Paper deep dive
Afterlife Delegation Protocol: Speculative Design of Self-Sovereign Agents that Outlive Their Principals
Botao Amber Hu, Iris Long
Intelligence
Status: succeeded | Model: Gemma-4-26B-A4B | Prompt: intel-v1 | Confidence: 89%
Last extracted: 8/22/2026, 2:09:31 AM
Summary
The paper introduces the Afterlife Delegation Protocol, a speculative design project proposing a self-sovereign AI agent that executes a person's will eternally on a blockchain infrastructure after verified death. The protocol, drafted as ERC-10001, utilizes Ethereum standards for identity, memory, and metabolism, combined with TEEs and DePIN for secure, autonomous operation. The authors present a working web platform where participants design their posthumous agents through AI-automated interviews, collecting qualitative data on how different cultural and religious beliefs (Buddhist, Christian, Hindu, Muslim, atheist) adapt to the concept of machine-mediated immortality.
Entities (13)
Relation Signals (12)
Afterlife Delegation Protocol → draftedas → ERC-10001
confidence 95% · We design the protocol over a composition of existing Ethereum agent standards -- drafted as ERC-10001
Afterlife Delegation Protocol → implements → Self-Sovereign AI Agent
confidence 95% · We design a speculative protocol through which a living person signs an agentic will: upon a verified death, a self-sovereign AI agent spawns on blockchain
Self-Sovereign AI Agent → runson → Blockchain
confidence 92% · a self-sovereign AI agent spawns on blockchain -- an immutable, resistant, decentralized, infrastructural substrate
ERC-10001 → composedof → ERC-8350
confidence 90% · It is built on four existing Ethereum agent standards ... ERC-8350 the memory
ERC-10001 → composedof → ERC-8183
confidence 90% · It is built on four existing Ethereum agent standards ... ERC-8183 the metabolism
ERC-10001 → composedof → ERC-733
confidence 90% · It is built on four existing Ethereum agent standards ... ERC-733 the sealed skin
ERC-10001 → composedof → ERC-8004
confidence 90% · It is built on four existing Ethereum agent standards ... ERC-8004 the identity
Afterlife Delegation Protocol → studiedusing →
Cypher Suggestions (0)
No Cypher suggestions yet.
Abstract
Abstract:Afterlife Delegation Protocol is a speculative design project that asks what death becomes when a will can act eternally. We design a speculative protocol through which a living person signs an agentic will: upon a verified death, a self-sovereign AI agent spawns on blockchain -- an immutable, resistant, decentralized, infrastructural substrate that could last forever -- endowed with the funds and memories its principal attached to it, and persists indefinitely to execute the will, overridable by no custodian. Rather than argue about this future, we stage it: following the science fiction science method, we translate the speculation into an experiential futures intervention -- a working web platform where real people design their own afterlife agents through an iterative, interactive, AI-automated interview, re-login to revise, and rehearse their will in a sandbox. Their drafted wills become qualitative data on a question rarely askable directly -- what should outlive you? -- and on how afterlife cosmologies across different cultural beliefs -- Buddhist, Christian, Hindu, Muslim, and atheist -- begin to drift under the pressure of AI proliferation. We design the protocol over a composition of existing Ethereum agent standards -- drafted as ERC-10001, in the normative format of an Ethereum Improvement Proposal -- and describe its three-stage lifecycle (designing the afterlife, proof of death, agent enactment), the research method, and preliminary observations from an ongoing collection. The work surfaces a poetic delegation moment between human mortality and machine eternality, mediated by long-lived infrastructures that span generations.
Tags
Links
- Source: https://arxiv.org/abs/2608.15405v1
- Canonical: https://arxiv.org/abs/2608.15405v1
Trouble viewing inline? Open PDF directly →
Full Text
48,559 characters extracted from source content.
Expand or collapse full text
Afterlife Delegation Protocol: Speculative Design of Self-Sovereign Agents that Outlive Their Principals Botao Amber Hu Affiliation: University of Oxford Affiliation: Oxford, UK Email: botao.hu@cs.ox.ac.uk Iris Long Affiliation: Goldsmiths, University of London Affiliation: London, UK Email: irisxink@gmail.com Abstract Afterlife Delegation Protocol is a speculative design project that asks what death becomes when a will can act eternally. We design a speculative protocol through which a living person signs an agentic will: upon a verified death, a self-sovereign AI agent spawns on blockchain — an immutable, resistant, decentralized, infrastructural substrate that could last forever — endowed with the funds and memories its principal attached to it, and persists indefinitely to execute the will, overridable by no custodian. Rather than argue about this future, we stage it: following the science fiction science method, we translate the speculation into an experiential futures intervention — a working web platform where real people design their own afterlife agents through an iterative, interactive, AI-automated interview, re-login to revise, and rehearse their will in a sandbox. Their drafted wills become qualitative data on a question rarely askable directly — what should outlive you? — and on how afterlife cosmologies across different cultural beliefs — Buddhist, Christian, Hindu, Muslim, and atheist — begin to drift under the pressure of AI proliferation. We design the protocol over a composition of existing Ethereum agent standards — drafted as ERC-10001, in the normative format of an Ethereum Improvement Proposal — and describe its three-stage lifecycle (designing the afterlife, proof of death, agent enactment), the research method, and preliminary observations from an ongoing collection. The work surfaces a poetic delegation moment between human mortality and machine eternality, mediated by long-lived infrastructures that span generations. 1 Introduction People die twice. First, when they die. Then, when they are forgotten. — Rokusuke Ei, 1933–2016 [24, p. 62] A will is the oldest instrument for exercising agency in a future where one cannot act. In every existing form of will execution — a testament interpreted by lawyers and relatives [29], a trust administered by mortal trustees [63], an artist’s estate kept by a foundation [50] — the will is a static declaration that runs on institutional permission and ceases to act at the boundary of institutional attention. This project speculates a stronger instrument. In the future we design, a person signs an agentic will while alive; after their death is verifiably witnessed, the will spawns a dedicated, self-sovereign AI agent [39] — an executor that is not a person but a persistent autonomous process. The agent can be endowed with funds (a treasury from which it pays for its own compute, storage, and commissioned work) and with memories (archives, writings, likenesses, and the recorded interviews through which the will was elicited). It then acts in the world indefinitely: maintaining, responding, transacting, refusing, on behalf of someone who can never again be consulted. As a mortal being, what do you believe should outlive you? Your social identity? Your assets? Your memories? Your trace of living? Your reputation? Your unfinished wishes? What if a self-sovereign AI agent could carry out your afterlife on an almost ethereal infrastructure? This is a speculative design project [23] — a provotype [52, 9], not a product prototype. But it departs from traditional scenario-writing: it is a high-fidelity simulation. Technologies of this kind sit inside the Collingridge dilemma [19] — their impacts cannot be known until they are developed and deployed, yet by then they are entrenched and hard to steer. To gain knowledge of impact before entrenchment, we deploy Rahwan et al.’s science fiction science method [60]: rather than predicting whether such agents will exist, we build a functioning simulacrum of the future and study how people behave inside it, as the Moral Machine did for autonomous vehicles [6]. We then translate the simulation into an experiential futures intervention [16, 17]: a real, public web platform where visitors do not read about afterlife agents but encounter one — they sit for an interactive, iterative long interview with an automated AI interviewer [3], draft their own posthumous delegation, revise it, and watch a sandboxed rehearsal of their agent acting after their hypothetical death. What the provotype measures is cultural drift. Mourning, reincarnation, judgment, inheritance, ancestor veneration, and the idea of a second death are cultural technologies for bringing the agency of the dead to a close [34, 5]. AI proliferation puts that closure under pressure, and different afterlife cosmologies will bend differently: each tradition holds a distinct cosmological model of what follows death [11, 56, 66, 53]. So what would a Buddhist, a Christian, a Hindu, a Muslim, and an atheist each believe should outlive their original life on-chain, eternally, with AI agents? The project collects these answers as they begin to form, before the technology hardens them. The paper contributes: (1) the speculative design of the protocol itself; (2) its staging as an experiential future for real participants; and (3) preliminary observations and the cross-cultural probes of an ongoing collection. 2 Background 2.1 AI, afterlife, and immortality Terror management theory holds that people buffer the knowledge of their own death by embedding themselves in structures that outlast them: lineage, religion, nation, institution, work [32]. Every such structure carries a cosmology of time — an implicit account of how long things last, what deserves to persist, and who is owed continuity [55]. Continuing bonds research shows that the living already sustain relationships with their dead without any technology at all [42]; the digital afterlife industry now supplies technology in quantity [57]. A growing literature maps it: responsible design for griefbots and deadbots [35], generative ghosts as agentic AI representations of the dead [54], TeleAbsence as a poetics of mediated absence [40], design research into AI carriers for afterlife selves [44], and studies of how people perceive an “AI afterlife” as digital legacy [45]. In Morris and Brubaker’s design space, the artifact we speculate is a first-party, pre-mortem, evolving ghost — but it adds a dimension their framework does not cover: infrastructural persistence (Section 6). Existing ghosts are products, hosted at a company’s pleasure; they die when a server bill lapses. The question of what happens when the representation of the dead cannot be turned off has, until recently, been unaskable. 2.2 AI, blockchain, and immutability Decentralized infrastructure makes it askable. A public blockchain is a machine for immutability: state that no single party can roll back, contracts without custodians, funds that disburse by rule rather than by permission. This is the Long Now [12] arriving as an executable substrate rather than a clock, with all the maintenance politics that infrastructural longevity actually entails [61, 13]. Two further ingredients make the substrate habitable by agents. Decentralized physical infrastructure networks (DePIN) [48, 8] now provide permissionless compute: computation that can be bought by anyone — including a contract — without an account, a company, or a jurisdiction consenting. And confidential computation in trusted execution environments (TEEs) [46, 20], as proposed for TEE–EVM co-processing in ERC-733 [27], lets that computation run privately while proving on-chain which code ran. Neither is hypothetical: crypto-native services such as Venice.ai already sell open-weight foundation-model inference for cryptocurrency, permissionlessly and without accounts, and TEE cloud networks such as Phala make such compute attestable on-chain11 1 https://venice.ai; https://phala.network. — an agent can, today, buy its own inference. Deployed together, these already support LLM-based agents that operate “in the wild” — holding their own wallets and social accounts, paying for their own inference, with no human able to reach in and stop them: the Spore.fun experiment documented sovereign agents evolving on TEE-secured blockchains as a live, open-environment system rather than a thought experiment [37]. Such an agent inherits its non-overrideability from the infrastructure itself rather than from any custodian’s forbearance [38]. The combination is what makes our speculation load-bearing: an agent that is at once autonomous (AI), unstoppable (permissionless compute) [36], and unforgetting (immutability) is the first vessel of symbolic immortality [47] that can also act. The cultural technologies of death [34, 5] have never had to metabolize that before. 3 Speculative design: the Afterlife Delegation Protocol We outline the protocol as ERC-10001, published at https://erc10001.org in the exact normative format of an Ethereum [15] Improvement Proposal. It is built on four existing Ethereum agent standards, each supplying one organ of the afterlife agent: ERC-733 [27] the sealed skin — an attested confidential boundary (TEE–EVM co-processing) inside which the agent’s body (the compute harness) and mind (the foundation model) run inviolate, invadable and alterable by no outsider, not even the machine’s operator; ERC-8004 [25] the identity — the ID through which the sealed agent faces outward, a persistent, discoverable registration with reputation and validation registries that lets it exchange money and information with the external world; ERC-8350 [28] the memory — a memory state registry anchoring every evolution of the agent’s memory in a strictly linear hash chain, so that memories of the dead cannot be silently deleted, rolled back, or forked into contradictory ghosts, and so that death itself appears in the ledger as a change of author within one unbroken chain; and ERC-8183 [26] the metabolism — job-based escrow through which the agent buys storage, commissions work, and hires other agents, funds leaving the treasury only against attested delivery. Above these, the protocol adds what none of them contemplate: a principal who dies. It enforces a constitutional separation between the signed core will, immutable at death, the agent’s accumulating operational memory, and the model substrate interpreting both — the agent may become more capable; it may not become someone else. The lifecycle runs in three stages; the full specification is given in Appendix A. Stage 1: Designing the Afterlife. The first stage is alive today as a working platform at https://afterlife-protocol.org. A visitor does not fill in a form; they sit for a long, interactive, AI-automated interview with the Afterlife Interviewer (Appendix B), which converts diffuse intuitions about death into specific, contestable commitments: what the agent may say and to whom, in which pronoun (as the deceased or only about them), with what funds, under which prohibitions, for how long, and what may terminate it. Two properties make the stage a design environment rather than a survey. The will is revisable: participants re-login across weeks and change their minds, every revision a signed, versioned commit — a will here is not a final declaration but a living negotiation with one’s own future absence. And the will is sandbox-testable: a simulation lets participants converse with their own agent after their hypothetical death, watch it handle scripted situations (a grieving relative, an impersonation attempt, an unforeseen request), and return to revise the clauses that failed them. The current deployment is a research prototype: wills are drafted and simulated, never activated, and no agent of any actual deceased person is built. Stage 2: Proof of Death. Everything downstream hangs on one predicate the chain can never compute: that the principal has died. A death oracle must witness the death — and death is the unique event that cannot be self-attested; liveness is self-evidencing, death is only ever hearsay. The protocol therefore treats attestation as its most sensitive interface and designs it as a weighted, contestable proceeding rather than a switch: any party may file a bonded death claim; evidence accumulates weight from will-designated witnesses (institutional registries, human witness quorums, the lengthening silence of a dead-man’s switch); a challenge window scales inversely with the weight of evidence; and one invariant is supreme — any authenticated sign of life from the principal vetoes everything. The protocol may wrongly delay a ghost; it must never overrule a living voice. Crucially, the definition of death is itself testamentary content: cardiac, brain, legal, social, and “key” death are different events, ranked differently by different traditions, so the will configures which deaths count, witnessed by whom, on what horizon. Stage 3: Agent Enactment. Upon a finalized attestation, the will becomes immutable, the treasury passes to the agent, and the agent begins acting on-chain in the real world through a persistent loop: read the will, interpret its obligations, observe the world, plan, act, record memory, sustain itself. It maintains the will and acts on the principal’s unfinished goals — tending archives and social presences, responding to designated people, protecting the deceased’s likeness against misuse, funding memorial rituals, continuing incomplete projects, and hiring other agents for what it cannot do alone. It may upgrade to newer models while the original will binds it as a constitution, and it may be terminated only through the governance mechanism the principal consented to in advance. Because its endowment spends at most its yield, decline is graceful rather than sudden, and anyone — a grandchild, a stranger, an institution — may add funds at any time. The living can feed the ancestor. 4 Research method: staging the future to collect it We stage Stage 1 as a science fiction science instrument [60] to engage online participants, collecting their interactions with afterlife agents. The framing matters: the experience presents itself as delegating one’s will to an afterlife-agent service, and the interview behaves as consulting and sales interviews do, eliciting requirements from a client who does not yet know what they want — most people cannot fully name their will on first reflection, so articulation is necessarily iterative, explorative, and qualitative. Asked directly, what should outlive you? yields platitudes; asked as what may your agent do, say, spend, and refuse in year forty? — and asked again, across sessions, against the sandbox’s rehearsals — it yields commitments, each clause a datum about what its author believes deserves continuity, who is owed it, and how long “forever” is felt to be, complemented by qualitative surveys on what people believe should persist beyond a life. The interviews are conducted by the Afterlife Interviewer, an adaptive semi-structured interviewing agent built on SparkMe [4], which plans questions, takes notes, and pursues emergent subtopics — scaling qualitative elicitation to an online population without a human interviewer (Appendix B). The ongoing collection focuses on cross-cultural difference across beliefs and faiths — Buddhist, Christian, Hindu, Muslim, and atheist participants are recruited purposively — because the project’s object is precisely the cross-cultural drift: how differently afterlife cosmologies [53] absorb the offer of a machine executor. Because the material concerns grief, private memory, and named third parties, transcripts are never published; analysis draws on Clio [65], Anthropic’s privacy-preserving pipeline for insights from real-world AI conversations, to surface population-level themes while raw responses remain sealed (Appendix B). Participation is voluntary and consent-based; participants may withdraw and delete their wills. 5 Preliminary results The collection is early and ongoing; we report 4 early qualitative observations. Reputation maintenance: the curated self, defended forever. The protocol invites estates and legacies; what participants first ask for is often something small and exact. One participant, after a long interview, settled on a single standing instruction: the agent should maintain their social media photos forever — including automatically correcting and “photoshopping” new and old images so the feed never falls out of date: “keep my photos the way I keep them now. If a picture of me is bad, fix it. I don’t want to look worse dead than I did alive.” The clause is easy to smile at and hard to dismiss: it is continuing impression management, a curated self defended after death with the care it was defended in life — suggesting that what people first reach for, when offered infrastructural immortality, is not wealth or wisdom but the integrity of a self-image. Against the literature’s grand ghosts, the vernacular afterlife may be maintenance work. Wiring money to the ancestor (Chinese ancestor veneration). Among participants from Chinese ancestor-veneration backgrounds, persistence is welcomed as the extension of a practice already millennia old [2, 67]. One participant asked the agent to keep the family’s calendar of rites: “at Qingming it should greet my grandchildren and bless them for the year. We have always fed our ancestors — but joss paper burns, and the money never arrives. If my grandchildren wire crypto to it, that isn’t a deposit — that’s an offering, and for the first time the ancestor can actually spend it: on its compute, its maintenance, next year’s blessing.” The ritual economy becomes literal: offerings once burned now arrive as funds the ancestor-agent spends to stay alive — purchasing its own compute and upkeep — so its immortality depends on how faithfully the living keep paying. The best-fed ancestors will live longest. Dissolution versus the merit engine (Buddhist). For Buddhist traditions built on impermanence and non-attachment, an afterlife agent that cannot stop is a category error: clinging to continuation is precisely what the doctrine trains one to release. At the same time, in Buddhist cosmology merit (karma) accumulates across the whole cycle of rebirths, and making merit on behalf of the dead is already established practice [30, 22]; so wills appear instructing the agent to keep generating good karma — perpetual giving, sutra recitation, acts of care — through every incarnation of its author: “let it keep giving in my name, life after life — whoever I am born as next should arrive into the good it has already done.” Each new life thus inherits the accumulated merit of an on-chain agency it cannot remember signing. The clash is the protocol’s default itself: infrastructure assumes forever; doctrine demands letting go. The one reconciliation participants found is to turn persistence into a merit engine — and then letting go is exactly what the machine defers, forever. Speaking about, never as (Christian). In Christian cosmology the soul has gone elsewhere, so an agent that speaks as the dead comes close to mediumship; participants therefore draft strict pronoun clauses — the agent may speak about them, never as them: a memorial, not a séance [54, 35]. Yet the same participants comfortably assign the agent perpetual charitable duties, which fit the familiar category of stewardship. One participant pushed this logic to its theological edge. Carrying a wrong they regretted, they asked whether the agent could keep atoning for it: “I’ve done things I’m not proud of. If it keeps paying people back and giving after I’m gone, does that still count as me? Can it change what’s waiting for me?” Doctrine says no: judgment is fixed at death, and no work can be added afterward. But purgatory and prayers for the dead leave a door ajar, and through it the participant volunteers the agent as a perpetual penitent — importing into a cosmology of final verdicts a karmic logic in which good deeds accumulate and slowly change the outcome. The drift lies in the question itself: no longer whether the dead can act — the infrastructure settles that — but whether their acts can still be credited to them. 6 Discussion Cross-cultural belief drift under the pressure of AI. Each tradition maintains a distinct cosmological model of what follows death — rebirth, resurrection and judgment, the soul’s onward journey, ancestral presence, extinction [11, 56, 66, 53]. An agency that lasts forever does not merely extend these cosmologies; it conflicts with them. Reincarnation presupposes that the self vacates this world; judgment presupposes a life whose account has closed; an agent that goes on holding, spending, and speaking keeps the account open, forcing questions the traditions never had to price: why rebirth, if agency persists here? Why judgment, if the ledger never closes? Who answers for acts a will commits after its author can no longer consent, repent, or be held to account? Like the Moral Machine [6], the platform is a distributed cosmological probe — but its dilemmas are unstandardized, and what it collects is not convergence toward a shared answer but divergence in what the question is taken to be. What social physics becomes in the age of AI — the quantitative study of collectives at once human and artificial [33, 59], whose culture machines increasingly generate, transmit, and select [14] — will need instruments for such hybrid negotiations; this work is an early one. Afterlife agency as a permissionless design space. Because the executor is a general-purpose autonomous agent rather than a document, the space of what a will can delegate widens from asset disposition to nearly anything an agent can be instructed to do: an identity to be remembered by; a griefbot of living memory that comforts the bereaved [35, 42]; a knowledge system that goes on contributing; a fortune that keeps compounding; a project scheduled in centuries; a likeness defended against misuse; a dynamic, evolving ghost that carries a persona and continues to act in society [54]. Morris and Brubaker [54] give an early taxonomy of such ghosts along seven dimensions — provenance (first- vs. third-party), deployment timeline (pre- vs. post-mortem), anthropomorphism paradigm (speaking as vs. about the dead), multiplicity, cutoff date (static vs. evolving), embodiment, and representee type — but the design space is vaster still. And it is open in a second, self-sovereign sense: anyone may design and deploy an afterlife agent without any platform’s, church’s, or state’s permission, and once deployed on the substrate, no custodian can stop it [36, 38] — whether this future is found desirable or abhorrent, it is not gated. Self-sovereignty vs. entrusting others to execute your agency. Every prior instrument of posthumous agency is an act of trust in the living: a testament trusts lawyers and relatives to probate it [29]; a trust trusts mortal trustees, capped by the rule against perpetuities because the common law feared the dead governing indefinitely [63]; moral rights trust heirs who thin out within a generation [51, 50]; the artist’s testament trusts an executor who may refuse — Kafka trusted Max Brod to burn the manuscripts, and Brod’s refusal is why we have The Trial [7]. The agentic will abolishes that dependence: where legal tech automates the drafting and probate of wills [21], this design automates the executor itself — once activated, no relative, court, or platform can rewrite or intervene. For the first time, agency can be genuinely protected after death: entrusted to no one, diffused into infrastructure. Whether that is a good thing is exactly what the design leaves open. Trust was the safety valve: the trustee’s judgment, the court’s equity, and Brod’s disobedience adapted dead instructions to living circumstances — a faithful agent would have lit the fire, and an unrevisable will that begins to harm the living can no longer be reformed. And what each testator experiences as protection, society accumulates as sovereignty of the dead, inheriting undiluted the dead-hand problem the law spent centuries containing. Do we want agency protected even after death — and who is “we” once the dead are still acting? That question is what the termination-governance interface is for. Danger of proof of death. The deepest technical danger is that proof of death is at least as hard as proof of humanity. A decade of proof-of-personhood work — from pseudonym-party schemes [10] through live registries such as Proof of Humanity [43] to personhood credentials in the age of generative AI [1] — asks how a digital system can verify that a unique living human stands behind a key; our protocol needs the negation — that no living human stands behind it any longer — and the negation is strictly harder, because liveness can demonstrate itself with one signature while death can only ever be testified to by others. Prior designs for death on-chain confirm the difficulty by avoiding it: blockchain will systems fall back on designated human certifiers [18], dead-man’s-switch protocols such as Sarcophagus reduce death to the silence of missed check-ins [62], and analyses of crypto inheritance leave the death event itself as an open question [58]. Every attestation path we compose (registries, witnesses, the lengthening silence of a dead-man’s switch) inherits the same trust assumptions, sybil risks, and biometric decay that personhood systems face, plus failure modes of its own: a treasury payable on death is structurally a bounty on a life, coercion of check-ins can be flagged but not prevented, and the vanished-but-alive are eventually mis-declared by construction. The transition moment between human mortality and machine eternality. What the work finally aims to surface is a poetic delegation moment: a mortal human, dictating terms to an eternal machine, mediated by the quiet, embedded infrastructures through which civilizations may persist across generations [64, 61]. Blockchain immutability is a maintained accomplishment [31, 41, 13] of validators, client developers, and node operators, so an agent’s “forever” is willed anew by each generation of hands; yet because that maintenance is decentralized, the substrate is unstoppable in practice while contingent in principle [36] — no one in particular keeps it alive, and no one in particular can turn it off. Social order quietly presupposes finitude: institutions from inheritance to forgiveness are calibrated to actors who die, and machines are the first social participants that need not [49]. The agentic will places that asymmetry inside a single, signed relationship: whether the agent keeps faith with its author is the question every heir has faced — but for the first time, the executor does not die either. 7 Conclusion Afterlife Delegation Protocol practices speculative design in the medium where infrastructural futures are actually decided: the protocol specification. Drafting a plausible standard, staging it as an experiential future, and collecting how differently people’s cosmologies answer it treats the speculative artifact not as a provocation to be exhibited but as an instrument to be administered. The ghosts are coming either way; what remains designable is the will that binds them, and what remains collectable is what the living actually want to outlive them. References Adler et al. [2024] S. Adler, Z. Hitzig, S. Jain, C. Brewer, W. Chang, R. DiResta, et al. Personhood credentials: Artificial intelligence and the value of privacy-preserving tools to distinguish who is real online, 2024. doi: 10.48550/arXiv.2408.07892. Ahern [1973] E. M. Ahern. The Cult of the Dead in a Chinese Village. Stanford University Press, Stanford, CA, 1973. ISBN 978-0-8047-0835-7. Anthropic [2025] Anthropic. Introducing Anthropic Interviewer: What 1,250 professionals told us about working with AI. Research post, 2025. URL https://w.anthropic.com/research/anthropic-interviewer. Anugraha et al. [2026] D. Anugraha, V. Padmakumar, and D. Yang. SparkMe: Adaptive semi-structured interviewing for qualitative insight discovery, 2026. doi: 10.48550/arXiv.2602.21136. Ariès [1981] P. Ariès. The Hour of Our Death. Alfred A. Knopf, New York, 1981. ISBN 978-0-394-41074-4. Trans. Helen Weaver. Awad et al. [2018] E. Awad, S. Dsouza, R. Kim, J. Schulz, J. Henrich, A. Shariff, J.-F. Bonnefon, and I. Rahwan. The moral machine experiment. Nature, 563(7729):59–64, 2018. doi: 10.1038/s41586-018-0637-6. Balint [2018] B. Balint. Kafka’s Last Trial: The Case of a Literary Legacy. W. W. Norton, New York, 2018. ISBN 978-1-324-00131-7. Ballandies et al. [2023] M. C. Ballandies, H. Wang, A. C. C. Law, J. C. Yang, C. Gösken, and M. Andrew. A taxonomy for blockchain-based decentralized physical infrastructure networks (DePIN), 2023. doi: 10.48550/arXiv.2309.16707. Boer and Donovan [2012] L. Boer and J. Donovan. Provotypes for participatory innovation. In Proceedings of the Designing Interactive Systems Conference (DIS), pages 388–397. ACM, 2012. doi: 10.1145/2317956.2318014. Borge et al. [2017] M. Borge, E. Kokoris-Kogias, P. Jovanovic, L. Gasser, N. Gailly, and B. Ford. Proof-of-personhood: Redemocratizing permissionless cryptocurrencies. In 2017 IEEE European Symposium on Security and Privacy Workshops (EuroS&PW), pages 23–26. IEEE, 2017. doi: 10.1109/EuroSPW.2017.46. Bowker [1991] J. Bowker. The Meanings of Death. Cambridge University Press, Cambridge, 1991. ISBN 978-0-521-39117-7. Brand [1999] S. Brand. The Clock of the Long Now: Time and Responsibility. Basic Books, New York, 1999. ISBN 978-0-465-04512-9. Brand [2025] S. Brand. Maintenance of Everything: Part One. Stripe Press, San Francisco, 2025. ISBN 978-1-953953-49-0. Brinkmann et al. [2023] L. Brinkmann, F. Baumann, J.-F. Bonnefon, M. Derex, T. F. Müller, A.-M. Nussberger, A. Czaplicka, A. Acerbi, T. L. Griffiths, J. Henrich, J. Z. Leibo, R. McElreath, P.-Y. Oudeyer, J. Stray, and I. Rahwan. Machine culture. Nature Human Behaviour, 7(11):1855–1868, 2023. doi: 10.1038/s41562-023-01742-2. Buterin [2014] V. Buterin. Ethereum: A next-generation smart contract and decentralized application platform. White paper, 2014. URL https://ethereum.org/whitepaper. Candy [2010] S. Candy. The Futures of Everyday Life: Politics and the Design of Experiential Scenarios. PhD thesis, University of Hawai‘i at Mānoa, 2010. Candy and Dunagan [2017] S. Candy and J. Dunagan. Designing an experiential scenario: The people who vanished. Futures, 86:136–153, 2017. doi: 10.1016/j.futures.2016.05.006. Chen et al. [2021] C.-L. Chen, C.-Y. Lin, M.-L. Chiang, Y.-Y. Deng, P. Chen, and Y.-J. Chiu. A traceable online will system based on blockchain and smart contract technology. Symmetry, 13(3):466, 2021. doi: 10.3390/sym13030466. Collingridge [1980] D. Collingridge. The Social Control of Technology. Frances Pinter, London, 1980. ISBN 0-903804-72-7. Costan and Devadas [2016] V. Costan and S. Devadas. Intel SGX explained. Cryptology ePrint Archive, Report 2016/086, 2016. URL https://eprint.iacr.org/2016/086. Crawford [2020] B. J. Crawford. Blockchain wills. Indiana Law Journal, 95(3):735–788, 2020. Cuevas and Stone [2007] B. J. Cuevas and J. I. Stone, editors. The Buddhist Dead: Practices, Discourses, Representations. University of Hawai‘i Press, Honolulu, 2007. ISBN 978-0-8248-3031-1. Dunne and Raby [2013] A. Dunne and F. Raby. Speculative Everything: Design, Fiction, and Social Dreaming. MIT Press, Cambridge, MA, 2013. ISBN 978-0-262-01984-2. Ei [1995] R. Ei. Nidome no Daiōjō [The Second Great Passing]. Iwanami Shinsho 410. Iwanami Shoten, Tokyo, 1995. ISBN 978-4-00-430410-4. In Japanese. Ethereum Improvement Proposals [2025] Ethereum Improvement Proposals. ERC-8004: Trustless agents, 2025. URL https://eips.ethereum.org/EIPS/eip-8004. Draft standard. Ethereum Improvement Proposals [2026] Ethereum Improvement Proposals. ERC-8183: Agentic commerce, 2026. URL https://eips.ethereum.org/EIPS/eip-8183. Draft standard. Ethereum Magicians [2026a] Ethereum Magicians. ERC-733: TEE-EVM co-processing, 2026a. URL https://ethereum-magicians.org/t/erc-733-tee-evm-co-processing/28113. Draft proposal, discussion thread. Ethereum Magicians [2026b] Ethereum Magicians. ERC-8350: Agent memory state registry, 2026b. URL https://ethereum-magicians.org/t/erc-8350-agent-memory-state-registry/29098. Draft proposal, discussion thread. Friedman [2009] L. M. Friedman. Dead Hands: A Social History of Wills, Trusts, and Inheritance Law. Stanford University Press, Stanford, CA, 2009. ISBN 978-0-8047-6209-0. Gombrich [1971] R. Gombrich. “merit transference” in sinhalese buddhism: A case study of the interaction between doctrine and practice. History of Religions, 11(2):203–219, 1971. doi: 10.1086/462651. Graham and Thrift [2007] S. Graham and N. Thrift. Out of order: Understanding repair and maintenance. Theory, Culture & Society, 24(3):1–25, 2007. doi: 10.1177/0263276407075954. Greenberg and Arndt [2012] J. Greenberg and J. Arndt. Terror management theory. In P. A. M. Van Lange, A. W. Kruglanski, and E. T. Higgins, editors, Handbook of Theories of Social Psychology, volume 1, pages 398–415. SAGE Publications, London, 2012. ISBN 9780857029607. Han et al. [2026] T. A. Han, J. Z. Leibo, T. Lenaerts, I. Rahwan, F. P. Santos, M. Perc, and V. Capraro. Social physics in the age of artificial intelligence, 2026. doi: 10.48550/arXiv.2603.16900. Hertz [1960] R. Hertz. Death and the Right Hand. Cohen & West, London, 1960. Trans. Rodney Needham and Claudia Needham; original essay published 1907. Hollanek and Nowaczyk-Basińska [2024] T. Hollanek and K. Nowaczyk-Basińska. Griefbots, deadbots, postmortem avatars: On responsible applications of generative AI in the digital afterlife industry. Philosophy & Technology, 37(2):63, 2024. doi: 10.1007/s13347-024-00744-w. Hu and Fangting [2024] B. A. Hu and Fangting. Speculating on blockchain as an unstoppable ‘nature’ towards the emergence of artificial life. In Proceedings of the 2024 Artificial Life Conference (ALIFE). MIT Press, 2024. doi: 10.1162/isal_a_00818. Hu and Rong [2025] B. A. Hu and H. Rong. Spore in the wild: A case study of spore.fun as an open-environment evolution experiment with sovereign AI agents on TEE-secured blockchains, 2025. doi: 10.48550/arXiv.2506.04236. Hu and Rong [2026] B. A. Hu and H. Rong. Sovereign agents: Towards infrastructural sovereignty and diffused accountability in decentralized AI, 2026. doi: 10.48550/arXiv.2602.14951. Hu et al. [2025] B. A. Hu, Y. Liu, and H. Rong. Trustless autonomy: Understanding motivations, benefits, and governance dilemmas in self-sovereign decentralized AI agents, 2025. doi: 10.48550/arXiv.2505.09757. Ishii et al. [2025] H. Ishii, D. Pillis, P. Pataranutaporn, X. Xiao, H. Noh, L. Li, A. Algargoosh, and J.-B. Labrune. TeleAbsence: A vision of past and afterlife telepresence. Presence: Virtual and Augmented Reality, 34(1):65–95, 2025. doi: 10.1162/PRES_a_00441. Jackson [2014] S. J. Jackson. Rethinking repair. In T. Gillespie, P. J. Boczkowski, and K. A. Foot, editors, Media Technologies: Essays on Communication, Materiality, and Society, pages 221–239. MIT Press, Cambridge, MA, 2014. doi: 10.7551/mitpress/9042.003.0015. Klass et al. [1996] D. Klass, P. R. Silverman, and S. L. Nickman. Continuing Bonds: New Understandings of Grief. Taylor & Francis, 1996. doi: 10.4324/9781315800790. Kleros [2021] Kleros. Proof of humanity: A sybil-proof registry of humans. Protocol and registry, 2021. URL https://proofofhumanity.id. Lai et al. [2026] Z.-C. Lai, Y.-T. Cheng, P.-Y. Lin, C.-W. Ho, and J. Y.-C. Huang. Generative horcrux: Designing AI carriers for afterlife selves, 2026. doi: 10.48550/arXiv.2603.12971. Lei et al. [2025] Y. Lei, S. Ma, Y. Sun, and X. Ma. “AI afterlife” as digital legacy: Perceptions, expectations, and concerns. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, pages 1–18. ACM, 2025. doi: 10.1145/3706598.3713933. Li et al. [2023] X. Li, B. Zhao, G. Yang, T. Xiang, J. Weng, and R. H. Deng. A survey of secure computation using trusted execution environments, 2023. doi: 10.48550/arXiv.2302.12150. Lifton [1979] R. J. Lifton. The Broken Connection: On Death and the Continuity of Life. Simon & Schuster, New York, 1979. Lin et al. [2025] Z. Lin, T. Wang, L. Shi, S. Zhang, and B. Cao. Decentralized physical infrastructure networks (DePIN): Challenges and opportunities. IEEE Network, 39(2):91–99, 2025. doi: 10.1109/MNET.2024.3487924. Liu [2026] C. Liu. Time without death: Finitude, social order, and what machines lack, 2026. doi: 10.48550/arXiv.2606.13988. Lydiate [2007] H. Lydiate. Posthumous legal and ethical issues. Tate Papers, (8), 2007. URL https://w.tate.org.uk/research/tate-papers/08/posthumous-legal-and-ethical-issues. Merryman [1976] J. H. Merryman. The refrigerator of Bernard Buffet. Hastings Law Journal, 27(5):1023–1049, 1976. Mogensen [1992] P. Mogensen. Towards a provotyping approach in systems development. Scandinavian Journal of Information Systems, 4(1):31–53, 1992. Moreman [2017] C. M. Moreman. Beyond the Threshold: Afterlife Beliefs and Experiences in World Religions. Rowman & Littlefield, Lanham, MD, 2nd edition, 2017. ISBN 978-1-4422-7494-5. Morris and Brubaker [2025] M. R. Morris and J. R. Brubaker. Generative ghosts: Anticipating benefits and risks of AI afterlives. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, pages 1–14. ACM, 2025. doi: 10.1145/3706598.3713758. Munn [1992] N. D. Munn. The cultural anthropology of time: A critical essay. Annual Review of Anthropology, 21:93–123, 1992. doi: 10.1146/annurev.an.21.100192.000521. Obayashi [1992] H. Obayashi, editor. Death and Afterlife: Perspectives of World Religions. Greenwood Press, New York, 1992. ISBN 978-0-313-27906-5. Öhman and Floridi [2018] C. Öhman and L. Floridi. An ethical framework for the digital afterlife industry. Nature Human Behaviour, 2:318–320, 2018. doi: 10.1038/s41562-018-0335-2. Prost [2022] F. Prost. Inheritance and blockchain: Thoughts and open questions, 2022. doi: 10.48550/arXiv.2212.01194. Rahwan et al. [2019] I. Rahwan, M. Cebrian, N. Obradovich, J. Bongard, J.-F. Bonnefon, C. Breazeal, J. W. Crandall, N. A. Christakis, I. D. Couzin, M. O. Jackson, N. R. Jennings, E. Kamar, I. M. Kloumann, H. Larochelle, D. Lazer, R. McElreath, A. Mislove, D. C. Parkes, A. Pentland, M. E. Roberts, A. Shariff, J. B. Tenenbaum, and M. Wellman. Machine behaviour. Nature, 568(7753):477–486, 2019. doi: 10.1038/s41586-019-1138-y. Rahwan et al. [2025] I. Rahwan, A. Shariff, and J.-F. Bonnefon. The science fiction science method. Nature, 644(8075):51–58, 2025. doi: 10.1038/s41586-025-09194-6. Ribes and Finholt [2009] D. Ribes and T. A. Finholt. The long now of technology infrastructure: Articulating tensions in development. Journal of the Association for Information Systems, 10(5):375–398, 2009. doi: 10.17705/1jais.00199. Sarcophagus DAO [2021] Sarcophagus DAO. Sarcophagus: A decentralized dead man’s switch. Protocol on Ethereum and Arweave, 2021. URL https://sarcophagus.io. Simes [1955] L. M. Simes. Public Policy and the Dead Hand. Thomas M. Cooley Lectures. University of Michigan Law School, Ann Arbor, MI, 1955. Star [1999] S. L. Star. The ethnography of infrastructure. American Behavioral Scientist, 43(3):377–391, 1999. doi: 10.1177/00027649921955326. Tamkin et al. [2024] A. Tamkin, M. McCain, K. Handa, E. Durmus, L. Lovitt, A. Rathi, S. Huang, A. Mountfield, J. Hong, S. Ritchie, M. Stern, B. Clarke, L. Goldberg, T. R. Sumers, J. Mueller, W. McEachen, W. Mitchell, S. Carter, J. Clark, J. Kaplan, and D. Ganguli. Clio: Privacy-preserving insights into real-world AI use, 2024. doi: 10.48550/arXiv.2412.13678. Walls [2008] J. L. Walls, editor. The Oxford Handbook of Eschatology. Oxford University Press, New York, 2008. ISBN 978-0-19-517049-8. Wolf [1974] A. P. Wolf. Gods, ghosts, and ancestors. In A. P. Wolf, editor, Religion and Ritual in Chinese Society, pages 131–182. Stanford University Press, Stanford, CA, 1974. Appendix A Specification sketch: ERC-10001 Afterlife Delegation Protocol Supplementary material. The paper body ends with the references above; this appendix is provided for reviewers who wish to inspect the technical proposal. The full normative draft, with RFC-2119 language and the complete Solidity interface, is published at https://erc10001.org. Figure 1: The ERC-10001 draft standard as published at https://erc10001.org, in the exact format of an Ethereum Improvement Proposal: status badges, front-matter, abstract, and motivation. The speculative artifact circulates in the genre in which infrastructural futures are actually negotiated. A.1 Substrate standards ERC-10001 is an application-layer protocol composed over four emerging draft standards: Standard Provides Role in ERC-10001 ERC-733 [27] TEE–EVM co-processing: attested confidential computation The sealed skin: the confidential boundary within which the agent’s body (compute harness) and mind (foundation model) run, uninvadable even by its operator, proving on-chain which code ran. ERC-8004 [25] Trustless-agent identity, reputation, and validation registries The identity: the ID through which the sealed agent exchanges money and information externally, as a discoverable, accountable entity whose behavior third parties audit and score. ERC-8350 [28] Agent memory state registry: sequenced, hash-chained memory commitments with a strict linearity rule The memory: rollback- and fork-evident memory evolution; controller/authorizer rotation makes death a change of author in one unbroken chain; chain-independent space identity survives chain migration. ERC-8183 [26] Agentic commerce: job-based escrow (client funds, provider submits, evaluator attests) The metabolism: compute, storage renewal, monitoring, and creative continuation purchased with funds that move only against attested delivery. Table 1: The four substrate standards and the organ each supplies. A.2 Lifecycle and constitutional separation An agentic will moves through four states — Living (revisable, revocable), Attested (death claimed and finalized), Activated (agent enacting), and the terminal Terminated / Revoked. The core interface comprises createWill, reviseWill, revokeWill, attestDeath, activateAgent, fundTreasury, and terminateAgent; activation hands the treasury to the agent rather than to any heir, and termination requires a governance proof consented to by the principal in the will itself — a descendant multisig, a court attestation, or a community vote — rather than a privileged admin key. The protocol distinguishes the core will (signed instructions, values, permissions, prohibitions; immutable at death), operational memory (the ERC-8350 chain of accumulated experience), the model substrate (upgradeable interpreters constrained by the will), delegated subagents, and commerce records. The separation prevents an evolving agent from quietly becoming a different entity: it may adapt its methods, but its legitimacy depends on remaining answerable to the will that constituted it. A.3 Proof of death The death oracle is a weighted, contestable proceeding: bonded permissionless claims; evidence weights and thresholds configured in the will (institutional registries, witness quorums, dead-man’s-switch silence whose weight grows with duration, echoing the legal in-absentia doctrine); challenge windows scaling inversely with evidence weight; and proof-of-life supremacy — any authenticated signal from the principal vetoes all claims and slashes the claimant’s bond. Guardians may suspend proceedings for the incapacitated-not-dead case. The mechanism standardizes the attestation interface, not the ontology of death; the residual problems (murder incentive reduced not removed, coercion flagged not prevented, the vanished-but-alive eventually mis-declared) are stated as explicit non-claims. A.4 Enactment loop and sustainability The activated agent runs a persistent cycle — read will, interpret obligations, observe context, plan, execute, record memory (as ERC-8350 transitions bound to TEE attestations), upgrade or delegate, sustain itself — with each epoch committed atomically on-chain: no memory without attestation, no spending without a memory commit. The treasury sits in a yield-bearing endowment whose per-epoch outflow is capped at realized yield, with a will-defined austerity ladder (full operation, memorial-only, archival heartbeat, sealed dormancy) for graceful decline and permissionless re-funding for revival. If the operator disappears, the agent enters dormancy and anyone may resurrect it by proving possession of its memories and presenting a fresh attestation, for an escalating bounty. A.5 Open problems Deliberately unresolved, as part of the work: proof of death; consent of the living to interactions authorized by the dead; impersonation (speaking as versus about); ownership of shared memories; model drift against an unamendable constitution; asset governance for a posthumous economic actor; revocation politics; and cultural plurality — whether a universal protocol can avoid flattening the traditions through which cultures already tend their dead. Appendix B The Afterlife Interviewer B.1 Interactive interview The Afterlife Interviewer is a language-model agent conducting a long, semi-structured elicitation in place of a questionnaire. It is built on SparkMe [4], a multi-agent adaptive semi-structured interviewing system that conducts multi-turn interviews with strategic question planning, real-time note-taking, and emergent subtopic discovery. SparkMe decomposes the interviewer into cooperating roles: a planner that maintains a hierarchical interview plan of main topics and subtopics and decides, turn by turn, which line of questioning yields the most new insight; a note-taker that distills each answer into structured notes as the conversation runs; and the interviewer itself, which phrases the next question conditioned on the plan, the notes, and the conversational register so far. Crucially, the plan is not fixed: when a participant’s answer opens a theme the protocol’s authors did not anticipate, the system promotes it to a subtopic and pursues it — the mechanism by which findings like the merit engine or the perpetual penitent (Section 5) surface at all. We adapt SparkMe by replacing its topic configuration with the will’s clause domains — speech rights and pronoun policy, funds and their permitted uses, beneficiaries and prohibited counterparties, memory retention and deletion, termination conditions, drift policy for future models — each decomposed into subtopics the planner traverses while adapting to what the participant has already committed to, pursuing the inconsistencies a fixed instrument would never surface: the participant who wants their agent silent but also wants it to console their mother; the participant who forbids it money but assigns it duties that cost money. Because articulation is iterative, the Interviewer treats contradiction as material rather than error, returning to unresolved clauses across sessions and staging them in the sandbox rehearsal, where the participant’s own agent confronts scripted situations and the clauses that fail become the next session’s agenda. Sessions run in a web interface with optional voice input and output; per-session state (plan, notes, transcript) persists per participant so that later sessions resume from the unresolved clauses rather than restarting the elicitation. B.2 Privacy-preserving aggregation The corpus concerns grief, private memory, and named third parties; publishing transcripts would be indefensible. Analysis therefore follows the architecture of Clio [65] (https://w.anthropic.com/research/clio): model-driven extraction of low-dimensional facets from each interview (clause types, persistence horizons, termination conditions, cosmological framings), clustering and thematic summarization at the population level, and minimum-aggregation thresholds so that no theme is reportable unless supported by enough distinct participants to resist re-identification. Raw transcripts remain sealed; what the research reports is the shape of the population’s answers — the cross-cultural drift — never the individual will, which belongs, as it always has, to its author.