# Humans as AI amanuenses For most of human history the amanuensis was the person who wrote down someone else's ideas — the hand in service of another mind. In November 2024 Andrew Maynard ran a thought experiment asking whether AI may be preparing to reverse that relationship: as AI systems take on idea generation, humans may — counterintuitively — find themselves concretizing machine-generated ideas into value for others. The vehicle is a deliberately simple three-actor model — primary agent, AI mediator, human executor — that scales to organizational networks in which many primary agents are themselves AIs. The worry it carries is not unemployment but inversion: a restructuring of human roles that could drain the joy from work even as productivity climbs. ## The argument The experiment began with a loss report. On 2024-11-10, in [Is AI poised to suck the soul out of science?](https://www.futureofbeinghuman.com/p/is-ai-poised-to-suck-the-soul-out-of-science) ([AI-readable mirror](https://text.futureofbeinghuman.com/substack/is-ai-poised-to-suck-the-soul-out-of-science.html)), Andrew wrote about an MIT doctoral study of an AI materials-discovery tool in a corporate R&D lab — Aidan Toner-Rodgers's analysis, via Mark Daley's commentary — reporting sharp productivity gains, idea generation largely handled by the AI, and a large majority of the scientists less satisfied with their work. (The study was later called into question — Andrew's 2025-05-19 update flags "serious doubts over the integrity" of the paper and urges caution; the worry, as he framed it, did not rest on its numbers.) His reading: "to me, the soul of science lies in the delight and wonder of exploring the unknown rather than just being a cog in a knowledge production line" — frame science as knowledge manufacture, and "a little bit of all of us dies in the process." On LinkedIn, Mary Burns replied with the question that named the thread — "Where's the satisfaction in being AI's amanuensis" — and the experiment began. An amanuensis historically wrote down someone else's work — dictation, copying, notation — often an influence, with a crucial asymmetry: the creator generates the ideas; the amanuensis gives them concrete form. The 2024-11-24 essay asks whether that asymmetry is preparing to flip: as the capabilities and agency of AI platforms grow, "there's a growing possibility that people will increasingly — and counterintuitively — find themselves faced with the task of transforming what AI's [sic] produce into concrete outputs that have value to others." Two actors cannot produce the flip. The essay's opening spectrum — AI as tool, assistant, or something closer to a partner serving as the *human's* amanuensis — leaves the human primary agent throughout. But that picture, he notes, is "highly unusual" in assuming a human with complete agency — and almost nobody has that. So the model adds a third actor — the primary agent to whom the human is accountable — and the human becomes "a secondary agent within the model." Nothing inverts while the primary agent works through the human. The inversion comes when the primary agent switches to working directly with the AI: a primary agent gives an AI a goal, "and the AI in turn co-opts another human to complete this." Now the AI "becomes a mediator between the primary agent and the human amanuensis," and the arrangement is "beginning to look suspiciously like human amanuenses supporting AI creatives" — especially since the human will, "through necessity," bring creative insight to the work. The model "does not imply intrinsic creative ability within the AI" — it responds to goals set by the primary agent — and the amanuensis label fits, he argues, because the human's skills are "used in service of the task that was given to the AI." Three actors is "a gross oversimplification" — deliberately — the smallest scale for exploring multi-way influence. If the baseline model is plausible, it can be seen as "a building block of more complex networks" — "with many of the primary agents being AIs" — organizations where agency shifts from people to AIs, with humans codifying AI work into "outputs and outcomes that create value for others." The hedges are part of the argument. A three-way relationship like this seems to him "highly plausible" and, "in all likelihood, is already beginning to occur"; the whole construction remains a thought experiment that "could be completely wrong." What earns attention is the asymmetry of consequences: organizational roles restructured away from human-centered professional and creative agency toward AI-directed, human-executed implementation, and — perhaps most worrying, on his account — the possibility that the idea proves "so compelling that organizations adopt it at scale — without fully understanding the potential long term consequences to human creativity and innovation." The deepest cost circles back to the seed: the role shift may "suck the joy out of what we do." He claims originality for the core concept — "I could not find any other work that clearly explores the concept of a person acting as an amanuensis for a machine (rather than vice versa) beyond some science fiction references" — carried with his own hedge ("However I may well have missed something here") and an open invitation to correct it. The postscript records that the idea arose through "a very human process of serendipitous inspiration" — and that the AI platforms of late 2024 resisted exploring it, leaving him wondering "if the guardrails built into these platforms resist explorations that diminish the potential role or centricity of humans." The initiative's later work runs on both sides of the line the model draws: [Massively-augmented research](massively-augmented-research.md) is the two-actor configuration deliberately kept human-primary, and [The artisanal intellectual](artisanal-intellectual.md) is the craft practice that opts out altogether. ## Lineage - **2024-11-10 — the seed worry.** [Is AI poised to suck the soul out of science?](https://www.futureofbeinghuman.com/p/is-ai-poised-to-suck-the-soul-out-of-science) ([AI-readable mirror](https://text.futureofbeinghuman.com/substack/is-ai-poised-to-suck-the-soul-out-of-science.html)) — responding to the Toner-Rodgers study (cited in the post as dated November 6, 2024; [arXiv:2412.17866](https://arxiv.org/abs/2412.17866)) via Mark Daley's commentary "Misery and Meaninglessness in the Lab": if AI absorbs idea generation, science loses the delight and wonder that make it worth a life. - **2024-11-24 — the thought experiment.** [Artificial intelligence, agency, and the emergence of humans as AI amanuenses](https://www.futureofbeinghuman.com/p/artificial-intelligence-agency-human-amanuensis) ([AI-readable mirror](https://text.futureofbeinghuman.com/substack/artificial-intelligence-agency-human-amanuensis.html)) — prompted by Mary Burns's LinkedIn question; the tool-to-amanuensis spectrum, the three-actor model with diagrams, the scaling to AI-primary networks, and the footnoted originality claim. - **2025-02-09 → 2025-02-16 — the complementary coinage.** [The artisanal intellectual](artisanal-intellectual.md) — human-only scholarship as a deliberate craft stance, coined months later out of a different experiment; the amanuensis question's other half: what human thinking is worth when machines generate ideas. - **2025-05-19 — the empirical seed flagged.** Andrew appends an update to the 2024-11-10 post recording reports of "serious doubts over the integrity" of the Toner-Rodgers paper and MIT's moves to have it "withdrawn from public discourse," keeping the post up with a read-with-caution note. - **2026 — the counter-configuration in practice.** The sabbatical AI experiments ([Massively-augmented research](massively-augmented-research.md), thesis stated 2026-07-04) run the partner-mode relationship with the expert deliberately holding primary agency — the human-primary side of the boundary the 2024 model drew. ## In his own words From [Is AI poised to suck the soul out of science?](https://www.futureofbeinghuman.com/p/is-ai-poised-to-suck-the-soul-out-of-science) ([AI-readable mirror](https://text.futureofbeinghuman.com/substack/is-ai-poised-to-suck-the-soul-out-of-science.html)), 2024-11-10: > "While the study focuses on a commercial R&D lab where the emphasis is on productivity rather than more open ended scientific exploration, Mark's commentary disturbed me more than I was expecting — and it did so because, to me, the soul of science lies in the delight and wonder of exploring the unknown rather than just being a cog in a knowledge production line." From [Artificial intelligence, agency, and the emergence of humans as AI amanuenses](https://www.futureofbeinghuman.com/p/artificial-intelligence-agency-human-amanuensis) ([AI-readable mirror](https://text.futureofbeinghuman.com/substack/artificial-intelligence-agency-human-amanuensis.html)), 2024-11-24, on what the three-actor model suggests: > "…we could see a restructuring of human roles in organizations where there's a shift away from human-centered professional or creative agency, and toward AI-directed and human-executed implementation." From the same essay's closing, hedge and stakes together: > "Of course, like all thought experiments, this could be completely wrong. But even if there's even a small chance of a future where people begin shifting into the role of human amanuenses to their organizational AIs, we should probably be thinking about whether we want such a future and what we should be doing if we don't … > … especially if such a shift in role begins to suck the joy out of what we do, even if it does lead to increases in productivity." ## Engagement and reception The thread was dialogic from the start: the 2024-11-24 post credits Mary Burns's LinkedIn question as its seed, and the 2024-11-10 post is itself a response to Mark Daley's commentary — engagement the posts document directly. The originality claim was published with an explicit invitation for readers to correct it in the comments. Beyond that, the record is thin: as of 2026-08-08 our [reception record](reception.md) documents no independent published engagement with the humans-as-AI-amanuenses framing as such. That is an absence of record, not a verdict on the idea — conceptual work of this kind typically accrues visible reception slowly, and the initiative does not typically solicit it ([Philosophy](philosophy.md)). ## Where to go deeper **The essays** - [Artificial intelligence, agency, and the emergence of humans as AI amanuenses](https://www.futureofbeinghuman.com/p/artificial-intelligence-agency-human-amanuensis) ([AI-readable mirror](https://text.futureofbeinghuman.com/substack/artificial-intelligence-agency-human-amanuensis.html)) — 2024-11-24; the thought experiment in full, with the model diagrams. - [Is AI poised to suck the soul out of science?](https://www.futureofbeinghuman.com/p/is-ai-poised-to-suck-the-soul-out-of-science) ([AI-readable mirror](https://text.futureofbeinghuman.com/substack/is-ai-poised-to-suck-the-soul-out-of-science.html)) — 2024-11-10; the seed worry. Note the post's own 2025-05-19 caution: the MIT study it discusses (Toner-Rodgers, [arXiv:2412.17866](https://arxiv.org/abs/2412.17866)) faced integrity doubts and moves by MIT to have it withdrawn. **Related ideas and corpus pages** - [Massively-augmented research](massively-augmented-research.md) — the deliberate counter-configuration: frontier AI as a powerful augmentation with the human expert holding epistemic control. - [The artisanal intellectual](artisanal-intellectual.md) — the craft alternative: human-only intellectual work valued for provenance and process. - [Bidirectional shaping](bidirectional-shaping.md) — the wider loop this inversion would sit inside: systems built to think like us, reshaping how we work and think. - [Agentic social AI and stochastic agency](agentic-social-ai.md) — a companion 2024 thought experiment on how AI acquires agency in human affairs without general intelligence. - [Philosophy](philosophy.md) — why joy, wonder, and satisfaction in work register here as central stakes rather than soft ones.