:10:50.355775+00:00", "status": "succeeded" } ], "git": [], "headers": [ { "cmd": [ "/usr/bin/curl", "--silent", "--location", "--compressed", "--proxy", "socks5://tor-socks-proxy:9150", "--head", "--max-time", "60", "--user-agent", "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/128.0.0.0 Safari/537.36 ArchiveBox/{VERSION} (+https://github.com/ArchiveBox/ArchiveBox/)", "https://matthewbutterick.com/extinction-level-capitalism.html" ], "cmd_version": "8.10.1", "end_ts": "2026-06-15T10:10:54.264466+00:00", "index_texts": null, "output": "headers.json", "pwd": "/data/archive/1781518242.322736", "schema": "ArchiveResult", "start_ts": "2026-06-15T10:10:54.223229+00:00", "status": "succeeded" } ], "htmltotext": [ { "cmd": [ "(internal) archivebox.extractors.htmltotext", "./{singlefile,dom}.html" ], "cmd_version": "0.8.5rc51", "end_ts": "2026-06-15T10:11:37.452617+00:00", "index_texts": [ "Matthew Butterick | Extinction-level capitalism (/elc/styles.css) Extinction-level capitalism a citizen\u2019s thoughtson AI risk AI is inher\u00adently polit\u00adical tech\u00adnology. If AI works as intended, it will grad\u00adu\u00adally corrode our liberal democ\u00adracy, risking an irre\u00adversible shift into another polit\u00adical and economic config\u00adu\u00adra\u00adtion. Among AI risks, this one deserves\u00a0more consid\u00ader\u00ada\u00adtion, because it requires no addi\u00adtional condi\u00adtions like malign actors or AI malfunc\u00adtion. AI only needs to amplify existing trends, espe\u00adcially around concen\u00adtra\u00adtion of capital. This damage will occur even assuming that in the near term, AI will broadly improve mate\u00adrial well-being. About MB I\u2019m a self-employed (https://practicaltypography.com/) author , (https://mbtype.com/) designer , (https://beautifulracket.com/) programmer , and (https://buttericklaw.com) lawyer . In 2022, I learned that my own works were in the training datasets of gener\u00ada\u00adtive-AI compa\u00adnies. In response, I (https://www.wired.com/story/matthew-butterick-ai-copyright-lawsuits-openai-meta/) invented the first set of lawsuits chal\u00adlenging the legality of these prac\u00adtices. I\u2019m currently (https://buttericklaw.com) co-counsel for plain\u00adtiffs in a number of AI cases. Though I discuss certain legal issues below, I am not your lawyer, and nothing here is held out as legal advice. These are my personal views as a citizen and economic actor; I speak only for myself. This piece is typeset in (https://mbtype.com/fonts/equity/) Equity , (https://mbtype.com/fonts/advocate/) Advo\u00adcate , and (https://mbtype.com/fonts/triplicate/) Trip\u00adli\u00adcate , fonts I designed. They can be licensed for your own polemics and pamphlets. Emergent effects Two billion years ago, the rock layers comprising what is now called the Colorado Plateau began to form: first igneous and meta\u00admor\u00adphic rocks, followed by many layers of sedi\u00admen\u00adtary rocks. About fifty million years ago, through tectonic action, this plateau gained thou\u00adsands of feet of eleva\u00adtion. About five million years ago, a river began to flow. The river carried silt and debris, scraping out the begin\u00adnings of a canyon. The river deep\u00adened the canyon, exposing its walls to weather and erosional forces that widened the canyon further. Today the waterway is the Colorado River. The geolog\u00adical forma\u00adtion is the Grand Canyon. The forma\u00adtion of the Grand Canyon required zero human agency. Zero tech\u00adnology. Zero coor\u00addi\u00adna\u00adtion among\u00a0the river, the land, and gravity. In that sense the Grand Canyon is an emer\u00adgent effect : a complex, unfore\u00adsee\u00adable output arising from simpler inputs. But we would never wonder whether the river is sentient. Or whether the river cares about the dirt that it carries out of the canyon. The water is just doing what water does: flowing down\u00adhill. The dirt just happens to be in the way. Inherently political technology Langdon Winner is a polit\u00adical theo\u00adrist. Winner wrote the excel\u00adlent and influ\u00aden\u00adtial essay (https://faculty.cc.gatech.edu/~beki/cs4001/Winner.pdf) \u201cDo Arti\u00adfacts Have Poli\u00adtics?\u201d (1980). Winner sought to debunk the tradi\u00adtional framing that \u201ctech\u00adnolo\u00adgies are \u2026\u00a0neutral tools that can be used well or poorly, for good, evil, or some\u00adthing in between.\u201d Instead, Winner proposes two ways that a tech\u00adnology can affect its polit\u00adical envi\u00adron\u00adment: The tech\u00adnology is designed to have certain polit\u00adical effects. For example, the (https://cs.stanford.edu/people/eroberts/cs201/projects/2010-11/FreedomOfInformationChina/category/great-firewall-of-china/index.html) Great Fire\u00adwall of China , a bundle of tech\u00adno\u00adlog\u00adical measures that limit Chinese citi\u00adzens\u2019 access to foreign infor\u00adma\u00adtion sources. Antipo\u00addally, the (https://www.torproject.org/) Tor Project intends to maxi\u00admize user anonymity and thwart govern\u00adment intru\u00adsion. The tech\u00adnology is inher\u00adently polit\u00adical. This is Winner\u2019s key analytic fulcrum. Winner describes two versions of inher\u00adently polit\u00adical tech\u00adnology. The first is where the tech\u00adnology \u201cactu\u00adally requires \u2026\u00a0a partic\u00adular set of social condi\u00adtions as [its] oper\u00adating envi\u00adron\u00adment.\u201d For instance, nuclear weapons: the only respon\u00adsible way to possess such dangerous tech\u00adnology is to place it within \u201ca central\u00adized, rigidly hier\u00adar\u00adchical chain of command \u2026\u00a0the [atom] bomb must be author\u00adi\u00adtarian; there is no other way.\u201d The second version is where the tech\u00adnology is \u201cstrongly compat\u00adible\u201d with a certain polit\u00adical arrange\u00adment (even if not strictly required) and thus tends to bring that arrange\u00adment to fruition. As an example, Winner considers the (https://www.plantsciences.ucdavis.edu/news/how-mechanical-tomato-harvester-prompted-food-movement) mechan\u00adical tomato harvester . Devel\u00adoped at (https://www.ucdavis.edu/research/discoveries/agricultural-discoveries/tomato-harvester) UC Davis in the 1950s, the machine was tremen\u00addously produc\u00adtive. But it was also expen\u00adsive. Only well-capi\u00adtal\u00adized tomato growers could afford it. The rest couldn\u2019t compete. According to Winner, the number of Cali\u00adfornia tomato growers dropped from ~4000 in the early 1960s to ~600 in 1973, costing ~32,000 jobs and the compounding nega\u00adtive effects on those commu\u00adni\u00adties. Winner summa\u00adrizes: What we see here \u2026\u00a0is an ongoing social process in which scien\u00adtific knowl\u00adedge, tech\u00adno\u00adlog\u00adical inven\u00adtion, and corpo\u00adrate profit rein\u00adforce each other in deeply entrenched patterns that bear the unmis\u00adtak\u00adable stamp of polit\u00adical and economic power \u2026\u00a0oppo\u00adnents of inno\u00adva\u00adtions like the tomato harvester are made to seem \u201cantitech\u00adnology\u201d or \u201cantiprogress\u201d. For the harvester is not merely the symbol of a social order that rewards some while punishing others; it is in a true sense an embod\u00adi\u00adment of that order. Not merely the symbol\u2014the embod\u00adi\u00adment. A facially neutral tech\u00adno\u00adlog\u00adical inven\u00adtion\u2014here, a tomato harvester\u2014can induce polit\u00adical effects. Those effects don\u2019t arise from flaws in the tech\u00adnology. To the contrary\u2014they arise from its effi\u00adcacy. How are the polit\u00adical effects deter\u00admined? Winner iden\u00adti\u00adfies two key early deci\u00adsions. The first is the binary ques\u00adtion of whether to pursue the tech\u00adnology at all. The second are choices about \u201cthe design or arrange\u00adment\u201d of the tech\u00adnology. Winner cautions: \u201c[t]o see the matter solely in terms of cost-cutting, effi\u00adciency, or the modern\u00adiza\u00adtion of equip\u00adment is to miss a deci\u00adsive element\u201d. That is, the polit\u00adical effects can possibly be coun\u00adtered, but first they must be acknowl\u00adedged. Of course, the best oppor\u00adtu\u00adnity to choose wisely is before the tech\u00adnology is widely intro\u00adduced, as capital and social invest\u00adment tends to entrench it: Because choices tend to become strongly fixed in mate\u00adrial equip\u00adment, economic invest\u00adment, and social habit, the orig\u00adinal flex\u00adi\u00adbility vanishes for all prac\u00adtical purposes once the initial commit\u00adments are made. In that sense tech\u00adno\u00adlog\u00adical inno\u00adva\u00adtions are similar to legisla\u00adtive acts or polit\u00adical found\u00adings that estab\u00adlish a frame\u00adwork for public order that will endure over many gener\u00ada\u00adtions. \u2026\u00a0The issues that divide or unite people in society are settled not only in the insti\u00adtu\u00adtions and prac\u00adtices of poli\u00adtics proper, but also, and less obvi\u00adously, in tangible arrange\u00adments of steel and concrete, wires and tran\u00adsis\u00adtors, nuts and bolts. Tech\u00adno\u00adlog\u00adical choices bear directly on the \u201cpublic order\u201d at large. When we don\u2019t take those choices seri\u00adously\u2014or we\u2019re persuaded to ignore them by those insisting that tech\u00adnology is just a neutral tool\u2014we risk polit\u00adical conse\u00adquences. Winner warns of compla\u00adcency. Once the tech\u00adnology arrives and becomes entrenched, the conver\u00adsa\u00adtion gets reframed as one of tech\u00adno\u00adlog\u00adical inevitabilism vs. anachro\u00adnism, and dissent is discour\u00adaged: \u201cthe kinds of reasoning that justify the adap\u00adta\u00adtion of social life to tech\u00adnical require\u00adments pop up as spon\u00adta\u00adneously as flowers in the spring\u00a0\u2026\u00a0After a certain point, those who cannot accept the hard require\u00adments and imper\u00ada\u00adtives will be dismissed as dreamers and fools.\u201d Liberal democracy The balance of power of democ\u00adracy is premised on the average person having leverage through creating economic value. If that\u2019s not present, I think things become kind of scary.\u2014a (https://www.axios.com/2025/05/28/ai-jobs-white-collar-unemployment-anthropic#:~:text=The%20balance%20of%20power%20of%20democracy%20is%20premised%20on%20the%20average%20person%20having%20leverage%20through%20creating%20economic%20value%2E%20If%20that%27s%20not%20present%2C%20I%20think%20things%20become%20kind%20of%20scary%2E) certain AI CEO Liberal democ\u00adracy is the polit\u00adical scien\u00adtist\u2019s term for the type of govern\u00adment preva\u00adlent (https://ourworldindata.org/grapher/liberal-democracy-index) among capi\u00adtalist economies since the Amer\u00adican and French Revo\u00adlu\u00adtions. The intel\u00adlec\u00adtual foun\u00adda\u00adtion of liberal democ\u00adracy arose during the Enlight\u00aden\u00adment, espe\u00adcially through the work of (https://plato.stanford.edu/entries/locke-political/) John Locke . Liberal democ\u00adracy empha\u00adsizes limited govern\u00adment, indi\u00advidual rights, and sepa\u00adra\u00adtion of powers\u2014in short, majority rule with excep\u00adtions and guardrails. (The term liberal democ\u00adracy doesn\u2019t connote liberals or Democ\u00adrats in the specific US polit\u00adical sense. But polit\u00adical parties of differing ideolo\u00adgies are a tradi\u00adtional feature of liberal democ\u00adra\u00adcies.) Today, most liberal democ\u00adra\u00adcies are in Europe, the Amer\u00adicas, and the Pacific Rim. Liberal democ\u00adracy is not a fixed set of immutable char\u00adac\u00adter\u00adis\u00adtics, but a bundle of graded values. All liberal democ\u00adra\u00adcies empha\u00adsize certain ones over others. In the aggre\u00adgate, some of these nations evolve toward stronger liber\u00adalism; others (https://freedomhouse.org/report/freedom-world/2026/growing-shadow-autocracy) evolve away . These degraded cases have some\u00adtimes been called illib\u00aderal democ\u00adracy : the observ\u00adable formal\u00adi\u00adties of liberal democ\u00adracy may still be observed\u2014e.g. multi\u00adparty elec\u00adtions, sepa\u00adra\u00adtion of powers\u2014but the lived reality is single-party rule and declining indi\u00advidual rights. That\u2019s not to say that liberal democ\u00adracy produces excel\u00adlent outcomes for all citi\u00adzens, all the time. It doesn\u2019t. At any moment, certain citi\u00adzens are dissat\u00adis\u00adfied\u2014say, because they\u00a0belong to a group whose rights are inad\u00ade\u00adquately protected or econom\u00adi\u00adcally margin\u00adal\u00adized. Liberal democ\u00adracy offers a process, not a result: grass\u00adroots demo\u00adc\u00adratic partic\u00adi\u00adpa\u00adtion can coalesce into policy change. But within the arena of competing polit\u00adical inter\u00adests, winners and losers neces\u00adsarily follow. Navi\u00adgating these differ\u00adences within a stable, accom\u00admoda\u00adtive frame\u00adwork is prefer\u00adable to a rigid one that buckles under these stresses\u2014say, through polit\u00adical revo\u00adlu\u00adtion, which tends to be messy and unpre\u00addictable. Capi\u00adtalism has tradi\u00adtion\u00adally been consid\u00adered a neces\u00adsary but not suffi\u00adcient condi\u00adtion for liberal democ\u00adracy. Why? A regu\u00adlated market economy encour\u00adages citizen partic\u00adi\u00adpa\u00adtion through (https://fee.org/articles/democracy-and-capitalism-are-mutually-reinforcing/) prop\u00aderty owner\u00adship and trans\u00adac\u00adtion . A prin\u00adcipal func\u00adtion of govern\u00adment is to define the economic condi\u00adtions of the state. Economic partic\u00adi\u00adpa\u00adtion mutu\u00adally rein\u00adforces demo\u00adc\u00adratic partic\u00adi\u00adpa\u00adtion. Prop\u00aderty owners will vote for those who protect their inter\u00adests. The rise of indus\u00adtrial capi\u00adtalism in the 19th century, and the wealth-redis\u00adtri\u00adb\u00adu\u00adtion mech\u00ada\u00adnisms that followed in the 20th, led to economic empow\u00ader\u00adment for more citi\u00adzens, and ulti\u00admately broader polit\u00adical empow\u00ader\u00adment. The converse also holds: economies premised on state or oligarchic control of some narrow class of assets haven\u2019t tended to evolve toward liberal democ\u00adracy. In prac\u00adtice, certain people in a capi\u00adtalist liberal democ\u00adracy tend to get increas\u00adingly rich. Absent coun\u00adter\u00admea\u00adsures, the wealthy gain control of the polit\u00adical appa\u00adratus, thwarting liberal-demo\u00adc\u00adratic norms. This tension between capital and poli\u00adtics is a long-consid\u00adered topic. A key early work was, of course, Karl Marx\u2019s Capital (about which more later). In the current era, Mancur Olson\u2019s book (https://dn720004.ca.archive.org/0/items/olson-mancur-rise-and-decline-of-nations-economic-growth-stagflation/Olson%2C_Mancur_Rise_and_Decline_of_Nations_Economic_Growth%2C_Stagflation.pdf) The Rise and Decline of Nations set out how small groups with a shared interest (which could include capital concen\u00adtra\u00adtion) can effec\u00adtively under\u00admine stable soci\u00adeties. More recently, econ\u00ado\u00admists Robert Reich ((https://cooperative-individualism.org/reich-robert_how-capitalism-is-killing-democracy-2007-sep-oct.pdf) \u201cHow Capi\u00adtalism is Killing Democ\u00adracy\u201d ), James Galbraith ((https://www.simonandschuster.com/books/The-Predator-State/James-K-Galbraith/9781416576211) The Predator State ), and Yanis Varo\u00adufakis ((https://www.penguinrandomhouse.com/books/751443/technofeudalism-by-yanis-varoufakis/) Tech\u00adnofeu\u00addalism: What Killed Capi\u00adtalism ) are among those who have studied the esca\u00adlating polit\u00adical conse\u00adquences of rising wealth inequality. The synthesis might be: as more wealth becomes concen\u00adtrated in the hands of fewer citi\u00adzens, liberal democ\u00adracy weakens, because whichever citi\u00adzens are losing economic rele\u00advance will also lose polit\u00adical rele\u00advance. A nation sending many of its citi\u00adzens toward economic irrel\u00ade\u00advance risks becoming polit\u00adi\u00adcally illib\u00aderal. The Skynet fallacy AI discourse often invokes sci-fi narra\u00adtives. (https://matthewbutterick.com/chron/domination.html#:~:text=Skynet%20fallacy) I\u2019ve called this the Skynet fallacy , after the (https://terminator.fandom.com/wiki/Skynet) Termi\u00adnator antag\u00ado\u00adnist , the most cited. But any sci-fi will do. For instance, (https://www.unilad.com/technology/elon-musk-ai-technology-terminator-humanity-930217-20230525) one AI CEO warned of AI \u201cgoing Termi\u00adnator\u201d; Stephen Hawking and other (https://www.independent.co.uk/news/science/stephen-hawking-transcendence-looks-at-the-implications-of-artificial-intelligence-but-are-we-taking-ai-seriously-enough-9313474.html) scien\u00adtists warned of AI \u201cdevel\u00adoping weapons we cannot even under\u00adstand\u201d; a second (https://www.bloomberg.com/news/articles/2023-12-12/sam-altman-defends-ai-thrust-days-after-retaking-openai-s-helm) AI CEO said we \u201cdon\u2019t need much imag\u00adi\u00adna\u00adtion [about AI risk] because we grew up with that in the media\u201d; a third (https://darioamodei.com/essay/the-adolescence-of-technology) AI CEO invoked the sci-fi movies Contact and 2001: A Space Odyssey in a piece about AI risk; a (https://www.politico.com/news/2024/05/12/ai-lobbyists-gain-upper-hand-washington-00157437) US congressman summa\u00adrized AI risk as \u201cevil robots rising up to take over the world\u201d; a well-known (https://www.slowboring.com/p/the-case-for-terminator-analogies) jour\u00adnalist advo\u00adcated for more Termi\u00adnator analo\u00adgies; a promi\u00adnent (https://intelligence.org/2018/02/28/sam-harris-and-eliezer-yudkowsky/#:~:text=annoyed%20with%20the%20journalists%20who%20insist%20on%20putting%20a%20picture%20of%20the%20Terminator) AI-risk pundit said he\u2019s \u201cannoyed\u201d with Termi\u00adnator analo\u00adgies yet (https://www.lesswrong.com/posts/uMQ3cqWDPHhjtiesc/agi-ruin-a-list-of-lethalities#:~:text=nanomachinery%20builds%20diamondoid%20bacteria) has suggested that AI will erad\u00adi\u00adcate humanity using hordes of toxic nanobots. Artis\u00adti\u00adcally, sci-fi movies exter\u00adnalize the awe and unease of tech\u00adno\u00adlog\u00adical confronta\u00adtion. It\u2019s easy to see why these metaphors have become part of AI-risk discourse. And yet. As AI puts down roots in our economy, the sci-fi framing hides more than it reveals. Sci-fi plots are opti\u00admized for cine\u00admatic impact. So as a metaphor for AI risk, they can lead to faulty intu\u00aditions. Among real\u00adistic AI risks, we can expect that most will be boring, slow, and depend on minimal extra tech\u00adnology. Whether AI will cause literal human extinc\u00adtion is esoteric\u2014a light\u00adning strike. But AI could easily induce future economic and polit\u00adical condi\u00adtions that most Amer\u00adi\u00adcans today would consider intol\u00ader\u00adable\u2014a cancer that extin\u00adguishes a certain way of life. Nobody\u2019s going to make a movie about boring AI risks. But they comprise the majority of worri\u00adsome AI outcomes. In 2003, philoso\u00adpher (https://nickbostrom.com/ethics/ai#:~:text=manufacture%20as%20many%20paperclips%20as%20possible) Nick Bostrom proposed his now well-known parable of the paper\u00adclip maxi\u00admizer. Bostrom imag\u00adines an advanced AI that is asked to make paper\u00adclips. Taking its mission seri\u00adously, the AI erad\u00adi\u00adcates humanity as it consumes all Earth resources to make more paper\u00adclips. Bostrom was illus\u00adtrating the AI control problem : ensuring an AI acts consis\u00adtently with human prior\u00adi\u00adties is diffi\u00adcult, even for osten\u00adsibly simple goals. Because when we say \u201cAI, make paper\u00adclips\u201d, the implied coda is \u201c\u2026\u00a0without killing everyone.\u201d But an AI can\u2019t know this ex nihilo . Bostrom softens the sci-fi flavor by choosing paper\u00adclips and not, say, laser-wielding robots. Bostrom\u2019s choice of an economic mech\u00ada\u00adnism of resource conver\u00adsion is apt. Even a mundane objec\u00adtive can produce outsize risk. We could further observe that on the paper\u00adclip-maxi\u00admizing path, human life would become dystopic long before literal extinc\u00adtion. As resources are depleted, humans would become tenants in a neofeu\u00addalist paper\u00adclip empire. (Paper\u00adclip Crisis: The Saga Begins \u2014opening soon.) Computer scien\u00adtist (https://people.eecs.berkeley.edu/~russell/) Stuart Russell also explored the control problem in his book (https://people.eecs.berkeley.edu/~russell/hc.html) Human Compat\u00adible . Russell calls one variant the gorilla problem : that \u201cances\u00adtors of the modern gorilla created \u2026\u00a0the genetic lineage leading to modern humans. How do the gorillas feel about this? \u2026\u00a0the consensus opinion would be very nega\u00adtive indeed.\u201d Despite shared lineage, gorillas and humans have incom\u00admen\u00adsu\u00adrable values. Nothing humans can do\u2014short of disap\u00adpearing\u2014would restore gorillas to their golden age. Russell\u2019s framing is a believ\u00adable analogy for the future rela\u00adtion\u00adship of humans and AI. Sure, gorillas didn\u2019t delib\u00ader\u00adately invent humans. But I take Russell to mean that the emer\u00adgent char\u00adac\u00adter\u00adis\u00adtics of these rela\u00adtion\u00adships are more conse\u00adquen\u00adtial than the intended ones. (In that sense, Russell echoes Langdon Winner .) The fact that we\u2019re inventing AI doesn\u2019t mean we will predict or control its gravest effects. Any more than gorillas could predict or control human domi\u00adnance of their ecosystem. The gorillas did their thing. We did ours. AI will do its thing too. It will take time to figure out what that is, exactly. How the West was won I recently read (https://www.penguinrandomhouse.com/books/323685/cadillac-desert-by-marc-reisner/) Cadillac Desert by (https://californiacurated.com/blog/cadillac-desert-how-marc-reisner-changed-the-way-we-see-water/) Marc Reisner , about the devel\u00adop\u00adment of water\u00a0resources in the western US between 1910 and 1980, espe\u00adcially the dam-building campaign of the federal (https://www.usbr.gov/) Bureau of Recla\u00adma\u00adtion . Reisner\u2019s book weaves several story\u00adlines: Engi\u00adneering and envi\u00adron\u00admental impacts . Dams concen\u00adtrate water. But they cause envi\u00adron\u00admental conse\u00adquences else\u00adwhere. Further\u00admore, Recla\u00adma\u00adtion\u2019s 20th-century projects were often based on opti\u00admistic projec\u00adtions of mete\u00ado\u00adro\u00adlog\u00adical water supply. Today, long-term drought condi\u00adtions chal\u00adlenge those projec\u00adtions. No amount of money or hydro\u00adlogic engi\u00adneering can change that. National US poli\u00adtics . In the early 20th century, newer Western states sought polit\u00adical clout in the federal govern\u00adment, which had been domi\u00adnated by Eastern states. Polit\u00adical clout followed economic growth, and to achieve growth, the Western states depended on one crit\u00adical but scarce input: water. In the West, the federal Bureau of Recla\u00adma\u00adtion was tasked with increasing water supplies. A polit\u00adical symbiosis emerged: congressmen from Western states voted to fund Recla\u00adma\u00adtion water projects; in turn, Recla\u00adma\u00adtion looked out for their constituents. Over decades, Reisner depicts this rela\u00adtion\u00adship as metas\u00adta\u00adsizing from prac\u00adti\u00adcality to corrup\u00adtion, in the sense of Recla\u00adma\u00adtion becoming beholden to a narrow polit\u00adical lane. In that sense, Recla\u00adma\u00adtion\u2019s dams arguably qualify as inher\u00adently polit\u00adical tech\u00adnology . State economies . Recla\u00adma\u00adtion\u2019s highly subsi\u00addized water boot\u00adstrapped Western economies, espe\u00adcially agri\u00adcul\u00adture. For a while, it worked as promised: Farmers got irri\u00adga\u00adtion. Cities got water. Western states grew and pros\u00adpered. But the projects worked so well that Western states wanted more. These states never weaned them\u00adselves from subsi\u00addized federal water, setting them on a path toward perma\u00adnent depen\u00addence. The parallel between water and AI is inexact. Water is a biolog\u00adical neces\u00adsity; AI is not. Recla\u00adma\u00adtion\u2019s projects worked (up to a point); AI may or may not. This is part of why AI propo\u00adnents have sought to raise the stakes. So far, AI has been (https://www.wheresyoured.at/ai-is-too-expensive/) grue\u00adsomely expen\u00adsive and deliv\u00adered (https://www.wheresyoured.at/ai-is-slowing-down/) middling results . Never\u00adthe\u00adless it\u2019s routinely depicted as a (https://www.foreignaffairs.com/united-states/artificial-intelligence-grand-bargain-buchanan-collins) geopo\u00adlit\u00adical fulcrum , a proxy for contin\u00aduing US excep\u00adtion\u00adalism. If Amer\u00adi\u00adcans don\u2019t adopt AI whole\u00adheart\u00adedly, (https://www.pbs.org/newshour/show/trump-considers-order-to-override-state-regulations-on-artificial-intelligence#:~:text=If%20we%20don%27t%2C%20then%20China%20will%20easily%20catch%20us%20in%20the%20A%2EI%2E%20race) we will be losers . Do you want to be a loser? Labor replacement Q: What is Big AI primarily selling? A: Labor replace\u00adment, with mass unem\u00adploy\u00adment as a likely conse\u00adquence. Some (https://www.nytimes.com/2026/05/22/opinion/ai-job-crisis-goldman-sachs.html) disagree or call it (https://nypost.com/2025/12/21/opinion/dont-trust-ai-doomers-tech-wont-steal-jobs-but-make-them-better/) doomerish . Why? It\u2019s exactly what AI grandees have been telling us. A certain AI CEO (https://www.cnbc.com/2025/01/21/bill-gates-my-favorite-book-on-ai.html) wrote that AI \u201cwill be hugely desta\u00adbi\u00adlizing for hundreds of millions\u201d and that AI tools \u201care funda\u00admen\u00adtally labor replacing\u201d. A certain AI company released a research (https://openai.com/index/gpts-are-gpts/) paper about \u201cthe labor market impact poten\u00adtial of large language models\u201d. That (https://www.theatlantic.com/technology/archive/2023/11/sam-altman-open-ai-fallout/676046/#:~:text=Jobs%20are%20definitely%20going%20to%20go%20away%2C%20full%20stop) AI CEO said \u201cjobs are defi\u00adnitely going to go away, full stop\u201d. A second AI CEO (https://www.axios.com/2025/05/28/ai-jobs-white-collar-unemployment-anthropic) said that in the near future, \u201c20% of people don\u2019t have jobs.\u201d A third AI CEO (https://finance.yahoo.com/news/probably-none-us-job-elon-200017498.html) predicted that farther out, \u201cprob\u00adably none of us will have a job.\u201d An (https://fortune.com/article/palantir-ceo-alex-karp-ai-humanities-jobs-vocational-training/) AI-adja\u00adcent CEO said that AI \u201cwill destroy human\u00adi\u00adties jobs\u201d. The ball is not hidden. Capital markets are already pricing in these expec\u00adta\u00adtions. Regard\u00adless of whether Big AI even\u00adtu\u00adally delivers mass labor replace\u00adment, today these compa\u00adnies seek to concen\u00adtrate capital as if they will. According to the Wash\u00adington Post, AI capital expen\u00addi\u00adture in 2026 (https://www.washingtonpost.com/technology/2026/02/07/ai-spending-economy-shortages/) is esti\u00admated to be $700 billion, a \u201cspending spree [that] has few prece\u00addents\u201d. Based on a (https://epoch.ai/blog/half-of-employed-ai-users-now-use-it-for-work/) recent survey of US workers, the (https://www.nbcnews.com/tech/tech-news/ai-job-work-replace-task-help-rcna267238) Global Part\u00adner\u00adship on Arti\u00adfi\u00adcial Intel\u00adli\u00adgence said that \u201cthe policy window to shape how AI trans\u00adforms work is prob\u00adably closing faster than most govern\u00adments realize.\u201d Extra\u00ador\u00addi\u00adnary invest\u00adment demands extra\u00ador\u00addi\u00adnary returns. In early 2026, after a certain billion\u00adaire tech CEO laid off 40% of his employees and attrib\u00aduted the deci\u00adsion to a new (https://www.theinformation.com/newsletters/the-briefing/jack-dorsey-proves-ai-job-cut-fears-true-ellisons-win-warner?rc=uyz1qi#:~:text=core%20thesis) \u201ccore thesis\u201d of AI, the company\u2019s (https://www.theinformation.com/newsletters/the-briefing/jack-dorsey-proves-ai-job-cut-fears-true-ellisons-win-warner?rc=uyz1qi#:~:text=core%20thesis) stock rose nearly 25%. He won\u2019t be the last. Whether these layoffs are based on actual AI bene\u00adfits or (https://hbr.org/2026/01/companies-are-laying-off-workers-because-of-ais-potential-not-its-performance) merely \u201cantic\u00adi\u00adpa\u00adtory\u201d is neither here nor there. Employers have strong incen\u00adtives to reduce head\u00adcount and increase AI spending before competi\u00adtors do. We will (https://www.financialexpress.com/life/technology-intel-amazon-meta-and-other-us-tech-firms-announce-mass-layoffs-in-2025-citing-cost-cutting-and-ai-as-reasons-3888709/) increas\u00adingly see both kinds of layoffs. Soft\u00adware program\u00admers are one set of conse\u00adquen\u00adtial, highly paid writers (https://www.nytimes.com/2026/03/12/magazine/ai-coding-programming-jobs-claude-chatgpt.html) likely to be replaced with AI . Else\u00adwhere (https://matthewbutterick.com/chron/layers-of-lawyers-and-liars.html#laws-two-vulnerabilities) I\u2019ve predicted that legal prac\u00adtice will also be seri\u00adously impaired. Why? Like program\u00admers, lawyers write about (https://typographyforlawyers.com/introduction.html) conse\u00adquen\u00adtial things and charge a lot. (We can expect that AI compa\u00adnies them\u00adselves, fond of (https://www.geekwire.com/2025/eat-your-own-dog-food-how-microsoft-popularized-one-of-the-yuckiest-terms-in-tech-history/) dogfooding , will perfor\u00adma\u00adtively (https://www.businessinsider.com/meta-plans-layoff-10-of-its-entire-staff-may-2026) set an example through their own AI-driven layoffs.) Two common objec\u00adtions to this conclu\u00adsion: \u201cAI won\u2019t replace workers, it will enhance them.\u201d That may turn out to be true. But for now, it\u2019s not what AI CEOs are pitching. And it\u2019s not how the (https://www.reuters.com/world/spacex-ipo-demand-is-approaching-four-times-oversubscribed-source-says-2026-06-09/) market is valuing Big AI. AI customers will natu\u00adrally pay less for a worker-assis\u00adtive tech\u00adnology than tech\u00adnology that reduces head\u00adcount, which is one of the biggest expenses at any employer. The labor-replace\u00adment story is every\u00adthing to every\u00adbody. If AI is just a neat way of, say, finding photos of your dog\u00a0on your phone, none of the AI bets pay out. \u201cAI will replace certain jobs, but also create new ones.\u201d Fair, but \u201ccreate new ones\u201d is skimpy on details. How do the jobs created compare to the ones lost? Do they pay more? Or less, as the supply of workers seeking work increases? Do the new jobs require different skills? Who bears the cost of the career switch? None of this is auto\u00admatic. As econ\u00ado\u00admist (https://www.nytimes.com/2026/04/30/opinion/ai-labor-work-force-silicon-valley.html#:~:text=the%20short%20run%20can%20be%20a%20lifetime) Carl Benedikt Frey said : \u201cMost econ\u00ado\u00admists will acknowl\u00adedge that tech\u00adno\u00adlog\u00adical progress can cause some adjust\u00adment prob\u00adlems in the short run. What is rarely noted is that the short run can be a life\u00adtime.\u201d After going hard on this narra\u00adtive for several years, AI CEOs are now (https://www.businessinsider.com/ai-job-impact-losses-shifting-sam-altman-mustafa-suleyman-comments-2026-6) soft\u00adening their tone , primarily in response to (https://www.theguardian.com/technology/2026/may/26/students-boo-pro-ai-graduation-speakers) public back\u00adlash . For instance, the AI CEO (https://www.theatlantic.com/technology/archive/2023/11/sam-altman-open-ai-fallout/676046/#:~:text=Jobs%20are%20definitely%20going%20to%20go%20away%2C%20full%20stop) who once said \u201cjobs are defi\u00adnitely going to go away\u201d now (https://www.businessinsider.com/ai-job-impact-losses-shifting-sam-altman-mustafa-suleyman-comments-2026-6#:~:text=the%20human%20part%20of%20the%20roles) wants us to know that watching AI answer his emails felt \u201cdehu\u00adman\u00adizing\u201d and that he now values \u201cthe human part of the roles\u201d. That\u2019s what it finally took\u2014using your own product? Ed Zitron (https://www.wheresyoured.at/the-revenge-of-the-business-idiot/#:~:text=have%20little%20idea%20what%20work%20looks%20like) has char\u00adac\u00adter\u00adized the AI market so far as a blink\u00adered conver\u00adsa\u00adtion between AI exec\u00adu\u00adtives and trad exec\u00adu\u00adtives, all of whom \u201chave little idea what work looks like\u201d. This is more a feature than a bug. To replace labor, AI doesn\u2019t need to (https://www.damiencharlotin.com/hallucinations/) actu\u00adally deliver compa\u00adrable outputs. It only needs to fulfill a certain exec\u00adu\u00adtive fantasy of \u201cwhat work looks like\u201d. That\u2019s much easier than, say, achieving AGI (however defined). On that view, AI may turn out to be a (https://www.economicshelp.org/blog/1164/economics/veblen-goods/) Veblen good \u2014an item that buyers value primarily for signaling their ability to afford it. One more small matter\u2014funding of the US govern\u00adment is premised on employ\u00adment. In 2023, (https://usafacts.org/articles/how-much-revenue-does-the-federal-government-collect/) the US got 48% of its revenue from income taxes and 36% from other payroll taxes. If AI causes 30% unem\u00adploy\u00adment, revenues will drop and expenses will jump. Of course, higher unem\u00adploy\u00adment also leads to lower aggre\u00adgate consumer spending. These would be fiscal condi\u00adtions unknown since the Great Depres\u00adsion, and utterly foreign to Amer\u00adi\u00adcans living today. Inter\u00admis\u00adsion On its current trajec\u00adtory, I\u2019m person\u00adally skep\u00adtical that AI can deliver economic returns. Over time, I expect that AI will evolve toward a bifur\u00adcated market: Consumer AI . Today, consumers have cheap, high-volume access to fron\u00adtier models for, e.g., retouching dog selfies and writing dating profiles. This is unsus\u00adtain\u00adable and there\u00adfore must end. I expect these uses will increas\u00adingly be handled by on-device models that are either free (perhaps open-source) or built into the OS. The only compa\u00adnies that can be prof\u00aditable in this cate\u00adgory will be those who control a device ecosystem and can subsi\u00addize model devel\u00adop\u00adment through indi\u00adrect revenue streams. Labor-replacing AI services . These will offer models with substan\u00adtially higher complexity and value, but at substan\u00adtially higher price. I expect most market compe\u00adti\u00adtion will be in this second cate\u00adgory. Though it\u2019s not yet apparent that this cate\u00adgory will discover any work\u00adable customer\u2013product equi\u00adlib\u00adrium. For customers, tokens have to be cheaper than the wages being replaced. For AI suppliers, tokens will have to be wildly prof\u00aditable, or these compa\u00adnies will have to find addi\u00adtional ways of extracting rents from customers. However it evolves, I expect this market will be smaller than today\u2019s data\u00adcenter investors foresee, and they will (https://moiglobal.com/lattice-work-hyperscalers/) follow the path of many fiber-optic investors of the 1990s: into bank\u00adruptcy court. The goodies economy Big AI\u2019s message to poten\u00adtial corpo\u00adrate customers of AI has centered on labor replace\u00adment. Correctly fore\u00adseeing that the workers them\u00adselves might take umbrage, Big AI has pitched them a sepa\u00adrate story: that workers will enjoy higher stan\u00addards of mate\u00adrial well-being under AI. One (https://www.foxbusiness.com/economy/musk-predicts-ai-create-universal-high-income-make-saving-money-unnecessary) billion\u00adaire AI CEO predicted \u201cuniversal high income\u201d and that \u201cAI and robots will provide any goods and services that you want.\u201d Another (https://finance.yahoo.com/news/sam-altman-wants-universal-extreme-124300850.html) billion\u00adaire AI CEO predicted \u201cuniversal extreme wealth for every\u00adbody\u201d. A (https://a16z.com/the-techno-optimist-manifesto/) billion\u00adaire AI venture capi\u00adtalist \u201cbelieve[s] in making everyone rich, every\u00adthing cheap, and every\u00adthing abun\u00addant\u201d and that \u201cthe ulti\u00admate result will be that all phys\u00adical goods become as cheap as pencils.\u201d A (https://finance.yahoo.com/news/bill-gates-says-ai-change-143901693.html) tech billion\u00adaire predicted \u201cmore goods and services with less labor\u201d and \u201calways-avail\u00adable, high-quality medical advice.\u201d Another (https://fortune.com/2026/03/04/vinod-khosla-investor-openai-free-ai-labor-no-work-few-jobs-abundance/) billion\u00adaire AI venture capi\u00adtalist predicted \u201cthe abun\u00addance of goods and services \u2026\u00a0will be very, very large. Prices will be very, very, low.\u201d Hope\u00adfully yachts will come down too. But wait\u2014will they? In the post-AI economy, will we all have yachts? Of course not. As usual, we should be skep\u00adtical of billion\u00adaires. Espe\u00adcially those suddenly\u00a0claiming to care about the purchasing power of workers. Are they fibbers or just fools? I think it\u2019s more subtle: these billion\u00adaires are performing an idio\u00adsyn\u00adcratic kind of class empathy, prob\u00adably at the behest of publi\u00adcists. This feeling doesn\u2019t come natu\u00adrally. So, like a cyborg reading a love poem, it\u2019s stilted. The persis\u00adtent lack of detail signals that these billion\u00adaires have no idea how, exactly, AI will accom\u00adplish this. But the subtext is clear: If toler\u00adating worker purchasing power is what it takes to become a tril\u00adlion\u00adaire, that\u2019s a sacri\u00adfice I\u2019m willing to make. Some\u00adtimes this theory goes by fancy names like pros\u00adperity or abun\u00addance . Some\u00adtimes it\u2019s linked to specific policy mech\u00ada\u00adnisms like universal basic income . It all comes back to the same idea. I\u2019m just going to call this stuff goodies . The theory is that under AI, citi\u00adzens are going to get some nice level of goodies. A few reasons we might be skep\u00adtical of AI-goodies theory. The price of any consumer good has a lower bound set by the cost of its inputs (if one is not selling at a loss). But labor is only one input. So even if AI reduces labor costs to $0, the cost of other inputs\u2014e.g., energy and raw mate\u00adrials\u2014remain. Even if AI lowers commodity prices by increasing their supply (say, through better discovery and extrac\u00adtion tech\u00adniques), they are still rival\u00adrous resources. Not all are avail\u00adable domes\u00adti\u00adcally. Postwar, rising global consump\u00adtion of oil in the West influ\u00adenced the (https://jacobin.com/2023/09/opec-founding-decolonization-oil-production-western-imperialism) forma\u00adtion of OPEC by future petrostates who sought to capture a share of oil-derived wealth. Simi\u00adlarly, in a post-AI US economy, foreign nations with crit\u00adical minerals\u2014e.g., lithium, cobalt, nickel\u2014may like\u00adwise cartelize them, raising costs and creating a price floor. Related kinds of supply control may also arise domes\u00adti\u00adcally. Today, a certain US agri\u00adcul\u00adture company uses (https://www.theguardian.com/environment/2013/feb/12/monsanto-sues-farmers-seed-patents) patented seeds to exert control over a market of osten\u00adsibly inde\u00adpen\u00addent farmers. If AI dramat\u00adi\u00adcally reduces the costs of manu\u00adfac\u00adturing and agri\u00adcul\u00adture, Big AI may seek to capture value from the facto\u00adries and farms that benefit (say, through a sales royalty). AI-goodies theory implies that Big AI will allow maximum produc\u00adtion of goods, leading to lowest prices. But the huge finan\u00adcial projec\u00adtions around AI\u2014set by Big AI itself\u2014may force them to do other\u00adwise. That is, Big AI\u2019s promises of world-histor\u00adical goodies for consumers and world-histor\u00adical profits for share\u00adholders may be incom\u00adpat\u00adible. If so\u2014guess who prevails? Still, to be fair to Big AI, I\u2019ll accept their argu\u00adment that citi\u00adzens will enjoy higher stan\u00addards of mate\u00adrial well-being under AI through goodies. But the goodies must be paid for. Somehow, wealth must be trans\u00adferred from Big AI to everyone else. These trans\u00adfers can happen by three basic methods: Direct . Big AI directly provides goodies to citi\u00adzens. As one (https://www.nytimes.com/2023/03/31/technology/sam-altman-open-ai-chatgpt.html#:~:text=grand%20idea) AI CEO put it , to \u201ccapture much of the world\u2019s wealth \u2026 and then redis\u00adtribute this wealth to the people\u201d. Tax and spend . Govern\u00adment taxes Big AI at some high rate, and spends the money on public goodies. But the idea that Amer\u00adican AI compa\u00adnies would consent to high tax rates to ensure public goodies is incon\u00adceiv\u00adable because it vests more control in govern\u00adment, which (https://builtin.com/artificial-intelligence/ai-super-pacs-2026-midterm-election-regulation) no free-marke\u00adteer believes should be so.\u2021 Market-based . Big AI dramat\u00adi\u00adcally reduces the cost of goods and services, which then flow into the market at extremely low prices. The flaw in this method is that AI will reduce prices of certain things more than others. For instance, it\u2019s easy to see how AI will make streaming movies free; food and health\u00adcare and trans\u00adporta\u00adtion, less easy. These methods aren\u2019t mutu\u00adally exclu\u00adsive. I expect the most likely method is the first, where Big AI directly provides the goodies, in the style of (https://savingplaces.org/stories/in-good-company) Gilded Age company towns . Big AI would avoid govern\u00adment inter\u00adven\u00adtion through taxa\u00adtion and redis\u00adtri\u00adb\u00adu\u00adtion. Big AI would control the goodies and who gets them. Big AI would be clearly under\u00adstood as the goodie provider. Assuming the massive profits mate\u00adri\u00adalize, Big AI could afford to make diverse baskets of goodies that include even things like (https://www.theguardian.com/commentisfree/2024/apr/10/amazon-ai-cashier-less-shops-humans-technology) food and (https://health.amazon.com/) health\u00adcare and (https://waymo.com/) trans\u00adporta\u00adtion . Keep in mind the goodies economy will play out against a back\u00addrop of wors\u00adening AI-induced labor-market effects\u2014high unem\u00adploy\u00adment, but also wage declines, under\u00adem\u00adploy\u00adment, misem\u00adploy\u00adment, and so forth. There\u00adfore, a side effect of the goodies economy is that what\u00adever we think consti\u00adtutes the social safety net will shift from a service of govern\u00adment to a service of private industry, specif\u00adi\u00adcally Big AI. One (https://darioamodei.com/essay/machines-of-loving-grace#:~:text=clear%20opportunity%20for%20AI%20to%20be%20used%20to%20help%20provision%20government%20services) AI CEO shame\u00adlessly predicts exactly this outcome: that there is \u201coppor\u00adtu\u00adnity for AI to be used to help provi\u00adsion govern\u00adment services \u2026\u00a0includ[ing] health services, the DMV, taxes, social secu\u00adrity, building code enforce\u00adment and so on.\u201d But don\u2019t worry\u2014he concludes that because \u201cpoorly imple\u00admented services\u201d cause \u201ccyni\u00adcism about govern\u00adment\u201d, AI inter\u00adme\u00addi\u00ada\u00adtion would \u201cstrengthen[] respect for demo\u00adc\u00adratic gover\u00adnance.\u201d And to be clear\u2014by \u201cAI\u201d, he means \u201cmy privately owned AI\u201d, and by \u201coppor\u00adtu\u00adnity \u2026\u00a0to help\u201d he means a (https://www.anthropic.com/news/anthropic-and-the-department-of-defense-to-advance-responsible-ai-in-defense-operations) massive govern\u00adment contract . This portends a polit\u00adical shift. Citi\u00adzens once treated govern\u00adment as a polit\u00adical bulwark against the encroach\u00adments and over\u00adreaches of industry. In the goodies economy, citi\u00adzens will tend to align with Big AI, since it provides the goodies, which for many citi\u00adzens may include basic needs. Under those condi\u00adtions, if Big AI has a conflict with govern\u00adment\u2014whose side will citi\u00adzens take? Indus\u00adtries have often insu\u00adlated them\u00adselves against polit\u00adical conse\u00adquences with popular prod\u00aducts. Arguably, in the last 25 years, the tech industry\u2014espe\u00adcially smart\u00adphone and social-media compa\u00adnies\u2014has bene\u00adfited from light regu\u00adla\u00adtory scrutiny rela\u00adtive to its size. Exam\u00adples from the last century include the auto\u00admo\u00adtive and fast-food indus\u00adtries, which like\u00adwise enjoyed decades of gentle regu\u00adla\u00adtion, even though both had serious public-health impacts. To improve public accep\u00adtance, Big AI has so far empha\u00adsized their fun consumer-facing prod\u00aducts. These prod\u00aducts\u2014e.g., gener\u00adating an image, or writing a story, or riding in a driver\u00adless car\u2014make AI seem harm\u00adless: the citizen thinks of their own usage as trivial and fun, and there\u00adfore models the aggre\u00adgate effect of AI as a sum of simi\u00adlarly trivial indi\u00advidual usages. Consis\u00adtent with this, Big AI has often framed these tools as (https://news.microsoft.com/features/democratizing-ai/#sm.0000v8bqc0rcdflis6g1wekfl455r:~:text=Democratizing%20AI) \u201cdemoc\u00adra\u00adtizing\u201d ((https://blog.samaltman.com/#:~:text=AI%20has%20to%20be%20democratized) ex.\u00a02 , (https://blog.google/innovation-and-ai/products/sundar-pichai-ai-action-summit/#:~:text=it%20will%20do%20more%20to%20democratize%20access%20to%20information) ex.\u00a03 ) implying that oppo\u00adnents of these systems are in some sense anti-democ\u00adracy (or as Winner predicted , antitech\u00adnology and antiprogress). In the goodies economy, as citi\u00adzens lose inde\u00adpen\u00addence as economic actors, they will also lose inde\u00adpen\u00addence as polit\u00adical actors. They will be captured by industry. Econ\u00ado\u00admist Yanis Varo\u00adufakis has called the current early stage of this arrange\u00adment (https://www.penguinrandomhouse.com/books/751443/technofeudalism-by-yanis-varoufakis/) \u201ctech\u00adnofeu\u00addalism\u201d , a term suggesting how late capi\u00adtalism erodes indi\u00advidual economic agency. Through its provi\u00adsion of goodies, Big AI intends to complete this tran\u00adsi\u00adtion. An economic histo\u00adrian might say that a polit\u00adical takeover by Big AI is unlikely, citing successful worker-led polit\u00adical move\u00adments in the US and Europe. If the US reaches 30% AI-induced unem\u00adploy\u00adment, the thinking goes, workers will rise up and demand change through polit\u00adical mech\u00ada\u00adnisms of our liberal democ\u00adracy. There\u2019s reason to doubt these paral\u00adlels. These events occurred earlier in the tech\u00adno\u00adlog\u00adical era, when industry and workers were adver\u00adsarial yet remained inter\u00adde\u00adpen\u00addent. Citi\u00adzens retained irre\u00adducible leverage with both industry (as laborers) and govern\u00adment (as voters). In the decades after the US Gilded Age, this combi\u00adna\u00adtion of economic and elec\u00adtoral power was crit\u00adical to the (https://www.history.com/articles/gilded-age-progressive-era-reforms) rising status of workers . But in a future post-AI economy, workers could lose their jobs because of Big AI while remaining depen\u00addent on Big AI for basic needs. How do you bite the hand that feeds? Further\u00admore, many previous conflicts between industry and workers were not gentle adjust\u00adments, but violent (say, the (https://www.smithsonianmag.com/history/what-the-luddites-really-fought-against-264412/) Luddite move\u00adment of the 1810s, and (https://www.pbs.org/wgbh/americanexperience/features/theminewars-labor-wars-us/) Gilded Age strikes of the late 1800s) or caused devas\u00adtating, perma\u00adnent social tran\u00adsi\u00adtions (say, the 1917 (https://encyclopedia.1914-1918-online.net/article/labour-movements-trade-unions-and-strikes-russian-empire/) Russian Revo\u00adlu\u00adtion ). Legally, shifting the social safety net from govern\u00adment to Big AI also means that citizen enti\u00adtle\u00adment to these goodies will shift from public statu\u00adtory and consti\u00adtu\u00adtional law to private-contract law. Private compa\u00adnies can with\u00adhold goodies for many reasons the govern\u00adment cannot. For instance: if you talk shit about the govern\u00adment, you will still get your unem\u00adploy\u00adment check, because the First Amend\u00adment says so. But no such restric\u00adtion on Big AI: if you talk shit about them, they will be free to revoke your goodies. Can\u2019t happen? Already does: (https://www.nclc.org/resources/comments-opposing-cfpb-proposal-to-reduce-number-of-consumer-reporting-agencies-supervised-by-the-bureau/) credit reports are a private system that largely controls who gets favor\u00adable interest rates. Or employer-provided health\u00adcare, a struc\u00adturally unnec\u00ades\u00adsary tradi\u00adtion that (https://www.ajmc.com/view/improve-american-health-care-by-moving-away-from-employer-sponsored-insurance) distorts incen\u00adtives for workers by making their continued health contin\u00adgent on continued employ\u00adment. Most likely, the polit\u00adical align\u00adment between citi\u00adzens and Big AI will not only be enforced by carrot, but also by stick. Of course, an industry seeking to induce polit\u00adical align\u00adment from citi\u00adzens would benefit not only from distrib\u00aduting goodies, but also weak\u00adening the govern\u00adment. In 2023, would you have believed this predic\u00adtion?\u00a0That an AI CEO would (https://www.cnbc.com/2024/07/13/elon-musk-endorses-trump-after-shots-were-fired-at-former-president-.html) endorse a pres\u00adi\u00adden\u00adtial candi\u00addate . That the candi\u00addate would win. That on his (https://www.npr.org/2025/02/04/nx-s1-5286314/department-of-government-efficiency-doge-explainer-elon-musk) first day in office , the new pres\u00adi\u00addent would install the AI CEO in the govern\u00adment with broad powers, in pursuit of a nebu\u00adlous project of cutting waste and saving money. That ulti\u00admately, this project would (https://www.civilservicestrong.org/update/doge-what-was-the-point) save little money and primarily impair govern\u00adment agen\u00adcies designed to protect the vulner\u00adable. That somehow the parts of the govern\u00adment that provided large contracts for the same AI CEO (https://theintercept.com/2025/04/03/doge-pentagon-budget-spacex-government-contracts-musk/) would remain untouched . To be fair, for centuries US busi\u00adnesses have sought to bend US public policy to suit their inter\u00adests, often with great success. This is a theme of Galbraith\u2019s book (https://www.simonandschuster.com/books/The-Predator-State/James-K-Galbraith/9781416576211) The Predator State . What\u2019s new is the mous\u00adtache-twirling flagrancy of the project, and that it was under\u00adtaken at the behest of the\u00a0newly elected pres\u00adi\u00addent. (https://lucid.substack.com/p/doges-real-aims-have-nothing-to-do) Some called it a coup, though one premised on abdi\u00adca\u00adtion of power rather than seizure. \u2021 As of June 2026, (https://www.cnbc.com/2026/06/05/trump-open-ai-altman-stake.html) reporting has suggested that an AI CEO has proposed to the current US pres\u00adi\u00addent that the US take an equity stake in the AI CEO\u2019s company (and maybe others). This arrange\u00adment would be very different from taxa\u00adtion. It fore\u00adsees a transfer of capital from the US govern\u00adment to the corpo\u00adra\u00adtion to buy equity. As a share\u00adholder, the US govern\u00adment would be enti\u00adtled to a share of future profits, which may never exist. It also poten\u00adtially creates a conflict of interest: as a Big AI share\u00adholder, would the US govt regu\u00adlate or legis\u00adlate the AI industry? Taxa\u00adtion, by contrast, requires no prepay\u00adment by the US govern\u00adment, and can be assessed on the current activ\u00adi\u00adties of the AI company, regard\u00adless of prof\u00aditability. Histor\u00adi\u00adcally, the US govern\u00adment has (https://home.treasury.gov/news/press-releases/tg179) taken equity stakes in private corpo\u00adra\u00adtions as an indi\u00adrect bailout. Which is maybe the better word for what is being proposed here. Absent this exigency, taxa\u00adtion is econom\u00adi\u00adcally and polit\u00adi\u00adcally prefer\u00adable. This AI CEO may also want to ask his chatbot what happened to the oil industry in Venezuela under Ch\u00e1vez . Know your enemy Oh Karl, the world isn\u2019t fairIt isn\u2019t and never will beThey tried out your planIt brought misery insteadIf you\u2019d seen how they worked itYou\u2019d be glad you were dead.\u2014Randy Newman (https://www.youtube.com/watch?v=BaEcGHd1IX8) \u201cThe World Isn\u2019t Fair\u201d To iden\u00adtify as a Marxist today is rare (at least in the US). More often, it\u2019s wielded as an anti-capi\u00adtalist epithet against someone who endorses socialist or Commu\u00adnist methods. True\u2014Marx himself supported these methods. After all, he did write (https://www.marxists.org/admin/books/manifesto/Manifesto.pdf) The Commu\u00adnist Mani\u00adfesto . But short\u00adhand Marxism tends to omit the reasoning that Marx, a philoso\u00adpher and histo\u00adrian, consid\u00adered vital. After all, he did write (https://www.marxists.org/archive/marx/works/1867-c1/) Capital . Marx saw class struggle as the key insti\u00adgator of history. He argued that capi\u00adtalism was polit\u00adi\u00adcally unstable, and neces\u00adsarily led to worker revolt. Marx\u2019s theory that economic value orig\u00adi\u00adnates in labor over\u00adlaps concep\u00adtu\u00adally with Locke\u2019s (https://press-pubs.uchicago.edu/founders/documents/v1ch16s3.html) labor theory of prop\u00aderty . But to Marx, capi\u00adtalism sepa\u00adrated\u2014or using his preferred term, (https://www.marxists.org/subject/alienation/index.htm) alien\u00adated \u2014workers from the fruits of that labor. In prac\u00adtice, Marx\u2019s theo\u00adries have had mixed success. But Marx\u2019s obser\u00adva\u00adtions in Capital about the inter\u00adac\u00adtion of workers and machines remain rele\u00advant, even though he was writing in the mid-1800s, when the pinnacle of tech\u00adnology was textile machinery. Looking back on worker upris\u00adings since the 1600s\u2014including the (https://www.smithsonianmag.com/history/what-the-luddites-really-fought-against-264412/) Luddite move\u00adment that famously protested those textile machines\u2014(https://www.marxists.org/archive/marx/works/1867-c1/ch15.htm#S6) Marx wrote : It took both time and expe\u00adri\u00adence before the workpeople learnt to distin\u00adguish between machinery and its employ\u00adment by capital, and to direct their attacks, not against the mate\u00adrial instru\u00adments of produc\u00adtion, but against the mode in which they are used. By protesting the textile machines, the Luddites were missing the bigger picture. The Luddites (and earlier workers) were raging against the literal machine\u2014the specific new tech\u00adnology. But the tech\u00adnology was a symptom, not a cause. Instead, workers needed to rage against the bigger, figu\u00adra\u00adtive machine\u2014the extrac\u00adtive capi\u00adtalist system. Marx\u2019s obser\u00adva\u00adtion has a subtler impli\u00adca\u00adtion too. New tech\u00adnology often holds itself out as the starting point of a narra\u00adtive: from now on, every\u00adthing is different . When we consider the tech\u00adnology alone, that narra\u00adtive domi\u00adnates. But when we zoom out and consider the histor\u00adical context, the new tech\u00adnology becomes the current endpoint of a much longer polit\u00adical narra\u00adtive. What would Marx say to AI critics\u2014social, legal, economic, polit\u00adical\u2014that have arisen so far? Maybe that we\u2019re missing the bigger picture. That as a human inven\u00adtion, AI may be the starting point of a new tech\u00adno\u00adlog\u00adical narra\u00adtive. But as an affront to human workers, it continues a long tradi\u00adtion of capi\u00adtalist tech\u00adnolo\u00adgies, begin\u00adning with the Indus\u00adtrial Revo\u00adlu\u00adtion (if not earlier). For their part, the capi\u00adtal\u00adists who own a new tech\u00adnology usually prefer to frame it as a starting point. Conve\u00adniently, this omits the grubby histor\u00adical context that might dampen the marketing. Similar sleight of hand animates the snide techie\u2019s favorite sick burn\u2014Luddite \u2014connoting a block\u00adhead (https://pmarca.substack.com/p/the-techno-optimist-manifesto#:~:text=continuous%20howling%20from%20Communists%20and%20Luddites) irra\u00adtionally opposed to tech\u00adnology. Of course, this usage is (https://www.nationalgeographic.com/history/article/luddite-industrial-revolution-anti-technology) delib\u00ader\u00adately ahis\u00adtor\u00adical \u2014read Brian Merchant\u2019s (https://www.hachettebookgroup.com/titles/brian-merchant/blood-in-the-machine/9780316487740/) Blood in the Machine . Perhaps there should be a companion term for a block\u00adhead (https://futurism.com/artificial-intelligence/marc-andreessen-sputters-ai-benefits) irra\u00adtionally committed to AI. These two frames for under\u00adstanding new tech\u00adnology under capi\u00adtalism also suggest two corre\u00adsponding analytic approaches to AI risk: AI as tech\u00adnology. On this view, AI is a starting point, and we consider the risks of what it can do that previous tech\u00adnolo\u00adgies could not. This isn\u2019t wrong, exactly. It works for certain char\u00adac\u00adter\u00adis\u00adtics. For instance: I can believe that AI will (https://www.axios.com/2026/04/07/anthropic-mythos-preview-cybersecurity-risks) facil\u00adi\u00adtate hacking and similar crim\u00adi\u00adnality like no tech\u00adnology before. That risk is specific to AI as new tech\u00adnology and deserves consid\u00ader\u00ada\u00adtion. AI as capi\u00adtalist instru\u00adment. But for other char\u00adac\u00adter\u00adis\u00adtics\u2014say, AI\u2019s effects on labor, public wealth, and the economy\u2014thinking of AI strictly as new tech\u00adnology embeds the error Marx warns against. For those char\u00adac\u00adter\u00adis\u00adtics, we should consider how AI may amplify or accel\u00aderate existing trends. Put differ\u00adently\u2014the ways in which AI is inher\u00adently polit\u00adical tech\u00adnology . In sum\u2014as new tech\u00adnology, AI acts as a creator of new risks; as a capi\u00adtalist instru\u00adment, it acts as an ampli\u00adfier or accel\u00ader\u00adator of existing risks and trends in the capi\u00adtalist system. The strongest version of this framing is that the \u201campli\u00adfier or accel\u00ader\u00adator\u201d part is precisely Why AI Exists. Unprece\u00addented billions have (https://www.washingtonpost.com/technology/2026/02/07/ai-spending-economy-shortages/) already been poured into AI. Why? Because it\u2019s expected to deliver more profit and concen\u00adtrate more wealth than any previous tech\u00adnology. But ampli\u00adfying these trends also means ampli\u00adfying the atten\u00addant polit\u00adical risks. Earlier I accepted the argu\u00adment that Big AI will deliver abun\u00addant goodies to citi\u00adzens\u2014which AI propo\u00adnents consider one of the strongest points in its favor. But doing so will likely increase align\u00adment between Big AI and citi\u00adzens, with nega\u00adtive polit\u00adical conse\u00adquences. As usual\u2014no free lunch. This shift in framing also shifts the eviden\u00adtiary context. When we think about AI risk, we\u2019re neces\u00adsarily making guesses about the future. But when we frame AI in the narrow sense of new tech\u00adnology, we\u2019re primarily consid\u00adering a time\u00adline that starts now. Whereas when we shift to thinking of AI as a capi\u00adtalist instru\u00adment, we\u2019re consid\u00adering a time\u00adline that starts centuries ago and has evolved contin\u00adu\u00adously into the present. We can and should study those existing economic and polit\u00adical trends, because those will likely shape the future trajec\u00adtory. Put differ\u00adently: AI may be new. But it\u2019s not immune to history. The poisoned chalice I think the biggest chal\u00adlenge to AI in this country is polit\u00adical unrest \u2026\u00a0Can you make more money? It\u2019s all irrel\u00ade\u00advant if the country blows up \u2026 If I were sitting here in private with my peers, I\u2019d be telling them \u2026 the country could blow up polit\u00adi\u00adcally. And none of us are going to make any money when the country blows up.\u2014(https://youtu.be/b-JCYM5jW0Y?t=2504) an AI-adja\u00adcent CEO Big AI\u2019s goal of labor replace\u00adment has two dimen\u00adsions some\u00adtimes over\u00adlooked, although they are both hidden in plain sight. The first relates to AI\u2019s effort to replace knowl\u00adedge workers. The other pertains to the compet\u00adi\u00adtive effects of doing so. AI replacing knowl\u00adedge workers. Labor replace\u00adment is the mech\u00ada\u00adnism. Struc\u00adturally, Big AI seeks to build a moat around US economic growth by normal\u00adizing AI as a (https://snscratchpad.com/posts/looking-ahead-2026/#:~:text=these%20new%20cognitive%20amplifier%20tools) cogni\u00adtive inter\u00adme\u00addiary where knowl\u00adedge work\u2014roughly, econom\u00adi\u00adcally valu\u00adable thinking and creativity\u2014will happen. This maneuver is part of the automa\u00adtion play\u00adbook, espe\u00adcially for the tech industry. During the rise of desktop automa\u00adtion in the 1980s and 1990s, tech compa\u00adnies sought to build moats by control\u00adling the file formats, programs, and plat\u00adforms where docu\u00adments were created. As the commer\u00adcial internet ripened during the 2000s, tech compa\u00adnies sought moats based on lever\u00adaging network effects\u2014espe\u00adcially the unholy dyad of adver\u00adtising and social media. How Big AI plans to profit from this inter\u00adme\u00addi\u00ada\u00adtion is an open ques\u00adtion. One AI company has suggested (https://www.theinformation.com/newsletters/applied-ai/openai-plans-take-cut-customers-ai-aided-discoveries) taking a cut of AI-assisted discov\u00aderies. The logis\u00adtics and legal\u00adi\u00adties would be boggling. Details\u2014what\u00adever. For now, AI compa\u00adnies largely agree on the first step: make workers depen\u00addent on AI to do their jobs, just as tech fore\u00adbears made workers depen\u00addent on a certain soft\u00adware program to share a file, or on a certain website to have friends. This time, however, the soft\u00adware ulti\u00admately consumes the worker. Big AI\u2019s timing is canny. In recent decades, intan\u00adgible assets have played an increasing role in global economic growth. An intan\u00adgible asset lacks phys\u00adical embod\u00adi\u00adment, and includes\u00a0formal intel\u00adlec\u00adtual-prop\u00aderty assets (e.g., copy\u00adrights, patents) but also informal (e.g., brand good\u00adwill, other knowl\u00adedge-work assets). Tangible assets are the oppo\u00adsite\u2014bricks, mortar, inven\u00adtory, and so forth. According to the WIPO, since 1995 global invest\u00adment in intan\u00adgible assets has steadily increased, and since 2009 (https://www.wipo.int/web-publications/world-intangible-investment-highlights-2025/en/world-intangible-investment-highlights-2025.html) has exceeded that of tangible assets. By wide margin, the US leads invest\u00adment in intan\u00adgible assets. For a long time, intan\u00adgible assets were under\u00advalued because tradi\u00adtional GDP measures and accounting prac\u00adtices (https://www.nber.org/digest/sep06/intangible-capital-and-economic-growth?page=1&perPage=50) excluded them . In the 1970s and 1980s, during the first big wave of IT invest\u00adment, the US expe\u00adri\u00adenced a coun\u00adter\u00adin\u00adtu\u00aditive slow\u00addown in produc\u00adtivity growth, an effect econ\u00ado\u00admist (https://www.nber.org/system/files/working_papers/w24001/w24001.pdf) Erik Bryn\u00adjolf\u00adsson called the \u201cproduc\u00adtivity paradox\u201d. Part of the reso\u00adlu\u00adtion to the produc\u00adtivity paradox is that returns from IT invest\u00adment were mani\u00adfesting not as labor produc\u00adtivity per se, but as intan\u00adgible assets. For busi\u00adnesses, IT invest\u00adment was worth\u00adwhile even if it didn\u2019t increase labor produc\u00adtivity, because it still gener\u00adated intan\u00adgible assets. (Another part of the reso\u00adlu\u00adtion to the produc\u00adtivity paradox is Bryn\u00adjolf\u00adsson\u2019s theory of (https://www.nber.org/system/files/working_papers/w25148/w25148.pdf) lagging comple\u00admen\u00adtary invest\u00adment , which will also likely apply to AI.) Never\u00adthe\u00adless, increasing labor produc\u00adtivity matters a lot to workers. Why? It has histor\u00adi\u00adcally been the biggest driver of wage growth. Which brings us to another\u2014maybe the most\u2014conse\u00adquen\u00adtial trend in the US economy over the last 50 years: wage stag\u00adna\u00adtion , or the decou\u00adpling of produc\u00adtivity and wage growth. As the story is typi\u00adcally told: from 1948 to 1973, wage growth (91.3%) basi\u00adcally matched produc\u00adtivity growth (96.7%). Since 1973, however, overall (https://www.epi.org/publication/charting-wage-stagnation/) produc\u00adtivity has grown (74.4%) much faster than wages (9.2%). Where has the money gone instead? Into the pockets of busi\u00adness owners, mostly. Over the same time\u00adframe, the oper\u00adating profits of busi\u00adnesses (as measured by net oper\u00adating surplus) has (https://fred.stlouisfed.org/series/W271RE1A156NBEA) steadily increased . Today, if wage growth had kept up with produc\u00adtivity gains since 1973, we\u2019d expect wages to be more than double. Together, these trends have been a long-term double whammy for US workers. Because of wage stag\u00adna\u00adtion, workers have been gener\u00adating increasing value for employers (as produc\u00adtivity) but not capturing via wages the same share of that value they once did. Because of the shift toward intan\u00adgible assets, employers have invested less in labor produc\u00adtivity, further inhibiting wage growth. As AI replaces knowl\u00adedge workers, we can expect it to amplify both trends. Will it? Recent labor-market research by one major AI company suggests that the greatest \u201ctheo\u00adret\u00adical capa\u00adbility\u201d for AI labor replace\u00adment lies with (https://www.anthropic.com/research/labor-market-impacts) tradi\u00adtional cate\u00adgories of knowl\u00adedge work: finance, program\u00adming, engi\u00adneering, science, legal, and the arts. A research group at (https://digitalplanet.tufts.edu/ai-and-the-emerging-geography-of-american-job-risk-page/) Tufts Univer\u00adsity was blunter : \u201cthe more AI helps you do your job, the more expend\u00adable you can become. Finance profes\u00adsionals, teachers/profes\u00adsors, creative profes\u00adsionals, accoun\u00adtants and audi\u00adtors, legal profes\u00adsionals\u201d are at risk. Even (https://www.nytimes.com/2023/08/24/upshot/artificial-intelligence-jobs.html#:~:text=It%E2%80%99s%20surprised%20most%20people%2C%20including%20me) Bryn\u00adjolf\u00adsson said he was \u201csurprised\u201d that knowl\u00adedge work would be so readily replaced. \u201cTech\u00adnology always makes certain jobs obso\u00adlete; new ones will arise.\u201d AI\u2019s predicted labor replace\u00adment is unprece\u00addented in three ways: the diver\u00adsity of tasks replaced; its outsize effect on highly educated workers; and the back\u00addrop of 50 years of wage stag\u00adna\u00adtion. Automa\u00adtion-driven tran\u00adsi\u00adtions aren\u2019t neces\u00adsarily easy, even when they\u2019re narrow and the economy can absorb the workers. Those who hand\u00adwave over the details should study (https://www.pbs.org/wgbh/americanexperience/features/theminewars-labor-wars-us/) histor\u00adical exam\u00adples . When you tell a large group of workers that their skills no longer have economic value, you risk a (https://www.smithsonianmag.com/history/what-the-luddites-really-fought-against-264412/) polit\u00adical\u00a0and social tinderbox . Recall (https://www.nytimes.com/2026/04/30/opinion/ai-labor-work-force-silicon-valley.html#:~:text=the%20short%20run%20can%20be%20a%20lifetime) Carl Benedikt Frey\u2019s comment : \u201cthe short run can be a life\u00adtime\u201d. Confi\u00adden\u00adtial to owners of capital : nothing I\u2019m saying above depends on a norma\u00adtive view about wealth distri\u00adbu\u00adtion. I person\u00adally oppose wage stag\u00adna\u00adtion and believe workers\u00a0should enjoy wage growth propor\u00adtional (at minimum) to produc\u00adtivity\u00a0growth. If you don\u2019t\u2014fine. The argu\u00adment persists. Even accepting that the AI labor tran\u00adsi\u00adtion is econom\u00adi\u00adcally rational and neces\u00adsary, it will still\u00a0likely have polit\u00adi\u00adcally desta\u00adbi\u00adlizing effects that you, owners of capital, will find unpleasant. Plus, you\u2019ll get punched in the wallet too\u2014keep reading. The compet\u00adi\u00adtive effects of replacing knowl\u00adedge workers . The previous section consid\u00adered how AI labor replace\u00adment will affect the rela\u00adtion\u00adship between workers and employers. But it will also induce more carniv\u00ado\u00adrous compe\u00adti\u00adtion among firms that adopt AI, and thus indi\u00adrectly among the owners of capital. Whereas the last 50 years of the US economy have featured wage stag\u00adna\u00adtion, the post-AI economy may feature some\u00adthing akin to capital stag\u00adna\u00adtion, for non-AI cate\u00adgories of capital. A growing economy used to be a rising tide that lifted all boats. Post-AI, it may only lift a handful of yachts. Everyone else will be in dry dock. Arguably that capital tran\u00adsi\u00adtion is (https://www.investopedia.com/the-mag-7-in-charts-how-big-tech-dominates-the-market-11866473) already underway . How might this work? The economy is increas\u00adingly driven by intan\u00adgible assets. Knowl\u00adedge workers provide those assets. If a company lays off its knowl\u00adedge workers in favor of adopting some vendor\u2019s AI, at first it gets the same produc\u00adtivity at much lower cost. So much win. But in doing so, the company commodi\u00adtizes its own output. If your company can auto\u00admate its output via AI, others can too. Your existing competi\u00adtors, certainly. But also upstarts who don\u2019t have your cost foot\u00adprint. What\u00adever intan\u00adgible assets AI can generate will be produced\u00a0in excess, leading to a defla\u00adtionary market for that asset. A company\u2019s knowl\u00adedge workers may be its greatest expense. But they also contribute to its compet\u00adi\u00adtive moat. Further\u00admore, the ongoing market value of a company is neces\u00adsarily tied to its differ\u00aden\u00adti\u00adated assets. A company valued at a billion dollars that adopts AI will then have to build some new asset not replic\u00adable with AI, or it won\u2019t be worth a billion much longer. Every\u00adthing that makes AI an excel\u00adlent cost collapser makes it an equally excel\u00adlent capital collapser.\u00a0A poisoned chalice. The greatest irony will be if AI-adopting compa\u00adnies\u2014having laid off knowl\u00adedge workers en masse\u2014hire them back merely to ensure that competi\u00adtors cannot. In recent tech history, there have (https://matthewbutterick.com/chron/the-broadway-paradox.html#:~:text=to%20deploy%20some%20of%20their%20economic%20power%20toward%20locking%20up%20highly%20skilled%20labor%20as%20a%20form%20of%20market%20compe%C2%ADti%C2%ADtion) already been signs of this dynamic. Consider large law firms, aka (https://law.usnews.com/law-firms/advice/articles/what-is-big-law) Big Law . Currently certain (https://www.businessinsider.com/legora-founder-harvey-legal-tech-ai-platform-competition-2026-3) legal-AI star\u00adtups license LLMs from Big AI and repackage them for Big Law at high prices. These star\u00adtups claim to add other special sauce. OK, sure. Where\u2019s the economic equi\u00adlib\u00adrium? If legal-AI star\u00adtups prove that money can be made selling AI to Big Law\u2014won\u2019t Big AI just sell to Big Law directly, and cut out the star\u00adtups? Or if legal-AI star\u00adtups prove that AI can effec\u00adtively provide legal services\u2014won\u2019t legal-AI star\u00adtups just sell to clients directly, and cut out Big Law? Won\u2019t members of Big Law that adopt AI have to (https://matthewbutterick.com/chron/layers-of-lawyers-and-liars.html#:~:text=embrace%20AI%2C%20and%20cut%20deeper%20into%20the%20partner%20ranks) lay off a lot of equity part\u00adners , because adop\u00adtion of AI will shrink profit margins? Won\u2019t the members of Big Law refusing AI have to\u00a0consol\u00adi\u00addate to preserve their margins? Or just cave to AI? (The tendency of competi\u00adtors to adopt similar prac\u00adtices is called (https://www.nytimes.com/2025/12/16/opinion/artists-creative-work-ai.html#:~:text=institutional%20isomorphism) insti\u00adtu\u00adtional isomor\u00adphism .) So it goes. Most states prevent nonlawyers from sharing in legal fees, so law firms will prob\u00adably remain a distinct set of enti\u00adties. But one plau\u00adsible equi\u00adlib\u00adrium is that legal-AI star\u00adtups disap\u00adpear (quickly), and members of Big Law consol\u00adi\u00addate (rela\u00adtively quickly) until there are only a handful left, all contracting directly with Big AI. Along these lines, I expect that to succeed finan\u00adcially, Big AI will likely need to demolish a signif\u00adi\u00adcant number of existing tech compa\u00adnies and (https://www.theinformation.com/articles/anthropic-blindsides-business-partners) grab their revenue for itself. By the process described above: Big AI essen\u00adtially uses its tech customers as an R&D facility. Big AI licenses models to these compa\u00adnies. Tech compa\u00adnies compete to adapt their busi\u00adnesses to AI. Once a concept is proven, Big AI directly takes over that market. The labor-replace\u00adment story will grow into a company-replace\u00adment story. Many of those tech compa\u00adnies\u2014and their share\u00adholders in the public markets\u2014may also find that AI is a poisoned chalice. The resource curse The theory is thought also to fit Middle Eastern oil exporters espe\u00adcially well. In this region, govern\u00adments\u2019 access to rents, in the form of oil revenue, may have freed them from the need for taxa\u00adtion of their peoples, and that this in turn freed them from the need for democ\u00adracy.\u2014(https://www.hks.harvard.edu/sites/default/files/centers/cid/files/publications/faculty-working-papers/233.pdf) Jeffrey Frankel Since AI is new, there are not yet exam\u00adples of post-AI economies. So we might ask a broader ques\u00adtion: what happens to nations whose economies are concen\u00adtrated around one inher\u00adently polit\u00adical tech\u00adnology? The most instruc\u00adtive compa\u00adra\u00adbles in the world economy are prob\u00adably petrostates : nations whose national wealth is concen\u00adtrated in oil or natural gas. The analogy is imper\u00adfect. Unlike AI, oil is anchored to geog\u00adraphy and limited in supply. On the other hand, oil has proven finan\u00adcial value; AI does not. Petrostates are vulner\u00adable to what econ\u00ado\u00admists call the resource curse , with two distinc\u00adtive sets of effects: Capital effects. The concen\u00adtrated resource tends to (https://www.hks.harvard.edu/sites/default/files/centers/cid/files/publications/faculty-working-papers/233.pdf) pull labor and capital from the rest of the economy, depriving other economic sectors, espe\u00adcially manu\u00adfac\u00adturing and agri\u00adcul\u00adture. This makes the national economy more depen\u00addent on imports for basic needs, and more sensi\u00adtive to price volatility of the resource. If the resource price declines suffi\u00adciently, the nation will be unable to keep affording imports, but also unable to fulfill those needs with domestic produc\u00adtion. Polit\u00adical effects. The value of the concen\u00adtrated resource creates what (https://www.hks.harvard.edu/sites/default/files/centers/cid/files/publications/faculty-working-papers/233.pdf#page=10) Jeffrey Frankel calls \u201ca polit\u00adical contest to capture owner\u00adship\u201d, which in turn encour\u00adages the emer\u00adgence of auto\u00adcratic or oligarchic insti\u00adtu\u00adtions captured by an economic elite who seek to retain control of the resource. The process is self-rein\u00adforcing in two ways. First: the economic elite uses its wealth to repress polit\u00adical oppo\u00adnents. Second: as the govern\u00adment derives more income from the concen\u00adtrated resource, it relies less on taxa\u00adtion of citi\u00adzens, which weakens demo\u00adc\u00adratic account\u00adability. According to the (https://www.cfr.org/backgrounders/venezuela-crisis) Council on Foreign Rela\u00adtions , \u201cpetrostates include Algeria, Cameroon, Chad, Ecuador, Indonesia, Iran, Kaza\u00adkhstan, Libya, Mexico, Nigeria, Oman, Qatar, Russia, Saudi Arabia, the United Arab Emirates, and Venezuela.\u201d No one would confuse this with a list of successful liberal democ\u00adra\u00adcies. Due to its increasing oil and gas produc\u00adtion, some also consider the (https://www.economist.com/finance-and-economics/2026/03/19/america-may-be-a-petrostate-but-the-energy-shock-still-hurts) US a petrostate . Which petrostates provide the most goodies to citi\u00adzens? The Gulf petrostates\u2014Qatar , UAE , Saudi Arabia , Oman . These nations distribute petrowealth to citi\u00adzens through substan\u00adtial public bene\u00adfits (e.g., housing, health\u00adcare, educa\u00adtion, public-sector jobs) while having low rates of taxa\u00adtion. The tradeoff, of course, is that these coun\u00adtries are not even notion\u00adally demo\u00adc\u00adratic. They are hered\u00adi\u00adtary monar\u00adchies. They offer none of the bene\u00adfits that citi\u00adzens in a liberal democ\u00adracy enjoy (or some\u00adtimes take for granted)\u2014demo\u00adc\u00adratic partic\u00adi\u00adpa\u00adtion, indi\u00advidual civil rights, freedom of speech. But the Gulf states were never liberal democ\u00adra\u00adcies. Can the resource curse send a liberal democ\u00adracy spiraling into autoc\u00adracy? (https://foreignpolicy.com/2018/07/16/how-venezuela-struck-it-poor-oil-energy-chavez/) Sure \u2014consider Venezuela . Today, Venezuela is an author\u00adi\u00adtarian regime. But from 1958 to the 1990s, it was a stable liberal democ\u00adracy. This changed with the 1998 elec\u00adtion of Hugo Ch\u00e1vez. After taking office, Ch\u00e1vez (https://www.cfr.org/backgrounders/venezuela-crisis) pursued populist economic poli\u00adcies funded by national oil wealth. Embold\u00adened by high oil prices, Ch\u00e1vez amended the consti\u00adtu\u00adtion to bolster his own power and installed loyal\u00adists at the top of Venezuela\u2019s national oil company (PDVSA). He redi\u00adrected oil revenues to a series of domestic social programs (the \u201cBoli\u00advarian missions\u201d). Ch\u00e1vez lever\u00adaged his resulting popu\u00adlarity among workers to consol\u00adi\u00addate his power, for instance by (https://www.nytimes.com/2009/02/16/world/americas/16venez.html) ending term limits and taking control of the judi\u00adciary. (Confi\u00adden\u00adtial to AI CEOs: think care\u00adfully before you offer an owner\u00adship stake to any US pres\u00adi\u00addent who has author\u00adi\u00adtarian and populist tenden\u00adcies.) After oil prices dropped in 2014, the resource curse hit Venezuela\u00a0hard, putting its economy into crisis and allowing Ch\u00e1vez\u2019s successor Nicol\u00e1s Maduro to estab\u00adlish a dicta\u00adtor\u00adship. Demon\u00adstrating a central irony of the resource curse: while the resource is gener\u00adating wealth and deliv\u00adering goodies to citi\u00adzens, author\u00adi\u00adtarian moves are depicted as essen\u00adtial to sustaining pros\u00adperity. After the economy weakens, author\u00adi\u00adtarian moves are depicted as essen\u00adtial to restoring pros\u00adperity. The ratchet only tightens. The river flows only down\u00adhill. Norway is another outlier. Norway is a robust liberal democ\u00adracy. After oil was discov\u00adered in its North Sea terri\u00adto\u00adrial waters in the 1960s, Norway created the state-owned (https://cases.som.yale.edu/equinor/norway-and-oil/statoil) Statoil to build and operate oil facil\u00adi\u00adties. In 1990, Norway estab\u00adlished the (https://www.nbim.no/en/about-us/about-the-fund/) Govern\u00adment Pension Fund Global , a sover\u00adeign-wealth fund to receive surplus oil revenues and invest them for the benefit of Norwe\u00adgian citi\u00adzens. According to Norges Bank, \u201c[t]he aim of the fund is to ensure that we use this money respon\u00adsibly, think long-term and so safe\u00adguard the future of the Norwe\u00adgian economy.\u201d In one sense it has worked: the GPFG currently holds over $2 tril\u00adlion in assets. What\u2019s the problem? Even Norway hasn\u2019t entirely avoided the resource curse. Rising sover\u00adeign wealth has deliv\u00adered tons of goodies to Norwe\u00adgians. But it has also made Norway\u2019s govern\u00adment less finan\u00adcially disci\u00adplined and (https://www.economist.com/finance-and-economics/2026/04/01/can-a-country-get-too-rich) created social conse\u00adquences like high dropout and unem\u00adploy\u00adment rates. (A popular 2025 Norwe\u00adgian book was called (https://booksfromnorway.com/books/2802-the-country-that-got-too-rich) The Country That Got Too Rich .) Keep in mind that Norway reached this predica\u00adment despite 50 years of careful prepa\u00adra\u00adtion and over\u00adsight. Inher\u00adently polit\u00adical tech\u00adnolo\u00adgies can upend even gener\u00ada\u00adtional prepa\u00adra\u00adtions. Those who would suggest that the US can follow a similar model with AI: easier said than done. Extinction-level capitalism Putting it all together: Among AI risks, we should take more seri\u00adously the poten\u00adtial conse\u00adquences of AI working as intended. AI is a capi\u00adtalist instru\u00adment. Its prin\u00adcipal func\u00adtion is to concen\u00adtrate capital. Its intended mech\u00ada\u00adnism is large-scale labor replace\u00adment. But it is also inher\u00adently polit\u00adical tech\u00adnology. As AI makes it harder for workers to capture value from their labor, they will increas\u00adingly have to rely on goodies from Big AI, priva\u00adtizing what were once func\u00adtions of govern\u00adment. If Big AI subsumes the func\u00adtions of workers and govern\u00adment, both will tend to realign polit\u00adi\u00adcally around Big AI\u2019s inter\u00adests. What\u00adever term describes this system, it is not liberal democ\u00adracy as US citi\u00adzens have tradi\u00adtion\u00adally under\u00adstood it. AI-centered capi\u00adtalism risks an extinc\u00adtion of demo\u00adc\u00adratic possi\u00adbility. It will be America. But it will no longer be Amer\u00adican. Epilogue Thank you for reading. Every citizen has a voice in these issues. Every citizen can partic\u00adi\u00adpate in the vital public conver\u00adsa\u00adtion about how we want AI to be part of our country\u2014our schools, our work, our fami\u00adlies. (https://archive.org/details/computerpowerhum0000weiz_v0i3/page/n9/mode/2up?q=%22powerful+tranquilizer%22) Joseph Weizen\u00adbaum : \u201cThe myth of tech\u00adno\u00adlog\u00adical and polit\u00adical and social inevitability is a powerful tran\u00adquil\u00adizer of the conscience.\u201d In the upcoming elec\u00adtion, vote for people who take your rights and inter\u00adests seri\u00adously. I support orga\u00adnized labor. Ulti\u00admately, Big AI is constrained by an incon\u00adve\u00adnient truth. Today, they need us more than we need them. As long as that remains true, we the people have the upper hand. " ], "output": "htmltotext.txt", "pwd": "/data/archive/1781518242.322736", "schema": "ArchiveResult", "start_ts": "2026-06-15T10:11:37.414609+00:00", "status": "succeeded" } ], "media": [ { "cmd": [ "/usr/local/bin/yt-dlp", "--restrict-filenames", "--trim-filenames", "128", "--write-description", "--write-info-json", "--write-annotations", "--write-thumbnail", "--no-call-home", "--write-sub", "--write-auto-subs", "--convert-subs=srt", "--yes-playlist", "--continue", "--no-abort-on-error", "--ignore-errors", "--geo-bypass", "--add-metadata", "--format=(bv*+ba/b)[filesize<=750m][filesize_approx<=?750m]/(bv*+ba/b)", "--skip-download", "--cache-dir=/data/yt-dlp-cache/", "--cookies=/data/yt-dlp-cache/cookies.txt", "--proxy=socks5://tor-socks-proxy:9150", "--no-playlist", "https://matthewbutterick.com/extinction-level-capitalism.html" ], "cmd_version": "2024.10.7", "end_ts": "2026-06-15T10:11:45.683903+00:00", "index_texts": [], "output": "media/", "pwd": "/data/archive/1781518242.322736", "schema": "ArchiveResult", "start_ts": "2026-06-15T10:11:39.730021+00:00", "status": "succeeded" } ], "mercury": [ { "cmd": [ "/home/archivebox/.npm/bin/postlight-parser", "https://matthewbutterick.com/extinction-level-capitalism.html" ], "cmd_version": "2.2.3", "end_ts": "2026-06-15T10:11:37.367793+00:00", "index_texts": null, "output": "mercury/", "pwd": "/data/archive/1781518242.322736", "schema": "ArchiveResult", "start_ts": "2026-06-15T10:11:33.489726+00:00", "status": "succeeded" } ], "pdf": [], "readability": [ { "cmd": [ "/home/archivebox/.npm/bin/readability-extractor", "/tmp/tmpud1m0pk3", "https://matthewbutterick.com/extinction-level-capitalism.html" ], "cmd_version": "0.0.11", "end_ts": "2026-06-15T10:11:23.513473+00:00", "index_texts": [ "AI is inher\u00adently polit\u00adical tech\u00adnology. If AI works as intended, it will grad\u00adu\u00adally corrode our liberal democ\u00adracy, risking an irre\u00adversible shift into another polit\u00adical and economic config\u00adu\u00adra\u00adtion. Among AI risks, this one deserves\u00a0more consid\u00ader\u00ada\u00adtion, because it requires no addi\u00adtional condi\u00adtions like malign actors or AI malfunc\u00adtion. AI only needs to amplify existing trends, espe\u00adcially around concen\u00adtra\u00adtion of capital. This damage will occur even assuming that in the near term, AI will broadly improve mate\u00adrial well-being.About MBI\u2019m a self-employed author, designer, programmer, and lawyer. In 2022, I learned that my own works were in the training datasets of gener\u00ada\u00adtive-AI compa\u00adnies. In response, I invented the first set of lawsuits chal\u00adlenging the legality of these prac\u00adtices. I\u2019m currently co-counsel for plain\u00adtiffs in a number of AI cases. Though I discuss certain legal issues below, I am not your lawyer, and nothing here is held out as legal advice. These are my personal views as a citizen and economic actor; I speak only for myself. This piece is typeset in Equity, Advo\u00adcate, and Trip\u00adli\u00adcate, fonts I designed. They can be licensed for your own polemics and pamphlets.Emergent effectsTwo billion years ago, the rock layers comprising what is now called the Colorado Plateau began to form: first igneous and meta\u00admor\u00adphic rocks, followed by many layers of sedi\u00admen\u00adtary rocks. About fifty million years ago, through tectonic action, this plateau gained thou\u00adsands of feet of eleva\u00adtion. About five million years ago, a river began to flow. The river carried silt and debris, scraping out the begin\u00adnings of a canyon. The river deep\u00adened the canyon, exposing its walls to weather and erosional forces that widened the canyon further. Today the waterway is the Colorado River. The geolog\u00adical forma\u00adtion is the Grand Canyon.The forma\u00adtion of the Grand Canyon required zero human agency. Zero tech\u00adnology. Zero coor\u00addi\u00adna\u00adtion among\u00a0the river, the land, and gravity. In that sense the Grand Canyon is an emer\u00adgent effect: a complex, unfore\u00adsee\u00adable output arising from simpler inputs.But we would never wonder whether the river is sentient. Or whether the river cares about the dirt that it carries out of the canyon. The water is just doing what water does: flowing down\u00adhill. The dirt just happens to be in the\u00a0way.Inherently political technologyLangdon Winner is a polit\u00adical theo\u00adrist. Winner wrote the excel\u00adlent and influ\u00aden\u00adtial essay \u201cDo Arti\u00adfacts Have Poli\u00adtics?\u201d (1980). Winner sought to debunk the tradi\u00adtional framing that \u201ctech\u00adnolo\u00adgies are \u2026\u00a0neutral tools that can be used well or poorly, for good, evil, or some\u00adthing in between.\u201d Instead, Winner proposes two ways that a tech\u00adnology can affect its polit\u00adical envi\u00adron\u00adment:The tech\u00adnology is designed to have certain polit\u00adical effects. For example, the Great Fire\u00adwall of China, a bundle of tech\u00adno\u00adlog\u00adical measures that limit Chinese citi\u00adzens\u2019 access to foreign infor\u00adma\u00adtion sources. Antipo\u00addally, the Tor Project intends to maxi\u00admize user anonymity and thwart govern\u00adment intru\u00adsion.The tech\u00adnology is inher\u00adently polit\u00adical. This is Winner\u2019s key analytic fulcrum. Winner describes two versions of inher\u00adently polit\u00adical tech\u00adnology. The first is where the tech\u00adnology \u201cactu\u00adally requires \u2026\u00a0a partic\u00adular set of social condi\u00adtions as [its] oper\u00adating envi\u00adron\u00adment.\u201d For instance, nuclear weapons: the only respon\u00adsible way to possess such dangerous tech\u00adnology is to place it within \u201ca central\u00adized, rigidly hier\u00adar\u00adchical chain of command \u2026\u00a0the [atom] bomb must be author\u00adi\u00adtarian; there is no other way.\u201d The second version is where the tech\u00adnology is \u201cstrongly compat\u00adible\u201d with a certain polit\u00adical arrange\u00adment (even if not strictly required) and thus tends to bring that arrange\u00adment to fruition.As an example, Winner considers the mechan\u00adical tomato harvester. Devel\u00adoped at UC Davis in the 1950s, the machine was tremen\u00addously produc\u00adtive. But it was also expen\u00adsive. Only well-capi\u00adtal\u00adized tomato growers could afford it. The rest couldn\u2019t compete. According to Winner, the number of Cali\u00adfornia tomato growers dropped from ~4000 in the early 1960s to ~600 in 1973, costing ~32,000 jobs and the compounding nega\u00adtive effects on those commu\u00adni\u00adties. Winner summa\u00adrizes:What we see here \u2026\u00a0is an ongoing social process in which scien\u00adtific knowl\u00adedge, tech\u00adno\u00adlog\u00adical inven\u00adtion, and corpo\u00adrate profit rein\u00adforce each other in deeply entrenched patterns that bear the unmis\u00adtak\u00adable stamp of polit\u00adical and economic power \u2026\u00a0oppo\u00adnents of inno\u00adva\u00adtions like the tomato harvester are made to seem \u201cantitech\u00adnology\u201d or \u201cantiprogress\u201d. For the harvester is not merely the symbol of a social order that rewards some while punishing others; it is in a true sense an embod\u00adi\u00adment of that order.Not merely the symbol\u2014the embod\u00adi\u00adment. A facially neutral tech\u00adno\u00adlog\u00adical inven\u00adtion\u2014here, a tomato harvester\u2014can induce polit\u00adical effects. Those effects don\u2019t arise from flaws in the tech\u00adnology. To the contrary\u2014they arise from its effi\u00adcacy.How are the polit\u00adical effects deter\u00admined? Winner iden\u00adti\u00adfies two key early deci\u00adsions. The first is the binary ques\u00adtion of whether to pursue the tech\u00adnology at all. The second are choices about \u201cthe design or arrange\u00adment\u201d of the tech\u00adnology. Winner cautions: \u201c[t]o see the matter solely in terms of cost-cutting, effi\u00adciency, or the modern\u00adiza\u00adtion of equip\u00adment is to miss a deci\u00adsive element\u201d. That is, the polit\u00adical effects can possibly be coun\u00adtered, but first they must be acknowl\u00adedged.Of course, the best oppor\u00adtu\u00adnity to choose wisely is before the tech\u00adnology is widely intro\u00adduced, as capital and social invest\u00adment tends to entrench\u00a0it:Because choices tend to become strongly fixed in mate\u00adrial equip\u00adment, economic invest\u00adment, and social habit, the orig\u00adinal flex\u00adi\u00adbility vanishes for all prac\u00adtical purposes once the initial commit\u00adments are made. In that sense tech\u00adno\u00adlog\u00adical inno\u00adva\u00adtions are similar to legisla\u00adtive acts or polit\u00adical found\u00adings that estab\u00adlish a frame\u00adwork for public order that will endure over many gener\u00ada\u00adtions. \u2026\u00a0The issues that divide or unite people in society are settled not only in the insti\u00adtu\u00adtions and prac\u00adtices of poli\u00adtics proper, but also, and less obvi\u00adously, in tangible arrange\u00adments of steel and concrete, wires and tran\u00adsis\u00adtors, nuts and bolts.Tech\u00adno\u00adlog\u00adical choices bear directly on the \u201cpublic order\u201d at large. When we don\u2019t take those choices seri\u00adously\u2014or we\u2019re persuaded to ignore them by those insisting that tech\u00adnology is just a neutral tool\u2014we risk polit\u00adical conse\u00adquences.Winner warns of compla\u00adcency. Once the tech\u00adnology arrives and becomes entrenched, the conver\u00adsa\u00adtion gets reframed as one of tech\u00adno\u00adlog\u00adical inevitabilism vs. anachro\u00adnism, and dissent is discour\u00adaged: \u201cthe kinds of reasoning that justify the adap\u00adta\u00adtion of social life to tech\u00adnical require\u00adments pop up as spon\u00adta\u00adneously as flowers in the spring\u00a0\u2026\u00a0After a certain point, those who cannot accept the hard require\u00adments and imper\u00ada\u00adtives will be dismissed as dreamers and fools.\u201dLiberal democracyThe balance of power of democ\u00adracy is premised on the average person having leverage through creating economic value. If that\u2019s not present, I think things become kind of scary.\u2014a certain AI\u00a0CEOLiberal democ\u00adracy is the polit\u00adical scien\u00adtist\u2019s term for the type of govern\u00adment preva\u00adlent among capi\u00adtalist economies since the Amer\u00adican and French Revo\u00adlu\u00adtions. The intel\u00adlec\u00adtual foun\u00adda\u00adtion of liberal democ\u00adracy arose during the Enlight\u00aden\u00adment, espe\u00adcially through the work of John Locke. Liberal democ\u00adracy empha\u00adsizes limited govern\u00adment, indi\u00advidual rights, and sepa\u00adra\u00adtion of powers\u2014in short, majority rule with excep\u00adtions and guardrails. (The term liberal democ\u00adracy doesn\u2019t connote liberals or Democ\u00adrats in the specific US polit\u00adical sense. But polit\u00adical parties of differing ideolo\u00adgies are a tradi\u00adtional feature of liberal democ\u00adra\u00adcies.) Today, most liberal democ\u00adra\u00adcies are in Europe, the Amer\u00adicas, and the Pacific\u00a0Rim.Liberal democ\u00adracy is not a fixed set of immutable char\u00adac\u00adter\u00adis\u00adtics, but a bundle of graded values. All liberal democ\u00adra\u00adcies empha\u00adsize certain ones over others. In the aggre\u00adgate, some of these nations evolve toward stronger liber\u00adalism; others evolve away. These degraded cases have some\u00adtimes been called illib\u00aderal democ\u00adracy: the observ\u00adable formal\u00adi\u00adties of liberal democ\u00adracy may still be observed\u2014e.g. multi\u00adparty elec\u00adtions, sepa\u00adra\u00adtion of powers\u2014but the lived reality is single-party rule and declining indi\u00advidual rights.That\u2019s not to say that liberal democ\u00adracy produces excel\u00adlent outcomes for all citi\u00adzens, all the time. It doesn\u2019t. At any moment, certain citi\u00adzens are dissat\u00adis\u00adfied\u2014say, because they\u00a0belong to a group whose rights are inad\u00ade\u00adquately protected or econom\u00adi\u00adcally margin\u00adal\u00adized. Liberal democ\u00adracy offers a process, not a result: grass\u00adroots demo\u00adc\u00adratic partic\u00adi\u00adpa\u00adtion can coalesce into policy change. But within the arena of competing polit\u00adical inter\u00adests, winners and losers neces\u00adsarily follow. Navi\u00adgating these differ\u00adences within a stable, accom\u00admoda\u00adtive frame\u00adwork is prefer\u00adable to a rigid one that buckles under these stresses\u2014say, through polit\u00adical revo\u00adlu\u00adtion, which tends to be messy and unpre\u00addictable.Capi\u00adtalism has tradi\u00adtion\u00adally been consid\u00adered a neces\u00adsary but not suffi\u00adcient condi\u00adtion for liberal democ\u00adracy. Why? A regu\u00adlated market economy encour\u00adages citizen partic\u00adi\u00adpa\u00adtion through prop\u00aderty owner\u00adship and trans\u00adac\u00adtion. A prin\u00adcipal func\u00adtion of govern\u00adment is to define the economic condi\u00adtions of the state. Economic partic\u00adi\u00adpa\u00adtion mutu\u00adally rein\u00adforces demo\u00adc\u00adratic partic\u00adi\u00adpa\u00adtion. Prop\u00aderty owners will vote for those who protect their inter\u00adests. The rise of indus\u00adtrial capi\u00adtalism in the 19th century, and the wealth-redis\u00adtri\u00adb\u00adu\u00adtion mech\u00ada\u00adnisms that followed in the 20th, led to economic empow\u00ader\u00adment for more citi\u00adzens, and ulti\u00admately broader polit\u00adical empow\u00ader\u00adment. The converse also holds: economies premised on state or oligarchic control of some narrow class of assets haven\u2019t tended to evolve toward liberal democ\u00adracy.In prac\u00adtice, certain people in a capi\u00adtalist liberal democ\u00adracy tend to get increas\u00adingly rich. Absent coun\u00adter\u00admea\u00adsures, the wealthy gain control of the polit\u00adical appa\u00adratus, thwarting liberal-demo\u00adc\u00adratic norms. This tension between capital and poli\u00adtics is a long-consid\u00adered topic. A key early work was, of course, Karl Marx\u2019s Capital (about which more later). In the current era, Mancur Olson\u2019s book The Rise and Decline of Nations set out how small groups with a shared interest (which could include capital concen\u00adtra\u00adtion) can effec\u00adtively under\u00admine stable soci\u00adeties. More recently, econ\u00ado\u00admists Robert Reich (\u201cHow Capi\u00adtalism is Killing Democ\u00adracy\u201d), James Galbraith (The Predator State), and Yanis Varo\u00adufakis (Tech\u00adnofeu\u00addalism: What Killed Capi\u00adtalism) are among those who have studied the esca\u00adlating polit\u00adical conse\u00adquences of rising wealth inequality. The synthesis might be: as more wealth becomes concen\u00adtrated in the hands of fewer citi\u00adzens, liberal democ\u00adracy weakens, because whichever citi\u00adzens are losing economic rele\u00advance will also lose polit\u00adical rele\u00advance. A nation sending many of its citi\u00adzens toward economic irrel\u00ade\u00advance risks becoming polit\u00adi\u00adcally illib\u00aderal.The Skynet fallacyAI discourse often invokes sci-fi narra\u00adtives. I\u2019ve called this the Skynet fallacy, after the Termi\u00adnator antag\u00ado\u00adnist, the most cited. But any sci-fi will do. For instance, one AI CEO warned of AI \u201cgoing Termi\u00adnator\u201d; Stephen Hawking and other scien\u00adtists warned of AI \u201cdevel\u00adoping weapons we cannot even under\u00adstand\u201d; a second AI CEO said we \u201cdon\u2019t need much imag\u00adi\u00adna\u00adtion [about AI risk] because we grew up with that in the media\u201d; a third AI CEO invoked the sci-fi movies Contact and 2001: A Space Odyssey in a piece about AI risk; a US congressman summa\u00adrized AI risk as \u201cevil robots rising up to take over the world\u201d; a well-known jour\u00adnalist advo\u00adcated for more Termi\u00adnator analo\u00adgies; a promi\u00adnent AI-risk pundit said he\u2019s \u201cannoyed\u201d with Termi\u00adnator analo\u00adgies yet has suggested that AI will erad\u00adi\u00adcate humanity using hordes of toxic nanobots.Artis\u00adti\u00adcally, sci-fi movies exter\u00adnalize the awe and unease of tech\u00adno\u00adlog\u00adical confronta\u00adtion. It\u2019s easy to see why these metaphors have become part of AI-risk discourse. And yet. As AI puts down roots in our economy, the sci-fi framing hides more than it reveals. Sci-fi plots are opti\u00admized for cine\u00admatic impact. So as a metaphor for AI risk, they can lead to faulty intu\u00aditions. Among real\u00adistic AI risks, we can expect that most will be boring, slow, and depend on minimal extra tech\u00adnology. Whether AI will cause literal human extinc\u00adtion is esoteric\u2014a light\u00adning strike. But AI could easily induce future economic and polit\u00adical condi\u00adtions that most Amer\u00adi\u00adcans today would consider intol\u00ader\u00adable\u2014a cancer that extin\u00adguishes a certain way of life. Nobody\u2019s going to make a movie about boring AI risks. But they comprise the majority of worri\u00adsome AI outcomes.In 2003, philoso\u00adpher Nick Bostrom proposed his now well-known parable of the paper\u00adclip maxi\u00admizer. Bostrom imag\u00adines an advanced AI that is asked to make paper\u00adclips. Taking its mission seri\u00adously, the AI erad\u00adi\u00adcates humanity as it consumes all Earth resources to make more paper\u00adclips. Bostrom was illus\u00adtrating the AI control problem: ensuring an AI acts consis\u00adtently with human prior\u00adi\u00adties is diffi\u00adcult, even for osten\u00adsibly simple goals. Because when we say \u201cAI, make paper\u00adclips\u201d, the implied coda is \u201c\u2026\u00a0without killing everyone.\u201d But an AI can\u2019t know this ex nihilo. Bostrom softens the sci-fi flavor by choosing paper\u00adclips and not, say, laser-wielding robots. Bostrom\u2019s choice of an economic mech\u00ada\u00adnism of resource conver\u00adsion is apt. Even a mundane objec\u00adtive can produce outsize risk. We could further observe that on the paper\u00adclip-maxi\u00admizing path, human life would become dystopic long before literal extinc\u00adtion. As resources are depleted, humans would become tenants in a neofeu\u00addalist paper\u00adclip empire. (Paper\u00adclip Crisis: The Saga Begins\u2014opening soon.)Computer scien\u00adtist Stuart Russell also explored the control problem in his book Human Compat\u00adible. Russell calls one variant the gorilla problem: that \u201cances\u00adtors of the modern gorilla created \u2026\u00a0the genetic lineage leading to modern humans. How do the gorillas feel about this? \u2026\u00a0the consensus opinion would be very nega\u00adtive indeed.\u201d Despite shared lineage, gorillas and humans have incom\u00admen\u00adsu\u00adrable values. Nothing humans can do\u2014short of disap\u00adpearing\u2014would restore gorillas to their golden age. Russell\u2019s framing is a believ\u00adable analogy for the future rela\u00adtion\u00adship of humans and AI. Sure, gorillas didn\u2019t delib\u00ader\u00adately invent humans. But I take Russell to mean that the emer\u00adgent char\u00adac\u00adter\u00adis\u00adtics of these rela\u00adtion\u00adships are more conse\u00adquen\u00adtial than the intended ones. (In that sense, Russell echoes Langdon Winner.) The fact that we\u2019re inventing AI doesn\u2019t mean we will predict or control its gravest effects. Any more than gorillas could predict or control human domi\u00adnance of their ecosystem. The gorillas did their thing. We did\u00a0ours.AI will do its thing too. It will take time to figure out what that is, exactly.How the West was wonI recently read Cadillac Desert by Marc Reisner, about the devel\u00adop\u00adment of water\u00a0resources in the western US between 1910 and 1980, espe\u00adcially the dam-building campaign of the federal Bureau of Recla\u00adma\u00adtion. Reisner\u2019s book weaves several story\u00adlines:Engi\u00adneering and envi\u00adron\u00admental impacts. Dams concen\u00adtrate water. But they cause envi\u00adron\u00admental conse\u00adquences else\u00adwhere. Further\u00admore, Recla\u00adma\u00adtion\u2019s 20th-century projects were often based on opti\u00admistic projec\u00adtions of mete\u00ado\u00adro\u00adlog\u00adical water supply. Today, long-term drought condi\u00adtions chal\u00adlenge those projec\u00adtions. No amount of money or hydro\u00adlogic engi\u00adneering can change\u00a0that.National US poli\u00adtics. In the early 20th century, newer Western states sought polit\u00adical clout in the federal govern\u00adment, which had been domi\u00adnated by Eastern states. Polit\u00adical clout followed economic growth, and to achieve growth, the Western states depended on one crit\u00adical but scarce input: water. In the West, the federal Bureau of Recla\u00adma\u00adtion was tasked with increasing water supplies. A polit\u00adical symbiosis emerged: congressmen from Western states voted to fund Recla\u00adma\u00adtion water projects; in turn, Recla\u00adma\u00adtion looked out for their constituents. Over decades, Reisner depicts this rela\u00adtion\u00adship as metas\u00adta\u00adsizing from prac\u00adti\u00adcality to corrup\u00adtion, in the sense of Recla\u00adma\u00adtion becoming beholden to a narrow polit\u00adical lane. In that sense, Recla\u00adma\u00adtion\u2019s dams arguably qualify as inher\u00adently polit\u00adical tech\u00adnology.State economies. Recla\u00adma\u00adtion\u2019s highly subsi\u00addized water boot\u00adstrapped Western economies, espe\u00adcially agri\u00adcul\u00adture. For a while, it worked as promised: Farmers got irri\u00adga\u00adtion. Cities got water. Western states grew and pros\u00adpered. But the projects worked so well that Western states wanted more. These states never weaned them\u00adselves from subsi\u00addized federal water, setting them on a path toward perma\u00adnent depen\u00addence.The parallel between water and AI is inexact. Water is a biolog\u00adical neces\u00adsity; AI is not. Recla\u00adma\u00adtion\u2019s projects worked (up to a point); AI may or may not. This is part of why AI propo\u00adnents have sought to raise the stakes. So far, AI has been grue\u00adsomely expen\u00adsive and deliv\u00adered middling results. Never\u00adthe\u00adless it\u2019s routinely depicted as a geopo\u00adlit\u00adical fulcrum, a proxy for contin\u00aduing US excep\u00adtion\u00adalism. If Amer\u00adi\u00adcans don\u2019t adopt AI whole\u00adheart\u00adedly, we will be losers. Do you want to be a loser?Labor replacementQ: What is Big AI primarily selling? A: Labor replace\u00adment, with mass unem\u00adploy\u00adment as a likely conse\u00adquence. Some disagree or call it doomerish. Why? It\u2019s exactly what AI grandees have been telling us. A certain AI CEO wrote that AI \u201cwill be hugely desta\u00adbi\u00adlizing for hundreds of millions\u201d and that AI tools \u201care funda\u00admen\u00adtally labor replacing\u201d. A certain AI company released a research paper about \u201cthe labor market impact poten\u00adtial of large language models\u201d. That AI CEO said \u201cjobs are defi\u00adnitely going to go away, full stop\u201d. A second AI CEO said that in the near future, \u201c20% of people don\u2019t have jobs.\u201d A third AI CEO predicted that farther out, \u201cprob\u00adably none of us will have a job.\u201d An AI-adja\u00adcent CEO said that AI \u201cwill destroy human\u00adi\u00adties jobs\u201d. The ball is not hidden.Capital markets are already pricing in these expec\u00adta\u00adtions. Regard\u00adless of whether Big AI even\u00adtu\u00adally delivers mass labor replace\u00adment, today these compa\u00adnies seek to concen\u00adtrate capital as if they will. According to the Wash\u00adington Post, AI capital expen\u00addi\u00adture in 2026 is esti\u00admated to be $700 billion, a \u201cspending spree [that] has few prece\u00addents\u201d. Based on a recent survey of US workers, the Global Part\u00adner\u00adship on Arti\u00adfi\u00adcial Intel\u00adli\u00adgence said that \u201cthe policy window to shape how AI trans\u00adforms work is prob\u00adably closing faster than most govern\u00adments realize.\u201dExtra\u00ador\u00addi\u00adnary invest\u00adment demands extra\u00ador\u00addi\u00adnary returns. In early 2026, after a certain billion\u00adaire tech CEO laid off 40% of his employees and attrib\u00aduted the deci\u00adsion to a new \u201ccore thesis\u201d of AI, the company\u2019s stock rose nearly 25%. He won\u2019t be the last. Whether these layoffs are based on actual AI bene\u00adfits or merely \u201cantic\u00adi\u00adpa\u00adtory\u201d is neither here nor there. Employers have strong incen\u00adtives to reduce head\u00adcount and increase AI spending before competi\u00adtors do. We will increas\u00adingly see both kinds of layoffs. Soft\u00adware program\u00admers are one set of conse\u00adquen\u00adtial, highly paid writers likely to be replaced with AI. Else\u00adwhere I\u2019ve predicted that legal prac\u00adtice will also be seri\u00adously impaired. Why? Like program\u00admers, lawyers write about conse\u00adquen\u00adtial things and charge a lot. (We can expect that AI compa\u00adnies them\u00adselves, fond of dogfooding, will perfor\u00adma\u00adtively set an example through their own AI-driven layoffs.)Two common objec\u00adtions to this conclu\u00adsion:\u201cAI won\u2019t replace workers, it will enhance them.\u201d That may turn out to be true. But for now, it\u2019s not what AI CEOs are pitching. And it\u2019s not how the market is valuing Big AI. AI customers will natu\u00adrally pay less for a worker-assis\u00adtive tech\u00adnology than tech\u00adnology that reduces head\u00adcount, which is one of the biggest expenses at any employer. The labor-replace\u00adment story is every\u00adthing to every\u00adbody. If AI is just a neat way of, say, finding photos of your dog\u00a0on your phone, none of the AI bets pay\u00a0out.\u201cAI will replace certain jobs, but also create new ones.\u201d Fair, but \u201ccreate new ones\u201d is skimpy on details. How do the jobs created compare to the ones lost? Do they pay more? Or less, as the supply of workers seeking work increases? Do the new jobs require different skills? Who bears the cost of the career switch? None of this is auto\u00admatic. As econ\u00ado\u00admist Carl Benedikt Frey said: \u201cMost econ\u00ado\u00admists will acknowl\u00adedge that tech\u00adno\u00adlog\u00adical progress can cause some adjust\u00adment prob\u00adlems in the short run. What is rarely noted is that the short run can be a life\u00adtime.\u201dAfter going hard on this narra\u00adtive for several years, AI CEOs are now soft\u00adening their tone, primarily in response to public back\u00adlash. For instance, the AI CEO who once said \u201cjobs are defi\u00adnitely going to go away\u201d now wants us to know that watching AI answer his emails felt \u201cdehu\u00adman\u00adizing\u201d and that he now values \u201cthe human part of the roles\u201d. That\u2019s what it finally took\u2014using your own product? Ed Zitron has char\u00adac\u00adter\u00adized the AI market so far as a blink\u00adered conver\u00adsa\u00adtion between AI exec\u00adu\u00adtives and trad exec\u00adu\u00adtives, all of whom \u201chave little idea what work looks like\u201d. This is more a feature than a bug. To replace labor, AI doesn\u2019t need to actu\u00adally deliver compa\u00adrable outputs. It only needs to fulfill a certain exec\u00adu\u00adtive fantasy of \u201cwhat work looks like\u201d. That\u2019s much easier than, say, achieving AGI (however defined). On that view, AI may turn out to be a Veblen good\u2014an item that buyers value primarily for signaling their ability to afford\u00a0it.One more small matter\u2014funding of the US govern\u00adment is premised on employ\u00adment. In 2023, the US got 48% of its revenue from income taxes and 36% from other payroll taxes. If AI causes 30% unem\u00adploy\u00adment, revenues will drop and expenses will jump. Of course, higher unem\u00adploy\u00adment also leads to lower aggre\u00adgate consumer spending. These would be fiscal condi\u00adtions unknown since the Great Depres\u00adsion, and utterly foreign to Amer\u00adi\u00adcans living today.The goodies economyBig AI\u2019s message to poten\u00adtial corpo\u00adrate customers of AI has centered on labor replace\u00adment. Correctly fore\u00adseeing that the workers them\u00adselves might take umbrage, Big AI has pitched them a sepa\u00adrate story: that workers will enjoy higher stan\u00addards of mate\u00adrial well-being under\u00a0AI.One billion\u00adaire AI CEO predicted \u201cuniversal high income\u201d and that \u201cAI and robots will provide any goods and services that you want.\u201d Another billion\u00adaire AI CEO predicted \u201cuniversal extreme wealth for every\u00adbody\u201d. A billion\u00adaire AI venture capi\u00adtalist \u201cbelieve[s] in making everyone rich, every\u00adthing cheap, and every\u00adthing abun\u00addant\u201d and that \u201cthe ulti\u00admate result will be that all phys\u00adical goods become as cheap as pencils.\u201d A tech billion\u00adaire predicted \u201cmore goods and services with less labor\u201d and \u201calways-avail\u00adable, high-quality medical advice.\u201d Another billion\u00adaire AI venture capi\u00adtalist predicted \u201cthe abun\u00addance of goods and services \u2026\u00a0will be very, very large. Prices will be very, very, low.\u201d Hope\u00adfully yachts will come down\u00a0too.But wait\u2014will they? In the post-AI economy, will we all have yachts? Of course not. As usual, we should be skep\u00adtical of billion\u00adaires. Espe\u00adcially those suddenly\u00a0claiming to care about the purchasing power of workers. Are they fibbers or just fools? I think it\u2019s more subtle: these billion\u00adaires are performing an idio\u00adsyn\u00adcratic kind of class empathy, prob\u00adably at the behest of publi\u00adcists. This feeling doesn\u2019t come natu\u00adrally. So, like a cyborg reading a love poem, it\u2019s stilted. The persis\u00adtent lack of detail signals that these billion\u00adaires have no idea how, exactly, AI will accom\u00adplish this. But the subtext is clear: If toler\u00adating worker purchasing power is what it takes to become a tril\u00adlion\u00adaire, that\u2019s a sacri\u00adfice I\u2019m willing to\u00a0make.Some\u00adtimes this theory goes by fancy names like pros\u00adperity or abun\u00addance. Some\u00adtimes it\u2019s linked to specific policy mech\u00ada\u00adnisms like universal basic income. It all comes back to the same idea. I\u2019m just going to call this stuff goodies. The theory is that under AI, citi\u00adzens are going to get some nice level of goodies.A few reasons we might be skep\u00adtical of AI-goodies theory. The price of any consumer good has a lower bound set by the cost of its inputs (if one is not selling at a loss). But labor is only one input. So even if AI reduces labor costs to $0, the cost of other inputs\u2014e.g., energy and raw mate\u00adrials\u2014remain. Even if AI lowers commodity prices by increasing their supply (say, through better discovery and extrac\u00adtion tech\u00adniques), they are still rival\u00adrous resources. Not all are avail\u00adable domes\u00adti\u00adcally. Postwar, rising global consump\u00adtion of oil in the West influ\u00adenced the forma\u00adtion of OPEC by future petrostates who sought to capture a share of oil-derived wealth. Simi\u00adlarly, in a post-AI US economy, foreign nations with crit\u00adical minerals\u2014e.g., lithium, cobalt, nickel\u2014may like\u00adwise cartelize them, raising costs and creating a price floor. Related kinds of supply control may also arise domes\u00adti\u00adcally. Today, a certain US agri\u00adcul\u00adture company uses patented seeds to exert control over a market of osten\u00adsibly inde\u00adpen\u00addent farmers. If AI dramat\u00adi\u00adcally reduces the costs of manu\u00adfac\u00adturing and agri\u00adcul\u00adture, Big AI may seek to capture value from the facto\u00adries and farms that benefit (say, through a sales royalty). AI-goodies theory implies that Big AI will allow maximum produc\u00adtion of goods, leading to lowest prices. But the huge finan\u00adcial projec\u00adtions around AI\u2014set by Big AI itself\u2014may force them to do other\u00adwise. That is, Big AI\u2019s promises of world-histor\u00adical goodies for consumers and world-histor\u00adical profits for share\u00adholders may be incom\u00adpat\u00adible. If so\u2014guess who prevails?Still, to be fair to Big AI, I\u2019ll accept their argu\u00adment that citi\u00adzens will enjoy higher stan\u00addards of mate\u00adrial well-being under AI through goodies. But the goodies must be paid for. Somehow, wealth must be trans\u00adferred from Big AI to everyone else. These trans\u00adfers can happen by three basic methods:Direct. Big AI directly provides goodies to citi\u00adzens. As one AI CEO put it, to \u201ccapture much of the world\u2019s wealth \u2026 and then redis\u00adtribute this wealth to the people\u201d.Tax and spend. Govern\u00adment taxes Big AI at some high rate, and spends the money on public goodies. But the idea that Amer\u00adican AI compa\u00adnies would consent to high tax rates to ensure public goodies is incon\u00adceiv\u00adable because it vests more control in govern\u00adment, which no free-marke\u00adteer believes should be so.\u2021Market-based. Big AI dramat\u00adi\u00adcally reduces the cost of goods and services, which then flow into the market at extremely low prices. The flaw in this method is that AI will reduce prices of certain things more than others. For instance, it\u2019s easy to see how AI will make streaming movies free; food and health\u00adcare and trans\u00adporta\u00adtion, less\u00a0easy.These methods aren\u2019t mutu\u00adally exclu\u00adsive. I expect the most likely method is the first, where Big AI directly provides the goodies, in the style of Gilded Age company towns. Big AI would avoid govern\u00adment inter\u00adven\u00adtion through taxa\u00adtion and redis\u00adtri\u00adb\u00adu\u00adtion. Big AI would control the goodies and who gets them. Big AI would be clearly under\u00adstood as the goodie provider. Assuming the massive profits mate\u00adri\u00adalize, Big AI could afford to make diverse baskets of goodies that include even things like food and health\u00adcare and trans\u00adporta\u00adtion.Keep in mind the goodies economy will play out against a back\u00addrop of wors\u00adening AI-induced labor-market effects\u2014high unem\u00adploy\u00adment, but also wage declines, under\u00adem\u00adploy\u00adment, misem\u00adploy\u00adment, and so forth. There\u00adfore, a side effect of the goodies economy is that what\u00adever we think consti\u00adtutes the social safety net will shift from a service of govern\u00adment to a service of private industry, specif\u00adi\u00adcally Big AI. One AI CEO shame\u00adlessly predicts exactly this outcome: that there is \u201coppor\u00adtu\u00adnity for AI to be used to help provi\u00adsion govern\u00adment services \u2026\u00a0includ[ing] health services, the DMV, taxes, social secu\u00adrity, building code enforce\u00adment and so on.\u201d But don\u2019t worry\u2014he concludes that because \u201cpoorly imple\u00admented services\u201d cause \u201ccyni\u00adcism about govern\u00adment\u201d, AI inter\u00adme\u00addi\u00ada\u00adtion would \u201cstrengthen[] respect for demo\u00adc\u00adratic gover\u00adnance.\u201d And to be clear\u2014by \u201cAI\u201d, he means \u201cmy privately owned AI\u201d, and by \u201coppor\u00adtu\u00adnity \u2026\u00a0to help\u201d he means a massive govern\u00adment contract.This portends a polit\u00adical shift. Citi\u00adzens once treated govern\u00adment as a polit\u00adical bulwark against the encroach\u00adments and over\u00adreaches of industry. In the goodies economy, citi\u00adzens will tend to align with Big AI, since it provides the goodies, which for many citi\u00adzens may include basic needs. Under those condi\u00adtions, if Big AI has a conflict with govern\u00adment\u2014whose side will citi\u00adzens\u00a0take?Indus\u00adtries have often insu\u00adlated them\u00adselves against polit\u00adical conse\u00adquences with popular prod\u00aducts. Arguably, in the last 25 years, the tech industry\u2014espe\u00adcially smart\u00adphone and social-media compa\u00adnies\u2014has bene\u00adfited from light regu\u00adla\u00adtory scrutiny rela\u00adtive to its size. Exam\u00adples from the last century include the auto\u00admo\u00adtive and fast-food indus\u00adtries, which like\u00adwise enjoyed decades of gentle regu\u00adla\u00adtion, even though both had serious public-health impacts.To improve public accep\u00adtance, Big AI has so far empha\u00adsized their fun consumer-facing prod\u00aducts. These prod\u00aducts\u2014e.g., gener\u00adating an image, or writing a story, or riding in a driver\u00adless car\u2014make AI seem harm\u00adless: the citizen thinks of their own usage as trivial and fun, and there\u00adfore models the aggre\u00adgate effect of AI as a sum of simi\u00adlarly trivial indi\u00advidual usages. Consis\u00adtent with this, Big AI has often framed these tools as \u201cdemoc\u00adra\u00adtizing\u201d (ex.\u00a02, ex.\u00a03) implying that oppo\u00adnents of these systems are in some sense anti-democ\u00adracy (or as Winner predicted, antitech\u00adnology and antiprogress).In the goodies economy, as citi\u00adzens lose inde\u00adpen\u00addence as economic actors, they will also lose inde\u00adpen\u00addence as polit\u00adical actors. They will be captured by industry. Econ\u00ado\u00admist Yanis Varo\u00adufakis has called the current early stage of this arrange\u00adment \u201ctech\u00adnofeu\u00addalism\u201d, a term suggesting how late capi\u00adtalism erodes indi\u00advidual economic agency. Through its provi\u00adsion of goodies, Big AI intends to complete this tran\u00adsi\u00adtion.An economic histo\u00adrian might say that a polit\u00adical takeover by Big AI is unlikely, citing successful worker-led polit\u00adical move\u00adments in the US and Europe. If the US reaches 30% AI-induced unem\u00adploy\u00adment, the thinking goes, workers will rise up and demand change through polit\u00adical mech\u00ada\u00adnisms of our liberal democ\u00adracy. There\u2019s reason to doubt these paral\u00adlels. These events occurred earlier in the tech\u00adno\u00adlog\u00adical era, when industry and workers were adver\u00adsarial yet remained inter\u00adde\u00adpen\u00addent. Citi\u00adzens retained irre\u00adducible leverage with both industry (as laborers) and govern\u00adment (as voters). In the decades after the US Gilded Age, this combi\u00adna\u00adtion of economic and elec\u00adtoral power was crit\u00adical to the rising status of workers. But in a future post-AI economy, workers could lose their jobs because of Big AI while remaining depen\u00addent on Big AI for basic needs. How do you bite the hand that feeds? Further\u00admore, many previous conflicts between industry and workers were not gentle adjust\u00adments, but violent (say, the Luddite move\u00adment of the 1810s, and Gilded Age strikes of the late 1800s) or caused devas\u00adtating, perma\u00adnent social tran\u00adsi\u00adtions (say, the 1917 Russian Revo\u00adlu\u00adtion).Legally, shifting the social safety net from govern\u00adment to Big AI also means that citizen enti\u00adtle\u00adment to these goodies will shift from public statu\u00adtory and consti\u00adtu\u00adtional law to private-contract law. Private compa\u00adnies can with\u00adhold goodies for many reasons the govern\u00adment cannot. For instance: if you talk shit about the govern\u00adment, you will still get your unem\u00adploy\u00adment check, because the First Amend\u00adment says so. But no such restric\u00adtion on Big AI: if you talk shit about them, they will be free to revoke your goodies. Can\u2019t happen? Already does:\u00a0credit reports are a private system that largely controls who gets favor\u00adable interest rates. Or employer-provided health\u00adcare, a struc\u00adturally unnec\u00ades\u00adsary tradi\u00adtion that distorts incen\u00adtives for workers by making their continued health contin\u00adgent on continued employ\u00adment. Most likely, the polit\u00adical align\u00adment between citi\u00adzens and Big AI will not only be enforced by carrot, but also by stick.Of course, an industry seeking to induce polit\u00adical align\u00adment from citi\u00adzens would benefit not only from distrib\u00aduting goodies, but also weak\u00adening the govern\u00adment. In 2023, would you have believed this predic\u00adtion?\u00a0That an AI CEO would endorse a pres\u00adi\u00adden\u00adtial candi\u00addate. That the candi\u00addate would win. That on his first day in office, the new pres\u00adi\u00addent would install the AI CEO in the govern\u00adment with broad powers, in pursuit of a nebu\u00adlous project of cutting waste and saving money. That ulti\u00admately, this project would save little money and primarily impair govern\u00adment agen\u00adcies designed to protect the vulner\u00adable. That somehow the parts of the govern\u00adment that provided large contracts for the same AI CEO would remain untouched. To be fair, for centuries US busi\u00adnesses have sought to bend US public policy to suit their inter\u00adests, often with great success. This is a theme of Galbraith\u2019s book The Predator State. What\u2019s new is the mous\u00adtache-twirling flagrancy of the project, and that it was under\u00adtaken at the behest of the\u00a0newly elected pres\u00adi\u00addent. Some called it a coup, though one premised on abdi\u00adca\u00adtion of power rather than seizure.\u2021 As of June 2026, reporting has suggested that an AI CEO has proposed to the current US pres\u00adi\u00addent that the US take an equity stake in the AI CEO\u2019s company (and maybe others). This arrange\u00adment would be very different from taxa\u00adtion. It fore\u00adsees a transfer of capital from the US govern\u00adment to the corpo\u00adra\u00adtion to buy equity. As a share\u00adholder, the US govern\u00adment would be enti\u00adtled to a share of future profits, which may never exist. It also poten\u00adtially creates a conflict of interest: as a Big AI share\u00adholder, would the US govt regu\u00adlate or legis\u00adlate the AI industry? Taxa\u00adtion, by contrast, requires no prepay\u00adment by the US govern\u00adment, and can be assessed on the current activ\u00adi\u00adties of the AI company, regard\u00adless of prof\u00aditability. Histor\u00adi\u00adcally, the US govern\u00adment has taken equity stakes in private corpo\u00adra\u00adtions as an indi\u00adrect bailout. Which is maybe the better word for what is being proposed here. Absent this exigency, taxa\u00adtion is econom\u00adi\u00adcally and polit\u00adi\u00adcally prefer\u00adable. This AI CEO may also want to ask his chatbot what happened to the oil industry in Venezuela under Ch\u00e1vez.Know your enemyOh Karl, the world isn\u2019t fairIt isn\u2019t and never will beThey tried out your planIt brought misery insteadIf you\u2019d seen how they worked itYou\u2019d be glad you were dead.\u2014Randy Newman \u201cThe World Isn\u2019t\u00a0Fair\u201dTo iden\u00adtify as a Marxist today is rare (at least in the US). More often, it\u2019s wielded as an anti-capi\u00adtalist epithet against someone who endorses socialist or Commu\u00adnist methods. True\u2014Marx himself supported these methods. After all, he did write The Commu\u00adnist Mani\u00adfesto. But short\u00adhand Marxism tends to omit the reasoning that Marx, a philoso\u00adpher and histo\u00adrian, consid\u00adered vital. After all, he did write Capital.Marx saw class struggle as the key insti\u00adgator of history. He argued that capi\u00adtalism was polit\u00adi\u00adcally unstable, and neces\u00adsarily led to worker revolt. Marx\u2019s theory that economic value orig\u00adi\u00adnates in labor over\u00adlaps concep\u00adtu\u00adally with Locke\u2019s labor theory of prop\u00aderty. But to Marx, capi\u00adtalism sepa\u00adrated\u2014or using his preferred term, alien\u00adated\u2014workers from the fruits of that labor.In prac\u00adtice, Marx\u2019s theo\u00adries have had mixed success. But Marx\u2019s obser\u00adva\u00adtions in Capital about the inter\u00adac\u00adtion of workers and machines remain rele\u00advant, even though he was writing in the mid-1800s, when the pinnacle of tech\u00adnology was textile machinery. Looking back on worker upris\u00adings since the 1600s\u2014including the Luddite move\u00adment that famously protested those textile machines\u2014Marx wrote:It took both time and expe\u00adri\u00adence before the workpeople learnt to distin\u00adguish between machinery and its employ\u00adment by capital, and to direct their attacks, not against the mate\u00adrial instru\u00adments of produc\u00adtion, but against the mode in which they are\u00a0used.By protesting the textile machines, the Luddites were missing the bigger picture. The Luddites (and earlier workers) were raging against the literal machine\u2014the specific new tech\u00adnology. But the tech\u00adnology was a symptom, not a cause. Instead, workers needed to rage against the bigger, figu\u00adra\u00adtive machine\u2014the extrac\u00adtive capi\u00adtalist system.Marx\u2019s obser\u00adva\u00adtion has a subtler impli\u00adca\u00adtion too. New tech\u00adnology often holds itself out as the starting point of a narra\u00adtive: from now on, every\u00adthing is different. When we consider the tech\u00adnology alone, that narra\u00adtive domi\u00adnates. But when we zoom out and consider the histor\u00adical context, the new tech\u00adnology becomes the current endpoint of a much longer polit\u00adical narra\u00adtive.What would Marx say to AI critics\u2014social, legal, economic, polit\u00adical\u2014that have arisen so far? Maybe that we\u2019re missing the bigger picture. That as a human inven\u00adtion, AI may be the starting point of a new tech\u00adno\u00adlog\u00adical narra\u00adtive. But as an affront to human workers, it continues a long tradi\u00adtion of capi\u00adtalist tech\u00adnolo\u00adgies, begin\u00adning with the Indus\u00adtrial Revo\u00adlu\u00adtion (if not earlier).For their part, the capi\u00adtal\u00adists who own a new tech\u00adnology usually prefer to frame it as a starting point. Conve\u00adniently, this omits the grubby histor\u00adical context that might dampen the marketing. Similar sleight of hand animates the snide techie\u2019s favorite sick burn\u2014Luddite\u2014connoting a block\u00adhead irra\u00adtionally opposed to tech\u00adnology. Of course, this usage is delib\u00ader\u00adately ahis\u00adtor\u00adical\u2014read Brian Merchant\u2019s Blood in the Machine. Perhaps there should be a companion term for a block\u00adhead irra\u00adtionally committed to\u00a0AI.These two frames for under\u00adstanding new tech\u00adnology under capi\u00adtalism also suggest two corre\u00adsponding analytic approaches to AI\u00a0risk:AI as tech\u00adnology. On this view, AI is a starting point, and we consider the risks of what it can do that previous tech\u00adnolo\u00adgies could not. This isn\u2019t wrong, exactly. It works for certain char\u00adac\u00adter\u00adis\u00adtics. For instance: I can believe that AI will facil\u00adi\u00adtate hacking and similar crim\u00adi\u00adnality like no tech\u00adnology before. That risk is specific to AI as new tech\u00adnology and deserves consid\u00ader\u00ada\u00adtion.AI as capi\u00adtalist instru\u00adment. But for other char\u00adac\u00adter\u00adis\u00adtics\u2014say, AI\u2019s effects on labor, public wealth, and the economy\u2014thinking of AI strictly as new tech\u00adnology embeds the error Marx warns against. For those char\u00adac\u00adter\u00adis\u00adtics, we should consider how AI may amplify or accel\u00aderate existing trends. Put differ\u00adently\u2014the ways in which AI is inher\u00adently polit\u00adical tech\u00adnology.In sum\u2014as new tech\u00adnology, AI acts as a creator of new risks; as a capi\u00adtalist instru\u00adment, it acts as an ampli\u00adfier or accel\u00ader\u00adator of existing risks and trends in the capi\u00adtalist system.The strongest version of this framing is that the \u201campli\u00adfier or accel\u00ader\u00adator\u201d part is precisely Why AI Exists. Unprece\u00addented billions have already been poured into AI. Why? Because it\u2019s expected to deliver more profit and concen\u00adtrate more wealth than any previous tech\u00adnology. But ampli\u00adfying these trends also means ampli\u00adfying the atten\u00addant polit\u00adical risks. Earlier I accepted the argu\u00adment that Big AI will deliver abun\u00addant goodies to citi\u00adzens\u2014which AI propo\u00adnents consider one of the strongest points in its favor. But doing so will likely increase align\u00adment between Big AI and citi\u00adzens, with nega\u00adtive polit\u00adical conse\u00adquences. As usual\u2014no free lunch.This shift in framing also shifts the eviden\u00adtiary context. When we think about AI risk, we\u2019re neces\u00adsarily making guesses about the future. But when we frame AI in the narrow sense of new tech\u00adnology, we\u2019re primarily consid\u00adering a time\u00adline that starts now. Whereas when we shift to thinking of AI as a capi\u00adtalist instru\u00adment, we\u2019re consid\u00adering a time\u00adline that starts centuries ago and has evolved contin\u00adu\u00adously into the present. We can and should study those existing economic and polit\u00adical trends, because those will likely shape the future trajec\u00adtory. Put differ\u00adently: AI may be new. But it\u2019s not immune to history.The poisoned chaliceI think the biggest chal\u00adlenge to AI in this country is polit\u00adical unrest \u2026\u00a0Can you make more money? It\u2019s all irrel\u00ade\u00advant if the country blows up \u2026 If I were sitting here in private with my peers, I\u2019d be telling them \u2026 the country could blow up polit\u00adi\u00adcally. And none of us are going to make any money when the country blows up.\u2014an AI-adja\u00adcent\u00a0CEOBig AI\u2019s goal of labor replace\u00adment has two dimen\u00adsions some\u00adtimes over\u00adlooked, although they are both hidden in plain sight. The first relates to AI\u2019s effort to replace knowl\u00adedge workers. The other pertains to the compet\u00adi\u00adtive effects of doing\u00a0so.AI replacing knowl\u00adedge workers. Labor replace\u00adment is the mech\u00ada\u00adnism. Struc\u00adturally, Big AI seeks to build a moat around US economic growth by normal\u00adizing AI as a cogni\u00adtive inter\u00adme\u00addiary where knowl\u00adedge work\u2014roughly, econom\u00adi\u00adcally valu\u00adable thinking and creativity\u2014will happen.This maneuver is part of the automa\u00adtion play\u00adbook, espe\u00adcially for the tech industry. During the rise of desktop automa\u00adtion in the 1980s and 1990s, tech compa\u00adnies sought to build moats by control\u00adling the file formats, programs, and plat\u00adforms where docu\u00adments were created. As the commer\u00adcial internet ripened during the 2000s, tech compa\u00adnies sought moats based on lever\u00adaging network effects\u2014espe\u00adcially the unholy dyad of adver\u00adtising and social media.How Big AI plans to profit from this inter\u00adme\u00addi\u00ada\u00adtion is an open ques\u00adtion. One AI company has suggested taking a cut of AI-assisted discov\u00aderies. The logis\u00adtics and legal\u00adi\u00adties would be boggling. Details\u2014what\u00adever. For now, AI compa\u00adnies largely agree on the first step: make workers depen\u00addent on AI to do their jobs, just as tech fore\u00adbears made workers depen\u00addent on a certain soft\u00adware program to share a file, or on a certain website to have friends. This time, however, the soft\u00adware ulti\u00admately consumes the worker.Big AI\u2019s timing is canny. In recent decades, intan\u00adgible assets have played an increasing role in global economic growth. An intan\u00adgible asset lacks phys\u00adical embod\u00adi\u00adment, and includes\u00a0formal intel\u00adlec\u00adtual-prop\u00aderty assets (e.g., copy\u00adrights, patents) but also informal (e.g., brand good\u00adwill, other knowl\u00adedge-work assets). Tangible assets are the oppo\u00adsite\u2014bricks, mortar, inven\u00adtory, and so forth. According to the WIPO, since 1995 global invest\u00adment in intan\u00adgible assets has steadily increased, and since 2009 has exceeded that of tangible assets. By wide margin, the US leads invest\u00adment in intan\u00adgible assets.For a long time, intan\u00adgible assets were under\u00advalued because tradi\u00adtional GDP measures and accounting prac\u00adtices excluded them. In the 1970s and 1980s, during the first big wave of IT invest\u00adment, the US expe\u00adri\u00adenced a coun\u00adter\u00adin\u00adtu\u00aditive slow\u00addown in produc\u00adtivity growth, an effect econ\u00ado\u00admist Erik Bryn\u00adjolf\u00adsson called the \u201cproduc\u00adtivity paradox\u201d. Part of the reso\u00adlu\u00adtion to the produc\u00adtivity paradox is that returns from IT invest\u00adment were mani\u00adfesting not as labor produc\u00adtivity per se, but as intan\u00adgible assets. For busi\u00adnesses, IT invest\u00adment was worth\u00adwhile even if it didn\u2019t increase labor produc\u00adtivity, because it still gener\u00adated intan\u00adgible assets. (Another part of the reso\u00adlu\u00adtion to the produc\u00adtivity paradox is Bryn\u00adjolf\u00adsson\u2019s theory of lagging comple\u00admen\u00adtary invest\u00adment, which will also likely apply to\u00a0AI.)Never\u00adthe\u00adless, increasing labor produc\u00adtivity matters a lot to workers. Why? It has histor\u00adi\u00adcally been the biggest driver of wage growth. Which brings us to another\u2014maybe the most\u2014conse\u00adquen\u00adtial trend in the US economy over the last 50 years: wage stag\u00adna\u00adtion, or the decou\u00adpling of produc\u00adtivity and wage growth. As the story is typi\u00adcally told: from 1948 to 1973, wage growth (91.3%) basi\u00adcally matched produc\u00adtivity growth (96.7%). Since 1973, however, overall produc\u00adtivity has grown (74.4%) much faster than wages (9.2%). Where has the money gone instead? Into the pockets of busi\u00adness owners, mostly. Over the same time\u00adframe, the oper\u00adating profits of busi\u00adnesses (as measured by net oper\u00adating surplus) has steadily increased. Today, if wage growth had kept up with produc\u00adtivity gains since 1973, we\u2019d expect wages to be more than double.Together, these trends have been a long-term double whammy for US workers. Because of wage stag\u00adna\u00adtion, workers have been gener\u00adating increasing value for employers (as produc\u00adtivity) but not capturing via wages the same share of that value they once did. Because of the shift toward intan\u00adgible assets, employers have invested less in labor produc\u00adtivity, further inhibiting wage growth. As AI replaces knowl\u00adedge workers, we can expect it to amplify both trends.Will it? Recent labor-market research by one major AI company suggests that the greatest \u201ctheo\u00adret\u00adical capa\u00adbility\u201d for AI labor replace\u00adment lies with tradi\u00adtional cate\u00adgories of knowl\u00adedge work: finance, program\u00adming, engi\u00adneering, science, legal, and the arts. A research group at Tufts Univer\u00adsity was blunter: \u201cthe more AI helps you do your job, the more expend\u00adable you can become. Finance profes\u00adsionals, teachers/profes\u00adsors, creative profes\u00adsionals, accoun\u00adtants and audi\u00adtors, legal profes\u00adsionals\u201d are at risk. Even Bryn\u00adjolf\u00adsson said he was \u201csurprised\u201d that knowl\u00adedge work would be so readily replaced.\u201cTech\u00adnology always makes certain jobs obso\u00adlete; new ones will arise.\u201d AI\u2019s predicted labor replace\u00adment is unprece\u00addented in three ways: the diver\u00adsity of tasks replaced; its outsize effect on highly educated workers; and the back\u00addrop of 50 years of wage stag\u00adna\u00adtion. Automa\u00adtion-driven tran\u00adsi\u00adtions aren\u2019t neces\u00adsarily easy, even when they\u2019re narrow and the economy can absorb the workers. Those who hand\u00adwave over the details should study histor\u00adical exam\u00adples. When you tell a large group of workers that their skills no longer have economic value, you risk a polit\u00adical\u00a0and social tinderbox. Recall Carl Benedikt Frey\u2019s comment: \u201cthe short run can be a life\u00adtime\u201d.Confi\u00adden\u00adtial to owners of capital: nothing I\u2019m saying above depends on a norma\u00adtive view about wealth distri\u00adbu\u00adtion. I person\u00adally oppose wage stag\u00adna\u00adtion and believe workers\u00a0should enjoy wage growth propor\u00adtional (at minimum) to produc\u00adtivity\u00a0growth. If you don\u2019t\u2014fine. The argu\u00adment persists. Even accepting that the AI labor tran\u00adsi\u00adtion is econom\u00adi\u00adcally rational and neces\u00adsary, it will still\u00a0likely have polit\u00adi\u00adcally desta\u00adbi\u00adlizing effects that you, owners of capital, will find unpleasant. Plus, you\u2019ll get punched in the wallet too\u2014keep reading.The compet\u00adi\u00adtive effects of replacing knowl\u00adedge workers. The previous section consid\u00adered how AI labor replace\u00adment will affect the rela\u00adtion\u00adship between workers and employers. But it will also induce more carniv\u00ado\u00adrous compe\u00adti\u00adtion among firms that adopt AI, and thus indi\u00adrectly among the owners of capital. Whereas the last 50 years of the US economy have featured wage stag\u00adna\u00adtion, the post-AI economy may feature some\u00adthing akin to capital stag\u00adna\u00adtion, for non-AI cate\u00adgories of capital. A growing economy used to be a rising tide that lifted all boats. Post-AI, it may only lift a handful of yachts. Everyone else will be in dry dock. Arguably that capital tran\u00adsi\u00adtion is already underway.How might this work? The economy is increas\u00adingly driven by intan\u00adgible assets. Knowl\u00adedge workers provide those assets. If a company lays off its knowl\u00adedge workers in favor of adopting some vendor\u2019s AI, at first it gets the same produc\u00adtivity at much lower cost. So much\u00a0win.But in doing so, the company commodi\u00adtizes its own output. If your company can auto\u00admate its output via AI, others can too. Your existing competi\u00adtors, certainly. But also upstarts who don\u2019t have your cost foot\u00adprint. What\u00adever intan\u00adgible assets AI can generate will be produced\u00a0in excess, leading to a defla\u00adtionary market for that asset. A company\u2019s knowl\u00adedge workers may be its greatest expense. But they also contribute to its compet\u00adi\u00adtive\u00a0moat.Further\u00admore, the ongoing market value of a company is neces\u00adsarily tied to its differ\u00aden\u00adti\u00adated assets. A company valued at a billion dollars that adopts AI will then have to build some new asset not replic\u00adable with AI, or it won\u2019t be worth a billion much longer. Every\u00adthing that makes AI an excel\u00adlent cost collapser makes it an equally excel\u00adlent capital collapser.\u00a0A poisoned chalice. The greatest irony will be if AI-adopting compa\u00adnies\u2014having laid off knowl\u00adedge workers en masse\u2014hire them back merely to ensure that competi\u00adtors cannot. In recent tech history, there have already been signs of this dynamic.Consider large law firms, aka Big Law. Currently certain legal-AI star\u00adtups license LLMs from Big AI and repackage them for Big Law at high prices. These star\u00adtups claim to add other special sauce. OK, sure. Where\u2019s the economic equi\u00adlib\u00adrium? If legal-AI star\u00adtups prove that money can be made selling AI to Big Law\u2014won\u2019t Big AI just sell to Big Law directly, and cut out the star\u00adtups? Or if legal-AI star\u00adtups prove that AI can effec\u00adtively provide legal services\u2014won\u2019t legal-AI star\u00adtups just sell to clients directly, and cut out Big Law? Won\u2019t members of Big Law that adopt AI have to lay off a lot of equity part\u00adners, because adop\u00adtion of AI will shrink profit margins? Won\u2019t the members of Big Law refusing AI have to\u00a0consol\u00adi\u00addate to preserve their margins? Or just cave to AI? (The tendency of competi\u00adtors to adopt similar prac\u00adtices is called insti\u00adtu\u00adtional isomor\u00adphism.) So it goes. Most states prevent nonlawyers from sharing in legal fees, so law firms will prob\u00adably remain a distinct set of enti\u00adties. But one plau\u00adsible equi\u00adlib\u00adrium is that legal-AI star\u00adtups disap\u00adpear (quickly), and members of Big Law consol\u00adi\u00addate (rela\u00adtively quickly) until there are only a handful left, all contracting directly with Big\u00a0AI.Along these lines, I expect that to succeed finan\u00adcially, Big AI will likely need to demolish a signif\u00adi\u00adcant number of existing tech compa\u00adnies and grab their revenue for itself. By the process described above: Big AI essen\u00adtially uses its tech customers as an R&D facility. Big AI licenses models to these compa\u00adnies. Tech compa\u00adnies compete to adapt their busi\u00adnesses to AI. Once a concept is proven, Big AI directly takes over that market. The labor-replace\u00adment story will grow into a company-replace\u00adment story. Many of those tech compa\u00adnies\u2014and their share\u00adholders in the public markets\u2014may also find that AI is a poisoned chalice.The resource curseThe theory is thought also to fit Middle Eastern oil exporters espe\u00adcially well. In this region, govern\u00adments\u2019 access to rents, in the form of oil revenue, may have freed them from the need for taxa\u00adtion of their peoples, and that this in turn freed them from the need for democ\u00adracy.\u2014Jeffrey FrankelSince AI is new, there are not yet exam\u00adples of post-AI economies. So we might ask a broader ques\u00adtion: what happens to nations whose economies are concen\u00adtrated around one inher\u00adently polit\u00adical tech\u00adnology? The most instruc\u00adtive compa\u00adra\u00adbles in the world economy are prob\u00adably petrostates: nations whose national wealth is concen\u00adtrated in oil or natural gas. The analogy is imper\u00adfect. Unlike AI, oil is anchored to geog\u00adraphy and limited in supply. On the other hand, oil has proven finan\u00adcial value; AI does\u00a0not.Petrostates are vulner\u00adable to what econ\u00ado\u00admists call the resource curse, with two distinc\u00adtive sets of effects:Capital effects. The concen\u00adtrated resource tends to pull labor and capital from the rest of the economy, depriving other economic sectors, espe\u00adcially manu\u00adfac\u00adturing and agri\u00adcul\u00adture. This makes the national economy more depen\u00addent on imports for basic needs, and more sensi\u00adtive to price volatility of the resource. If the resource price declines suffi\u00adciently, the nation will be unable to keep affording imports, but also unable to fulfill those needs with domestic produc\u00adtion.Polit\u00adical effects. The value of the concen\u00adtrated resource creates what Jeffrey Frankel calls \u201ca polit\u00adical contest to capture owner\u00adship\u201d, which in turn encour\u00adages the emer\u00adgence of auto\u00adcratic or oligarchic insti\u00adtu\u00adtions captured by an economic elite who seek to retain control of the resource. The process is self-rein\u00adforcing in two ways. First: the economic elite uses its wealth to repress polit\u00adical oppo\u00adnents. Second: as the govern\u00adment derives more income from the concen\u00adtrated resource, it relies less on taxa\u00adtion of citi\u00adzens, which weakens demo\u00adc\u00adratic account\u00adability.According to the Council on Foreign Rela\u00adtions, \u201cpetrostates include Algeria, Cameroon, Chad, Ecuador, Indonesia, Iran, Kaza\u00adkhstan, Libya, Mexico, Nigeria, Oman, Qatar, Russia, Saudi Arabia, the United Arab Emirates, and Venezuela.\u201d No one would confuse this with a list of successful liberal democ\u00adra\u00adcies. Due to its increasing oil and gas produc\u00adtion, some also consider the US a petrostate.Which petrostates provide the most goodies to citi\u00adzens? The Gulf petrostates\u2014Qatar, UAE, Saudi Arabia, Oman. These nations distribute petrowealth to citi\u00adzens through substan\u00adtial public bene\u00adfits (e.g., housing, health\u00adcare, educa\u00adtion, public-sector jobs) while having low rates of taxa\u00adtion. The tradeoff, of course, is that these coun\u00adtries are not even notion\u00adally demo\u00adc\u00adratic. They are hered\u00adi\u00adtary monar\u00adchies. They offer none of the bene\u00adfits that citi\u00adzens in a liberal democ\u00adracy enjoy (or some\u00adtimes take for granted)\u2014demo\u00adc\u00adratic partic\u00adi\u00adpa\u00adtion, indi\u00advidual civil rights, freedom of speech.But the Gulf states were never liberal democ\u00adra\u00adcies. Can the resource curse send a liberal democ\u00adracy spiraling into autoc\u00adracy? Sure\u2014consider Venezuela. Today, Venezuela is an author\u00adi\u00adtarian regime. But from 1958 to the 1990s, it was a stable liberal democ\u00adracy. This changed with the 1998 elec\u00adtion of Hugo Ch\u00e1vez. After taking office, Ch\u00e1vez pursued populist economic poli\u00adcies funded by national oil wealth. Embold\u00adened by high oil prices, Ch\u00e1vez amended the consti\u00adtu\u00adtion to bolster his own power and installed loyal\u00adists at the top of Venezuela\u2019s national oil company (PDVSA). He redi\u00adrected oil revenues to a series of domestic social programs (the \u201cBoli\u00advarian missions\u201d). Ch\u00e1vez lever\u00adaged his resulting popu\u00adlarity among workers to consol\u00adi\u00addate his power, for instance by ending term limits and taking control of the judi\u00adciary. (Confi\u00adden\u00adtial to AI CEOs: think care\u00adfully before you offer an owner\u00adship stake to any US pres\u00adi\u00addent who has author\u00adi\u00adtarian and populist tenden\u00adcies.) After oil prices dropped in 2014, the resource curse hit Venezuela\u00a0hard, putting its economy into crisis and allowing Ch\u00e1vez\u2019s successor Nicol\u00e1s Maduro to estab\u00adlish a dicta\u00adtor\u00adship. Demon\u00adstrating a central irony of the resource curse: while the resource is gener\u00adating wealth and deliv\u00adering goodies to citi\u00adzens, author\u00adi\u00adtarian moves are depicted as essen\u00adtial to sustaining pros\u00adperity. After the economy weakens, author\u00adi\u00adtarian moves are depicted as essen\u00adtial to restoring pros\u00adperity. The ratchet only tightens. The river flows only down\u00adhill.Norway is another outlier. Norway is a robust liberal democ\u00adracy. After oil was discov\u00adered in its North Sea terri\u00adto\u00adrial waters in the 1960s, Norway created the state-owned Statoil to build and operate oil facil\u00adi\u00adties. In 1990, Norway estab\u00adlished the Govern\u00adment Pension Fund Global, a sover\u00adeign-wealth fund to receive surplus oil revenues and invest them for the benefit of Norwe\u00adgian citi\u00adzens. According to Norges Bank, \u201c[t]he aim of the fund is to ensure that we use this money respon\u00adsibly, think long-term and so safe\u00adguard the future of the Norwe\u00adgian economy.\u201d In one sense it has worked: the GPFG currently holds over $2 tril\u00adlion in assets. What\u2019s the problem? Even Norway hasn\u2019t entirely avoided the resource curse. Rising sover\u00adeign wealth has deliv\u00adered tons of goodies to Norwe\u00adgians. But it has also made Norway\u2019s govern\u00adment less finan\u00adcially disci\u00adplined and created social conse\u00adquences like high dropout and unem\u00adploy\u00adment rates. (A popular 2025 Norwe\u00adgian book was called The Country That Got Too Rich.) Keep in mind that Norway reached this predica\u00adment despite 50 years of careful prepa\u00adra\u00adtion and over\u00adsight. Inher\u00adently polit\u00adical tech\u00adnolo\u00adgies can upend even gener\u00ada\u00adtional prepa\u00adra\u00adtions. Those who would suggest that the US can follow a similar model with AI: easier said than\u00a0done.Extinction-level capitalismPutting it all together: Among AI risks, we should take more seri\u00adously the poten\u00adtial conse\u00adquences of AI working as intended. AI is a capi\u00adtalist instru\u00adment. Its prin\u00adcipal func\u00adtion is to concen\u00adtrate capital. Its intended mech\u00ada\u00adnism is large-scale labor replace\u00adment. But it is also inher\u00adently polit\u00adical tech\u00adnology. As AI makes it harder for workers to capture value from their labor, they will increas\u00adingly have to rely on goodies from Big AI, priva\u00adtizing what were once func\u00adtions of govern\u00adment. If Big AI subsumes the func\u00adtions of workers and govern\u00adment, both will tend to realign polit\u00adi\u00adcally around Big AI\u2019s inter\u00adests. What\u00adever term describes this system, it is not liberal democ\u00adracy as US citi\u00adzens have tradi\u00adtion\u00adally under\u00adstood it. AI-centered capi\u00adtalism risks an extinc\u00adtion of demo\u00adc\u00adratic possi\u00adbility. It will be America. But it will no longer be Amer\u00adican.EpilogueThank you for reading.Every citizen has a voice in these issues. Every citizen can partic\u00adi\u00adpate in the vital public conver\u00adsa\u00adtion about how we want AI to be part of our country\u2014our schools, our work, our fami\u00adlies.Joseph Weizen\u00adbaum: \u201cThe myth of tech\u00adno\u00adlog\u00adical and polit\u00adical and social inevitability is a powerful tran\u00adquil\u00adizer of the conscience.\u201d\u00a0In the upcoming elec\u00adtion, vote for people who take your rights and inter\u00adests seri\u00adously.I support orga\u00adnized labor.Ulti\u00admately, Big AI is constrained by an incon\u00adve\u00adnient truth. Today, they need us more than we need them. As long as that remains true, we the people have the upper\u00a0hand." ], "output": "readability/", "pwd": "/data/archive/1781518242.322736", "schema": "ArchiveResult", "start_ts": "2026-06-15T10:11:19.799283+00:00", "status": "succeeded" } ], "screenshot": [], "singlefile": [], "title": [ { "cmd": [ "/usr/bin/curl", "--silent", "--location", "--compressed", "--proxy", "socks5://tor-socks-proxy:9150", "--max-time", "60", "--user-agent", "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/128.0.0.0 Safari/537.36 ArchiveBox/{VERSION} (+https://github.com/ArchiveBox/ArchiveBox/)", "https://matthewbutterick.com/extinction-level-capitalism.html" ], "cmd_version": "8.10.1", "end_ts": "2026-06-15T10:11:19.079830+00:00", "index_texts": null, "output": "Matthew Butterick | Extinction-level capitalism", "pwd": "/data/archive/1781518242.322736", "schema": "ArchiveResult", "start_ts": "2026-06-15T10:11:19.042957+00:00", "status": "succeeded" } ], "wget": [] }, "icons": null, "is_archived": true, "is_static": false, "latest": { "archive_org": "https://web.archive.org/web/20260615101221/https://matthewbutterick.com/extinction-level-capitalism.html", "dom": "output.html", "favicon": "favicon.ico", "git": null, "media": "media/", "pdf": null, "screenshot": null, "singlefile": null, "title": "Matthew Butterick | Extinction-level capitalism", "warc": null, "wget": null }, "link_dir": "/data/archive/1781518242.322736", "newest_archive_date": "2026-06-15T10:11:45.842835+00:00", "num_failures": 0, "num_outputs": 9, "oldest_archive_date": "2026-06-15T10:10:50.355775+00:00", "path": "/extinction-level-capitalism.html", "schema": "Link", "scheme": "https", "snapshot_abid": "snp_01KV5C64HZ7C2BBB0101RGADM5", "snapshot_id": "2e3d1056-8812-47b2-b07e-b67d31053685", "sources": [ "/data/sources/1781518241-import.txt" ], "tags": null, "tags_str": "", "timestamp": "1781518242.322736", "title": "Matthew Butterick | Extinction-level capitalism", "url": "https://matthewbutterick.com/extinction-level-capitalism.html" }