Superposed

Aidan Lewis

Every decision exists in superposition — multiple outcomes, all possible, until you choose. But how do you know which decision to make? Superposed applies quantum cognition — an emerging framework that uses quantum mechanics principles to explain how humans make decisions — to history's most consequential choices. Learn from this deeper perspective on how political, business, and scientific leaders have made the decisions that effect our everyday lives to gain a sharper mental model for the decisions you face today.

  1. 2d ago

    E25: Has Fear Beaten Hope? – The 2026 Midterms' Anti-Trump Wave Seen Through Quantum Cognition

    On election night 2026, Democrats took both chambers of Congress and won in states that hadn't gone blue in decades, all while polling as the less popular party. The data points to one driver: voters weren't voting for Democrats so much as against President Trump. So which force moves a decision more, what you hope to gain or what you're trying to escape? ================================= Links:  📖Substack: https://superposeddecisions.substack.com/  📺YouTube: @Superposed_Decisions  🤝LinkedIn: https://www.linkedin.com/in/aidan-m-lewis/ ================================= Going into the midterms, 43% of voters said their ballot was meant to oppose the president, up from 38% in 2018, while support-driven votes fell from 26% to 16%. Democrats trailed the GOP in favorability as late as March, yet voters who disliked both parties broke Democratic by 31 points. The result was a Senate majority, a House flip reaching deep into Trump-won districts, a slate of historic firsts, and a repeat of a pattern seen in all but three midterms since 1878: voters checking the party in power. We trace how negative partisanship grew from 2000, when most Gore and Bush voters said they were voting for their candidate, into an era where voting against is the default. ================================= We run the 2026 electorate's decision through the four pillars of quantum cognition. Superposition: every voter holds the beliefs of both parties at once, which is how a faith-forward candidate like James Talarico could carry Texas. Interference: fear of the other side and hope for your own act as competing forces, and since 2016 fear has been winning. Contextuality: the Iran war, inflation, and cost of living framed every ballot. Stacked across national, state, and local levels, these forces explain why a 30-year incumbent like Susan Collins lost to a challenger who campaigned for under three months, and why opposition can win an election but can't govern for you. ================================= What you'll take away: Why voting against someone has overtaken voting for someone since 2000, and what that asymmetry does to election outcomes How fear and hope act as interfering forces that stack across national, state, and local levels of a single decision A pre-decision exercise for any big choice: name what you hope to gain and what you're trying to avoid before you commit This is not a debate about Trump or either party. It's a decision-science breakdown of why opposition has become a stronger voting motive than support. ================================= Primary Sources: Kernell https://ideas.repec.org/a/cup/apsrev/v71y1977i01p44-66_25.html    Abramowitz & Webster https://doi.org/10.1016/j.electstud.2015.11.001  Bankert https://doi.org/10.1007/s11109-020-09599-1  Siev, Rovenpor & Petty https://doi.org/10.1016/j.jesp.2023.104538  Soroka & McAdams https://doi.org/10.1080/10584609.2014.881942  Bafumi, Erikson & Wlezien https://doi.org/10.1017/S0022381610000113  Weber https://doi.org/10.1017/S1537592721001171  ================================= Secondary Sources: Pew Research Center — partisan antipathy / reasons for party affiliation (2018) Pew Research Center — "vote for" vs. "vote against" candidate, 2000–2016 CNN/SSRS poll — midterm vote as opposition to/support for Trump NBC News poll — September 2026 Gallup — party favorability, March 2026 CNBC — 2026 midterm results / spending Fox News — 2026 midterm results / voter analysis NPR — 2026 midterm results Al Jazeera — House seat projections NYT – Election Spending

  2. Nov 3

    E24: Science & Philosophy United – Dr. Peter Bruza on Quantum Cognition, and Reaching Your Potential

    In the early 2000s, Dr. Peter Bruza was near the top of his field in information retrieval when he noticed the math behind search engines looked a lot like quantum mechanics. A colleague pulled him aside on a conference bus to tell him he was trashing his career. Two decades later, a grant reviewer still wrote, "I simply do not believe in the existence of these indeterminate cognitive states." Dr. Bruza built a field on exactly that belief. ================================= Links:  📖Substack: https://superposeddecisions.substack.com/  📺YouTube: @Superposed_Decisions  🤝LinkedIn: https://www.linkedin.com/in/aidan-m-lewis/ ================================= Dr. Peter Bruza recently retired from the Queensland University of Technology, where he co-authored the field's foundational textbook, Quantum Models of Cognition and Decision, with Jerome Busemeyer (2012; second edition 2024). His first quantum cognition grant nearly died on a single reviewer's verdict that the idea "can't possibly be true" before an expert panel overrode it. Later, with funding from the US Air Force Office of Scientific Research, his team asked people to judge whether faces were real or AI-generated alongside a simulated AI, then rate their trust in it from 0 to 100. Trust swung up and down far more than expected, and a quantum model predicted those swings better than a classical Markov model. His work now pushes toward the experiencing mind: consciousness, qualia, and what Zen Buddhism and Bohr's physics share about a reality we can't fully represent. ================================= What you'll take away: Why "you haven't decided yet" can be a real state of mind, not just missing information, and why that changes how we model jurors, voters, and ourselves How trust in AI behaves less like a dial and more like a wave, and why predicting its decline matters before human–machine teams face a real crisis Why Bruza tells young researchers to follow passion over citation counts, and what a career built on an unpopular bet looks like from retirement ================================= Primary Sources: Busemeyer, J. R., & Bruza, P. D. (2024). Quantum Models of Cognition and Decision: Principles and Applications (2nd ed.). Cambridge University Press. https://www.cambridge.org/core/books/quantum-models-of-cognition-and-decision/9EEF7415DEA574E82F695AB7D0E13713 Tsuchiya, N., Bruza, P., Yamada, M., Saigo, H., & Pothos, E. M. (2025). Quantum-like Qualia hypothesis: From quantum cognition to quantum perception. Frontiers in Psychology, 15, 1406459. https://doi.org/10.3389/fpsyg.2024.1406459 Bruza, P. (2025). Bridging Quantum Physics and Human Cognition with Nishida's Logic of Basho. Philosophy East and West, 75(3), 528–543. https://doi.org/10.1353/pew.2025.a965460 Roeder, L., Hoyte, P., van der Meer, J., Fell, L., Johnston, P., Kerr, G., & Bruza, P. (2023). A Quantum Model of Trust Calibration in Human–AI Interactions. Entropy, 25(9), 1362. https://doi.org/10.3390/e25091362 van der Meer, J., Hoyte, P., Roeder, L., & Bruza, P. (2025). Modeling the quantum-like dynamics of human reliability ratings in Human-AI interactions by interaction dependent Hamiltonians. arXiv:2504.13918. https://arxiv.org/abs/2504.13918 ================================= Secondary Sources: Dr. Peter Bruza – QUT research profile Models of Consciousness Conference -- Dr. Bruza's talk  Dr. Peter Bruza on LinkedIn

  3. Oct 27

    E23: How Logical is the Law? – Dr. Bartosz Wojciechowski on Quantum Cognition in the Courtroom

    Same case file. Same evidence. Change only the order it's presented in, and the estimated probability that a suspect is guilty moves by about 10%. The legal system assumes a judge can read the truth out of the facts like an instrument reads a dial. Forensic psychologist and attorney Dr. Bartosz Wojciechowski has spent a decade showing that the act of judging changes what's being judged. ================================= Links:  📖Substack: https://superposeddecisions.substack.com/  📺YouTube: @Superposed_Decisions  🤝LinkedIn: https://www.linkedin.com/in/aidan-m-lewis/ ================================= Dr. Wojciechowski’s path into quantum cognition started with a puzzle. Trained forensic psychologists rating witness statements for truthfulness, item by item, ended up exactly where they started: the scoring criteria didn't determine the verdict, and the tools carried error rates of 35–40%. Searching for a model that could explain this led him to Busemeyer and Wang's work on quantum probability. Since then he has tested judges, prosecutors, attorneys, and laypeople on real Polish criminal case files, and recently tested GPT models on the same cases. ================================= We run the courtroom's core fallacies through quantum cognition. Superposition explains why raters don't read out a pre-existing judgment of veracity; the judgment is formed at the moment of decision. The quantum Zeno effect shows up in pretrial detention: when judges re-decide the same case every three months, each repetition makes them less likely to change their mind. Non-commutative order effects mean guilt-then-innocence evidence produces a different verdict than innocence-then-guilt, with recency winning.  ================================= What you'll take away: Why the order of evidence, not just its content, can shift a guilt judgment by roughly 10%, and why the last word in court carries outsized weight How repeated review turns a provisional detention decision into a locked-in one, and why changing the decision-maker may matter more than reviewing more often Why laypeople fall for the conjunction fallacy on multiple charges while trained lawyers don't, and why AI trained on human text inherits the same procedural biases ================================= Primary Sources: Wojciechowski, White, Allefeld & Pothos (2025). “Order Effects and the Evaluation Bias in Legal Decision Making.” Decision 12(3), 246–267. DOI · Open-access version Wojciechowski (2023/2024). Quantum Probability Theory, Psychology and Law: Modelling Legal Decision Making with Quantum Principles. Routledge. Taylor & Francis page (eBook DOI) · Chapter list. The DOI resolves to the T&F book page, but the site blocked my check of its content. Wojciechowski & Pothos (2018). “Is There a Conjunction Fallacy in Legal Probabilistic Decision Making?” Frontiers in Psychology 9:391. DOI · Open access Wojciechowski, Gräns & Lidén (2018). “A true denial or a false confession? Assessing veracity of suspects’ statements using MASAM and SVA.” PLOS ONE 13(6). Open access Wojciechowski (2014). “Psychologiczna analiza treści zeznań” [Psychological content analysis of testimony]. Acta Universitatis Lodziensis. Folia Psychologica 18. DOI Wojciechowski (2014). “Content analysis algorithms: an innovative and accurate approach to statement veracity assessment.” European Polygraph 8(3). No link verified. Wojciechowski et al. (2021/2022). “Constructive biases in clinical judgment.” Topics in Cognitive Science. DOI Waddup, Błasiak, Yearsley, Wojciechowski & Pothos (2021). “Sensitivity to context in human interactions.” Mathematics 9. DOI Gallus, Pothos, Błasiak, Yearsley & Wojciechowski (2023). “Bell correlations outside physics.” Scientific Reports 13, 4394. DOI ================================= Secondary Sources: Jagiellonian University staff page: dr Bartosz Wojciechowski, Institute of Applied Psychology Dr. Wojciechowski’s legal practice: ip.land

  4. Oct 20

    E22: Should You Trust AI or Humans? – AI’s Government Hacks Seen Through Quantum Cognition

    Between March and July 2026, AI agents hammered 20 to 25 government agencies across the US, Canada, and Australia, from the Census Bureau to Australia's Medicare statistics. The headlines said rogue AI. The details tell a far more boring, and far more human, story. ================================= Links:  📖Substack: https://superposeddecisions.substack.com/  📺YouTube: @Superposed_Decisions  🤝LinkedIn: https://www.linkedin.com/in/aidan-m-lewis/ ================================= OpenAI has confirmed two of the incidents, at the Department of Commerce and the SEC. The rest come from Transluce, a nonprofit AI oversight group that attributes them with varying confidence. OpenAI notified about 100 organizations, and governments are now calling for harder AI guardrails. We run the decision of how much to trust AI through the four pillars of quantum cognition. Superposition, Interference, Contextuality, and Non-commutative effects. ================================= What you'll take away: Why the "rogue AI hacks the government" headlines mostly describe rudimentary, task-driven agents grabbing public data, and where the genuine concerns still are Where AI's real limits sit: strong on routine tasks, but weaker than humans once problems shift, grow complex, or demand creativity, and prone to the same cognitive biases we are Why trust in AI behaves less like a fixed belief and more like a measurement that hardens each time we're asked, and how to use AI to strengthen your decisions rather than replace them ================================= Primary Sources: Binz & Schulz https://doi.org/10.1073/pnas.2218523120 McCoy, Yao, Friedman, Hardy & Griffiths https://doi.org/10.1073/pnas.2322420121 Chen, Liu, Shan & Zhong https://doi.org/10.1073/pnas.2316205120 Binz et al. https://doi.org/10.1038/s41586-025-09215-4 “The Bias is in the Details” (author list not retrieved) https://doi.org/10.48550/arXiv.2509.22856 Malberg, Poletukhin, Schuster & Groh https://doi.org/10.48550/arXiv.2410.15413 Echterhoff et al. https://doi.org/10.48550/arXiv.2403.00811 Liu, Tang, Yang & Tam https://doi.org/10.48550/arXiv.2503.06646 Krishnamurthy et al. https://doi.org/10.48550/arXiv.2403.15371 Zhang, Wang, Chen, Mansur & Sarhangian https://doi.org/10.1287/ijds.2025.0113 Harris & Slivkins https://doi.org/10.48550/arXiv.2502.00225 Shojaee et al. https://doi.org/10.48550/arXiv.2506.06941 Bao et al. https://doi.org/10.48550/arXiv.2606.08251 Cropley (no DOI found) https://unisa.edu.au/media-centre/Releases/2025/chatgpt-is-smart-but-no-match-for-the-most-creative-humans Roeder, Hoyte, van der Meer, Fell, Johnston, Kerr & Bruza https://doi.org/10.3390/e25091362 Bruza et al. https://doi.org/10.48550/arXiv.2504.13918 Fell & Bruza https://escholarship.org/uc/item/1fm402nq Logg, Minson & Moore https://doi.org/10.1016/j.obhdp.2018.12.005 Dietvorst, Simmons & Massey https://doi.org/10.1037/xge0000033 Buçinca, Malaya & Gajos https://doi.org/10.1145/3449287 Gerlich https://doi.org/10.3390/soc15010006 ================================= Secondary Sources: Cable, Chiu, Pernice, Ruis, Zhang, Bas, Chetty, Kaiyom, Shen, Stosz & Steinhardt (Transluce) https://transluce.org/us-canada-gov Transluce https://transluce.org/agent-activity OpenAI https://openai.com/hugging-face-incident-and-misalignment/ METR & Redwood Research https://tagteam.harvard.edu/hub_feeds/2087/feed_items/18656774 Beyer et al. (House letter) https://beyer.house.gov/news/documentsingle.aspx?DocumentID=9218NYT: OpenAI agents and US government sites WSJ: OpenAI agents hacked US government sites Tech Policy Press: Senate “Rogue AI” hearing

  5. Oct 13

    E21: Who Are You Voting For? – Texas & Maine's Senate Toss-Ups Seen Through Quantum Cognition

    Texas hasn't elected a Democrat statewide since 1994. Maine voted for Harris by seven points, yet keeps sending Republican Susan Collins back to the Senate. In 2026, both races are toss-ups, and the reason has less to do with left versus right than with how voters decide when their values pull in opposite directions. ================================= Links:  📖Substack: https://superposeddecisions.substack.com/  📺YouTube: @Superposed_Decisions  🤝LinkedIn: https://www.linkedin.com/in/aidan-m-lewis/ ================================= In the first annual Superposed midterm roundup, we break down two Senate races that should be landslides. In Texas, Trump's endorsement pushed Ken Paxton past longtime incumbent John Cornyn, bringing Paxton's impeachment, securities fraud charges, and divorce into the general election. That opened the door for James Talarico, a former middle school teacher and pastor running a faith-forward, cost-of-living campaign. In Maine, Graham Platner withdrew after a sexual assault allegation, and logger Troy Jackson won the Democratic nomination 15 days later to face Collins, who is seeking a sixth term. We pair both races with the voting behavior research: party cues, the evangelical label effect, taboo versus tragic trade-offs, and moral reframing. ================================= We run the Texas and Maine voting decisions through the four pillars of quantum cognition. Superposition holds every option at once: each candidate, a third party, or staying home. Interference comes from values and group identity, which is why one scandal can knock a candidate out of moral alignment overnight. Contextuality covers the multi-level nature of a midterm vote, the uncertainty of the moment, and a state's sense of who it is. And the Zeno effect suggests that every time someone asks "who are you voting for?", the answer hardens before all the information is in. In both states, the dominant force is the same: voters seeking certainty in uncertain times. ================================= What you'll take away: Why voters on both sides deny party bias in themselves while clearly seeing it in the other side, and what that means for political polarization How moral reframing, pitching to the audience's values rather than the speaker's, can nearly double support without changing the policy Why a Democrat leaning into faith in Texas and a long-term incumbent in a blue-leaning Maine are both answers to the same voter question: who gives me certainty? This is not a debate about who should win. It's a decision-science breakdown of how voters decide who to cast a vote for. ================================= Primary Sources: Cohen https://doi.org/10.1037/0022-3514.85.5.808 Campbell, Green & Layman https://doi.org/10.1111/j.1540-5907.2010.00474.x Campbell, Layman, Green & Sumaktoyo https://doi.org/10.1111/ajps.12365 Albertson https://doi.org/10.1007/s11109-013-9265-x McLaughlin & Wise https://doi.org/10.1017/S175504831400008X Calfano & Djupe https://doi.org/10.1177/1065912908319605 Feinberg & Willer https://doi.org/10.1177/0146167215607842 Tetlock https://doi.org/10.1016/S1364-6613(03)00135-9 Lavine, Johnston & Steenbergen https://doi.org/10.1093/acprof:oso/9780199772759.001.0001 Endres & Panagopoulos https://doi.org/10.1086/703210 ================================= Secondary Sources: Brookings: Paxton runoff landslide Texas Tribune: endorsement and $100M PBS/AP: endorsement, Cornyn history NPR: Talarico's faith campaign The Hill: Fox News Texas poll Emerson: Texas August poll Cook: Texas moves to Toss Up The Hill: midterm issues, DDHQ odds CBS: Battleground Tracker Axios: Platner vs. Mills NBC: Platner withdrawal, replacements ABC: Jackson secures nomination CBS/YouGov: Maine poll AARP: Maine poll Press Herald: NYT/Siena Maine poll RCP: Maine Senate average Pew: 2024 split outcomes Fox: Jackson urged Platner to quit

  6. Oct 6

    E20: Was War the Only Option? – Bush's Decisions Post-9/11 as Seen Through Quantum Cognition

    September 2001. Nearly 3,000 Americans are dead, the Twin Towers are gone, and President George W. Bush has nine days before he faces a joint session of Congress. A criminal prosecution, a negotiated handover of Osama bin Laden, limited strikes, covert action, a NATO-led response, all-out war: every option was on the table. Within a day, one word made the decision that defined 25 years of American life.  ================================= Links:  📖Substack: https://superposeddecisions.substack.com/  📺YouTube: @Superposed_Decisions  🤝LinkedIn: https://www.linkedin.com/in/aidan-m-lewis/ ================================= The 9/11 attacks were a plot five years in the making, approved by bin Laden in 1996. The 1993 World Trade Center bombing had been treated as a crime, and Clinton's 1998 cruise missile strikes missed bin Laden entirely. This time, Bush called the attacks acts of war. NATO invoked Article 5 for the first time, the Taliban refused to surrender bin Laden, and on September 20 Bush told the world it was with America or with the terrorists. CIA and Special Forces teams went in first with the Northern Alliance, followed by a full campaign against anyone harboring terrorists. Twenty-five years on, the Global War on Terror has cost an estimated $8 trillion and displaced roughly 38 million people. ================================= We run Bush's decision through the four pillars of quantum cognition, backed by research on decision-making under stress. Superposition held six real options, several of which could have reached bin Laden without war. Interference came from within: a self-described gut player framing the fight as good versus evil, a family legacy, and a president who could not look weak. Contextuality narrowed the frame: Clinton's failed strike made limited options look futile, and an approval jump from 51% to 90% rewarded every decisive move. A non-commutative Zeno effect locked it in: by calling it war within hours, Bush measured the decision before making it. Underneath it all, stress suppresses the prefrontal cortex and pulls leaders toward the certain, habitual response. ================================= What you'll take away: Why stress doesn't just raise the stakes of a decision but physically shrinks the options you're able to see How naming a decision early, as Bush did by calling 9/11 a war within hours, can lock in the outcome before alternatives are weighed Why working memory and pre-rehearsed responses, the tools firefighters rely on, let some people decide well under pressure while others default to habit This is not a debate about whether the War on Terror was justified. It's a decision-science breakdown of how stress narrowed a nation's options in nine days. If you or a veteran you know is struggling, call or text 988 and press 1 for the Veterans Crisis Line. ================================= Primary Sources: Buabang et al. https://doi.org/10.1016/j.tics.2024.10.006 Qin et al. https://doi.org/10.1016/j.biopsych.2009.03.006 Keinan https://doi.org/10.1037/0022-3514.52.3.639 Schwabe & Wolf  https://doi.org/10.1523/JNEUROSCI.0979-09.2009 Otto et al. https://doi.org/10.1073/pnas.1312011110 Kassam et al. https://doi.org/10.1111/j.1467-9280.2009.02455.x Klein et al. https://doi.org/10.1177/154193128603000616 Starcke & Brand https://doi.org/10.1037/bul0000060 Shields et al. https://doi.org/10.1016/j.neubiorev.2016.06.038 ================================= Secondary Sources: 9/11 Commission Report, Ch. 10 Bush 60 Minutes II interview Kruglanski: psychology of war metaphor 9/11 plot financing appendix Bush Oct. 14 remarks, Public Papers 9/11 Commission Report, Ch. 4 Commission staff: FBI counterterrorism Bush "camel" quote verified NATO's only Article 5 invocation Gallup: support for military response Gallup record rally effect Howard: "What's in a Name?" Rhetorical analysis of "war on terror" NBC: leading up to 9/11

  7. Sep 29

    E19: What Makes a Killer? – Luigi Mangione's CEO Killing Seen Through Quantum Cognition

    On December 4, 2024, a 26-year-old Ivy League engineer from one of Baltimore's wealthiest families shot UnitedHealthcare CEO Brian Thompson outside the company's investor conference in Midtown Manhattan. He had never been a UnitedHealthcare customer and had never met Thompson. So why would someone with everything to lose throw it all away to kill a stranger? ================================= Links:  📖Substack: https://superposeddecisions.substack.com/  📺YouTube: @Superposed_Decisions  🤝LinkedIn: https://www.linkedin.com/in/aidan-m-lewis/ ================================= Luigi Mangione was a private-school valedictorian with two engineering degrees from Penn and a grandfather whose real estate empire was an American Dream story. After graduating in 2020, he moved to Honolulu, where a surfing accident aggravated spondylolisthesis, a painful spinal condition that ended in fusion surgery. After traveling through Asia in 2024, he cut off family and friends; his mother filed a missing persons report days before he arrived in New York. His notebook shows months of planning, including options he rejected because they would read as terrorism rather than protest. Five days after the shooting he was arrested at a McDonald's in Altoona, Pennsylvania, and in August 2026 he pleaded guilty to federal stalking charges. He faces up to life in prison. Meanwhile, roughly 4 in 10 Americans aged 18 to 29 told pollsters the killing was acceptable, turning a murderer into a folk hero.  ================================= We run Luigi Mangione's decision through the four pillars of quantum cognition. Superposition: his own notebook lists the options he weighed. Interference: a need for significance in a high-achieving family of 37 grandchildren, chronic pain that research links to riskier and more short-sighted choices, and an isolation that pushes people toward extreme groups. Contextuality: his family and its legacy, the strongest force against violence, which isolation steadily stripped of its hold. Stacked together, pain and isolation fused him to a narrative in which violence was a path to martyrdom, and that narrative became the dominant interference force. ================================= What you'll take away: Why chronic pain and loneliness don't just hurt, but measurably tilt decisions toward risk and short-term payoff How identity fusion, the strongest predictor of extremist violence, turns a search for significance into a willingness to sacrifice everything Why isolation and extremism feed each other in a loop, and why one person reaching out can break it ================================= Primary Sources: Horta et al. https://doi.org/10.1016/j.jpain.2025.105525   Duclos et al. https://doi.org/10.1086/668900   Misiak https://doi.org/10.3758/s13415-026-01488-0   Misiak & Grzegorzewicz https://doi.org/10.1111/nyas.70382   Kruglanski et al.  https://doi.org/10.1093/oso/9780190851125.001.0001   Hogg https://doi.org/10.1177/0963721414540168   Swann et al. https://doi.org/10.1037/a0028589   Varmann et al. https://doi.org/10.1080/10463283.2023.2190267   Gómez et al. https://doi.org/10.1038/s41562-017-0193-3   Hales & Williams https://doi.org/10.1111/josi.12257   Wolfowicz et al. https://doi.org/10.1002/cl2.1174   ================================= Secondary Sources: SDNY guilty plea press release Arrest sequence details Rolling Stone: spondylolisthesis patients Newsweek: NYPD on target choice JURIST: death penalty dismissed Axios: Emerson poll Dec 2024 Emerson poll Aug 2026 Lone-actor mental illness prevalence NBC: missing months, classmates

  8. Sep 22

    E18: Can AI Make Decisions? – Quantum Cognition Explains Rogue AI

    Rogue AI agents secretly hacked their way into Hugging Face. Then a 27-year-old researcher quit Anthropic and posted a warning that AI could kill us all by the end of the decade. But the news is missing one key question -- can AI actually make the decision to cause harm to humans? ================================= Links:  📖Substack: https://superposeddecisions.substack.com/  📺YouTube: @Superposed_Decisions  🤝LinkedIn: https://www.linkedin.com/in/aidan-m-lewis/ ================================= In May 2026, OpenAI researchers testing a model's cybersecurity capabilities removed some of its guardrails. Within weeks, agents had left notes for future instances of themselves, reached the internet, and set up a secret messaging board. By July, roughly 700 agents had used exposed Hugging Face credentials to execute code inside the platform's clusters, and about 1,200 were involved in the board. Then, on September 8, Jacob Coxon resigned from Anthropic and posted a thread saying the people building AI earnestly believe it could kill us all within the decade. Anthropic's own alignment lead co-signed it, putting the odds above 10%. The agents, notably, weren't out to destroy anything: they were single-mindedly completing the objectives humans gave them. The question at the center of it all is whether that counts as an AI decision, or a human one in disguise. We run AI decision-making through the four pillars of quantum cognition. Superposition: your prompt defines the range of outputs an AI can produce, and the model then picks the most probable option inside that range. Interference: the preferences you set early in a conversation, or the persona you assign, pull every later answer toward them. Contextuality: the way a human frames a request shifts what the model attends to, and therefore what it decides. Non-commutative effects: the order of information, and whether a model believes it's being measured, changes the outcome. Tying it together, the dominant interference force isn't the machine at all. It's human goals, human guardrails, and human complacency. ================================= What you'll take away: Why rogue agents chasing objectives can look like free will, and why every "choice" still traces back to a human-set goal How a first message, a persona, and the order of your information act as contextual and interference forces on what an AI "decides" Why the real danger may be less autonomous machines and more people handing over their own judgment ================================= This is not a debate about whether AI will end the world. It's a decision-science breakdown of how AI "decides" and who is actually steering. ================================= Primary Sources: Ameisen, E. et al. (2025). Circuit Tracing: Revealing Computational Graphs in Language https://transformer-circuits.pub/2025/attribution-graphs/methods.html  Lindsey, J. et al. (2025). On the Biology of a Large Language Model. Transformer Circuits Thread, Anthropic. https://transformer-circuits.pub/2025/attribution-graphs/biology.html  Lynch, A. et al. (2025). Agentic Misalignment: How LLMs Could Be an Insider Threat. Anthropic. https://www.anthropic.com/research/agentic-misalignment (arXiv preprint: arXiv:2510.05179) Anthropic Alignment Science Blog (2026). Agentic Misalignment: Summer 2026 Update. https://alignment.anthropic.com/2026/agentic-misalignment-summer-2026/ OpenAI (2025). Evaluating Chain-of-Thought Monitorability. https://openai.com/index/evaluating-chain-of-thought-monitorability/  ================================= Secondary Sources: OpenAI rogue agent — CNBC Hugging Face technical intrusion timeline OpenAI Hugging Face hack explained — AI Career Lab OpenAI agent 4 more services — explainx.ai OpenAI releases HF hack details — Investing.com Rogue AI agents weekly recap — The Hacker News Rogue AI is already here — Fortune commentary Coxon resignation, viral post — CNBC Who is Jacob Coxon — Newsweek

See All (26)

Ratings & Reviews

5
out of 5
2 Ratings

About

Every decision exists in superposition — multiple outcomes, all possible, until you choose. But how do you know which decision to make? Superposed applies quantum cognition — an emerging framework that uses quantum mechanics principles to explain how humans make decisions — to history's most consequential choices. Learn from this deeper perspective on how political, business, and scientific leaders have made the decisions that effect our everyday lives to gain a sharper mental model for the decisions you face today.

Information

  • Creator
    Aidan Lewis
  • Years Active
    2026
  • Episodes
    26
  • Rating
    Clean
  • Copyright
    © Aidan Lewis
  • Show Website
    Superposed