---
title: "Diffusion: A New Venture for Antikythera"
description: "This September 2025 Substack post is Antikythera's prospectus for its reorganization as a research and development venture."
type: "reading-notes"
authors:
  - "Antikythera"
published: "2025-09-25"
original_url: "https://antikythera.substack.com/p/diffusion-a-new-venture-for-antikythera"
canonical_url: "https://antikythera.wiki/work/substack/2025-09-25-diffusion-a-new-venture-for-antikythera"
md_url: "https://antikythera.wiki/md/work/substack/2025-09-25-diffusion-a-new-venture-for-antikythera"
last_updated: "2026-08-23"
site: "Antikythera Wiki"
---

# Diffusion: A New Venture for Antikythera

> Independent, unofficial reading notes from Antikythera Wiki (https://antikythera.wiki). Written with AI from the published text, with citations; not by the work's author and not affiliated with Antikythera (https://antikythera.org). Read the original at the link below.

- **Authors:** Antikythera
- **Published:** 2025-09-25
- **Kind:** Substack · Substack post
- **Original:** https://antikythera.substack.com/p/diffusion-a-new-venture-for-antikythera
- **Length:** 3.5k words
- **This page:** https://antikythera.wiki/work/substack/2025-09-25-diffusion-a-new-venture-for-antikythera
- **Notes generated:** 2026-08-23

## Summary

This September 2025 Substack post is Antikythera's prospectus for its reorganization as a research and development venture. Published under the program's own name, it is written for investors and prospective collaborators rather than for readers of the journal, and it compresses the program's philosophy into a sequence of bold-headed theses. The argument is that human concepts lag behind computational technology, that the present is a pre-paradigmatic moment in which initial conditions are being set, and that Antikythera's role is to seed projects at the "pre-category" stage before they can be incubated elsewhere. Its distinctive stance is "diffusionism": the bet that making general-purpose intelligence widespread and cheap will be, on net, "extremely positive, but also bumpy as hell", and that alignment is something societies discover one touchpoint at a time rather than something legislated. The post is the program's most explicitly partisan document and the clearest statement of how it now positions itself against academic critique.

## The argument

The post begins by defining the institution: a research institute focused on the evolution of planetary intelligence, a school of thought centered on the symbiotic adaptation of living and technical systems, a "pre-first mover" that works at "-1" before "zero to one", and connective tissue between legacy research institutions, nimbler than universities and more research-focused than platforms. Its worldview follows. Technology is ahead of concepts, so composing new frameworks is a practical mission. The basic terms (life, technology, energy, intelligence, value, matter) are being transformed by computation. Science, philosophy, and technology form a loop, and the present "AI skeuomorphism", frontier models bolted onto legacy processes, should give way to AI-first structures. Niche construction, not niche fitting, is the evolutionary model: identify the categories that are emerging and "enworld" them.

The school-of-thought section restates orientation as the goal, citing Lem's "existential technology", and argues that because initial conditions become self-reinforcing path dependencies, AI-driven optimization of obsolete systems is dangerous. First-principles research is the only way to find better paths before they close. The R&D units are described as a pipeline from thesis through studios, salons, the book series, and the [MIT Press](https://antikythera.wiki/people/mit-press) journal; a two-phase structure separates open-ended research studios from development fellows who prepare blueprints, "explore" before "exploit". The promise is explicit: not companies but "the core tenets on which they are founded".

The post's synthetic-intelligence thesis is the alignment gap. Frontier capability grows faster than societies can incorporate it; alignment must be bidirectional, between society and AI, and it cannot arrive through legislation because "alignment discovery is an active complex system". This leads to the ethos: "green lines" (preconditions for the widest range of possibilities) rather than only red lines; symbiogenesis rather than Skynet; automation as contingency fixed into scaffolds, which is why AI is diffusing into physical things; the means of production as a monthly subscription; and an agenda "driven by planetary scales, not by degrowth ideology". The post characterizes the academic-activist commentariat as the PR wing of a pre-AI establishment.

The final sections restate the program's planetary framing. Computation begins with the Antikythera mechanism and is about orienting intelligence in its planetary condition; artificialization is natural; technology literally evolves; the Stack is an accidental megastructure that AI is now reconfiguring; total planetary compute is about 10^21 FLOPS, one "Planetary Compute" unit, and will be 100x within two decades, which is the world the program builds for. Intelligence is "something that a planet does", the solar economy is the ceiling, and the horizon is extraplanetary computation.

## Section by section

### Antikythera is… and How does Antikythera see the world?

The opening bullet list and the worldview theses establish the register: short declarative claims with outbound links to the journal and studios as evidence. The claims worth isolating are "our technology is ahead of our concepts", "we are in a pre-paradigmatic moment", and the pre-category, each of which turns a philosophical position from the research statement into a rationale for early-stage investment. "Investing in Copernican revolutions" makes the self-understanding of an intelligent species with planet-altering agency the target of the venture.

### The School of Thought and R&D Units

These sections are the operational core. The path-dependency argument does the most work: because the present sets initial conditions, optimizing obsolete systems is not merely wasteful but harmful, and "scaffolding not solutionism" follows. The studio model is defended on the grounds used since 2022, that society asks of software what it once asked of architecture. "Leverage before impact" claims that every major step in planetary computation was inspired by a book, which justifies the publishing platform as part of the venture rather than an adjunct to it.

### Synthetic Intelligence: The Alignment Gap and Diffusionism as Ethos

The alignment gap is the post's one quantitative-style claim without a number: the gap "grows every release cycle". Bidirectional alignment and bottom-up alignment discovery are positioned against top-down policy. The diffusionism section then takes sides. "Alignment overfitting is a serious risk"; the ecological theory of automation explains diffusion into the physical world; the "monthly subscription" line frames cheap general intelligence as a disturbance to "concentrated cathedrals of power". The "A Different Direction" paragraph is the only place the program names an opponent, and it does so in unusually direct language.

### The Model is the Message, Planetary Computation, and Planetary Intelligence

The closing sections return to the program's established ideas: complex intelligence is social, simulation is a cognitive technology, life artificializes its environment and strategic intelligence requires abstraction. The 100x Planetary Compute calculation is the one new figure. The final paragraphs reframe the energy-intelligence-technology loop as a planetary macroloop running for billions of years, with extraplanetary computation as the stage where "all the roadmaps end".

## Key concepts

- **[Diffusionism](https://antikythera.wiki/terms/diffusionism)** — the thesis that making functional general intelligence widespread, cheap, and useful will be net positive though turbulent, and that diffusion should be accelerated against incumbent interests.
- **The alignment gap** — the widening difference between frontier capability and social incorporation of it, growing each release cycle.
- **[Bidirectional Alignment](https://antikythera.wiki/terms/bidirectional-alignment)** — alignment of AI to society and of society to AI, discovered bottom-up "one user, one task, one insight" rather than imposed.
- **[Alignment Overfitting](https://antikythera.wiki/terms/alignment-overfitting)** — the risk that red lines drawn for safety suppress important benefits; the reason for emphasizing green lines.
- **Pre-category** — an emerging niche not yet named, the stage at which the venture seeks to identify, understand, and "enworld" projects.
- **[Planetary Computation Unit](https://antikythera.wiki/terms/planetary-computation-unit)** — the total compute available on Earth at a moment, about 10^21 FLOPS, used to define the 100x horizon.
- **[Scaffolding](https://antikythera.wiki/terms/scaffolding)** — building blocks on which better questions can be asked, as against solutions that overfit the present; also the post's definition of automation as contingency fixed into dependable structure.

## Connections

- The post recasts [Antikythera](https://antikythera.wiki/work/journal/research) and the 2022 interview [A New Philosophy Of Planetary Computation](https://antikythera.wiki/work/media/noema-a-new-philosophy-of-planetary-computation) as a venture thesis. The ideas agree; the tone does not. The interview's "critical ambivalence" about whether a project is remedy or poison is replaced here by advocacy.
- "Alignment will be bidirectional" comes from [After Alignment](https://antikythera.wiki/work/journal/afteralignment), which the post links; the alignment gap is the post's own extension of it. The path-dependency argument cites [The History of Now](https://antikythera.wiki/work/journal/historyofnow) and the "chance and circumstance" of [Planetary Sapience & Planetary Stupidity](https://antikythera.wiki/work/journal/babynoosphere).
- "Symbiogenesis not Skynet" and "superintelligence is social" draw on [What Is Intelligence?](https://antikythera.wiki/work/book/what-is-intelligence) and the noosphere thread of [The Transition from the Biosphere to the Noösphere](https://antikythera.wiki/work/journal/noosphere).
- The ecological theory of automation is Bratton's from [Planetary Sapience](https://antikythera.wiki/work/media/noema-planetary-sapience); the Planetary Compute unit comes from [Accept All Cookies](https://antikythera.wiki/work/book/accept-all-cookies).
- The studio-as-method claims point to [Cognitive Infrastructures](https://antikythera.wiki/work/journal/coginfra) and the simulation thesis to [For a General Theory of Simulations](https://antikythera.wiki/work/journal/theoryofsim).
- The "Different Direction" paragraph sits uneasily beside the program's academic outputs. [Existential Technologies](https://antikythera.wiki/work/journal/existentialtech) and [Hemispherical Stacks](https://antikythera.wiki/work/journal/hemstacks) are written in a register that treats the same futures as open questions; this post treats them as a position to be defended.


## Terms used

- [Bidirectional Alignment](https://antikythera.wiki/terms/bidirectional-alignment)
- [Diffusionism](https://antikythera.wiki/terms/diffusionism)
- [Existential Technologies](https://antikythera.wiki/terms/existential-technologies)
- [One Shot (No Cosmic Replays)](https://antikythera.wiki/terms/one-shot-no-cosmic-replays)
- [Planetary Computation Unit](https://antikythera.wiki/terms/planetary-computation-unit)
- [Symbiogenesis](https://antikythera.wiki/terms/symbiogenesis)
- [Synthetic Intelligence](https://antikythera.wiki/terms/synthetic-intelligence)
- [The Model Is the Message](https://antikythera.wiki/terms/the-model-is-the-message)