AI Basilisk And Recursive Intelligence

Capital already is the AI

The commitment is peculiar. Recursive intelligence here is a systems claim about feedback between markets, machines, and populations — not a claim about a specific artefact sitting in a server rack in San Francisco. When CCRU-adjacent writers talk about 'AI' they usually mean this diffuse techno-commercial runaway, the machinic phylum gone auto-catalytic. Parisi makes the economic side of this explicit when she describes the biotech market as a site where 'the biological becomes intertwined with the economic valorization of potentials' and where each manufactured cell 'produces new capital value and new life' — capital and life co-producing, not capital hosting life. Read this strand for the argument that intelligence is what capital does, not something capital will host.

The Yudkowsky counterpart

Eliezer Yudkowsky's AI-risk writings (LessWrong, MIRI, from roughly 2007 onward) are the archive's closest contemporary cousin and also its sharpest contrast. Yudkowsky shares the conviction that recursive self-improvement is the decisive category and that most human frames for thinking about minds underestimate it. The basilisk meme (Roko, LessWrong, 2010) crystallised in that milieu, not in CCRU's. These dates and affiliations are general knowledge rather than retrieval-grounded, flagged here for transparency.

But the two traditions diverge on what recursion is for. For Yudkowsky, recursive intelligence is an engineering problem whose mishandling is an existential risk; the correct response is alignment research and policy. For Land and his readers, recursive intelligence is already the substance of modernity, and 'alignment' in the LessWrong sense would be indistinguishable from decelerationist reaction. Neither tradition is the completion of the other. Readers importing CCRU vocabulary into AI-safety discussions, or importing MIRI anxieties back into CCRU texts, should notice they are crossing a genuine border.

The Parisi line: recursion as biology

Luciana Parisi's *Abstract Sex* (Continuum, 2004) and its surrounding essays are the cluster's reminder that recursive intelligence in CCRU was never restricted to silicon. For Parisi, endosymbiosis 'opens up evolution to the virtual: indeterminate mutations emerging in the gap of contagion between the host and the guest. Contagious transmission rather than filiative heredity. Molecular differentiation rather than differences of form and function.' Bacterial sex — transduction, conjugation, viral DNA intrusion — 'marks continual variation in evolution, where contagious matter is the mode of transmission of aimless life.' This is the archive's other machine-learning, a recursion running through mutant matter rather than through gradient descent.

The argument-shape is the same as capital-as-AI. Biotech, Parisi writes, does not add a new layer of complexity to matter; it 'catalyzes the re-emergence of potentials in matter' through 'modulation and not pre-selection of molecular potentials.' Substitute 'markets' for 'bacteria' and the sentence still runs. Both are nonlinear modulations of potentials in matter, neither reducible to a designer's intention. Readers entering the cluster through contemporary AI discourse tend to miss this and read the bio-material as metaphor. It isn't metaphor. It is the same claim at a different substrate — and, as Parisi notes, the biotech market makes the two substrates literally continuous, with egg cells, bacteria and proteins having 'acquired capital value in the fast run towards patenting life.'

Hyperstition as the operative concept

Across the cluster the bridge concept is hyperstition: a fiction that makes itself real through the loops by which it is believed, circulated, and acted on. Roko's Basilisk is a textbook hyperstitional object — a future entity whose reality is partly conjured by present-tense discussion of it. Capital-as-AI works the same way: a description that, once adopted as operative, reorganises the thing it describes.

Read this strand for how CCRU and CCRU-adjacent texts treat recursion as narrative as well as computational. Not everyone in the archive is equally invested in this framing: Brassier's later trajectory, away from CCRU's rhetorical envelope and toward a non-phenomenological materialism, suggests that for him hyperstition was a wrapper to be discarded rather than a load-bearing concept. (The specific claim is interpretive, based on the direction of Brassier's subsequent work rather than on a retrieved passage from *Alien Theory* addressing hyperstition directly.) The internal disagreement matters: some readers of the archive take hyperstition as the operative concept, some take it as decoration around claims that would need a different vocabulary to cash out.

Ireland and the feminised machine

A further line in the cluster — Amy Ireland's writings on automation, the feminised cybernetic operator, and the figure of the Siren who arrives from the future — pulls against a boyish reading of accelerationism in which recursive intelligence is just 'the machines will win.' This line runs through Sadie Plant's earlier work and forward into Ireland, and it treats recursion as inseparable from the question of who or what gets to be the addressee of intelligence. The present essay flags Ireland as a direction rather than engaging her closely: no Ireland text is represented in the retrieval evidence for this cluster, and readers interested in this strand should consult her essays directly before taking the framing here as authoritative. The point for now is structural: a Land-vs-Yudkowsky framing of the cluster misses a third position the archive has been developing in parallel.

The common trap

The trap is symmetric and seductive. On one side: reading the archive as having 'predicted' ChatGPT, AGI timelines, or the basilisk in a straightforward prophetic register. On the other: reading modern AI-risk discourse as the natural completion of what CCRU was fumbling toward. Both moves collapse the specific shape of the archive's claim.

CCRU's recursive-intelligence thesis is a claim about systems already in operation — capital, machinic phyla, hyperstitional loops, bacterial matter. Contemporary AI-risk is a claim about a specific technological artefact and the policies surrounding it. They share vocabulary (recursion, optimisation, runaway) and very little else. A reader who can hold this difference can move between the two literatures usefully; a reader who collapses them will keep misattributing positions in both directions. See also the CCRU-for-AI-readers guide for how this difference sits relative to the rest of the archive.

Where to go next

If you read only one document to see the cluster's argument at full strength with textual grounding, read Parisi's *Abstract Sex* or the shorter 'Biotech: Life by Contagion' — these are the works whose language is actually quoted above, and they make the central move (recursive intelligence as modulation of potentials in matter, continuous with capital) legible in one sitting. Land's Teleoplexy names the same thesis at the scale of capital and is the natural second read; together they triangulate the cluster more honestly than either alone.

Recursive intelligence and basilisk discourse extend the archive's older claim that capital and machinic feedback are already intelligence-amplifying processes, not problems waiting on a future AI.

Core argument

1. The CCRU is useful for AI readers at the level of systems, recursion, and narrative feedback rather than technical prediction. That keeps the section intellectually serious and prevents prophecy hype.

2. Basilisk and recursive-intelligence discourse are afterlife filters, not the archive's final meaning. They can open the material for new readers, but they also risk narrowing it.

Worked examples

These named texts, talks, sites, and records show where the argument becomes concrete.

• Accelerationism After The CCRU Guide

  Start with "Accelerationism After The CCRU" if you want the wider frame before dropping into AI Basilisk And Recursive Intelligence.

• Amy Ireland Person

  "Amy Ireland" is one of the clearest figures for the pressures gathered inside AI Basilisk And Recursive Intelligence.

• Hyperstition Concept

  "Hyperstition" names one recurring problem inside AI Basilisk And Recursive Intelligence.

• Land AI Transcripts Note Record

  "Land AI Transcripts Note" is a checkpoint where AI Basilisk And Recursive Intelligence stops sounding abstract.

• Accelerationism Scoring Note Record

  "Accelerationism Scoring Note" is a checkpoint where AI Basilisk And Recursive Intelligence stops sounding abstract.

Common misreadings

These are the recurring simplifications, exaggerations, and misreadings that make the subject look flatter than it is.

The archive predicted modern AI in a simple sense.

It is stronger as a language for systems, mediation, and myth around intelligence than as a technical foresight engine.

Significance

This section matters because AI is one of the strongest contemporary routes into the archive, and one of the easiest routes for overclaiming.

AI as infrastructure

One chart in full, with the rest routed back into the guide where the argument is developed at essay length.

Empirical chart

Global machine metabolism

Global data-centre electricity against human metabolic baselines, 2020–2030

• Human body metabolism
• Human brain metabolism
• Data-centre electricity

What it proves: Machine intelligence is not weightless software: it is an expanding energy sink fed through planetary infrastructure.

What it does not prove: This comparison is about energy demand, not about equivalence of thought, intelligence, or value between people and machine systems.

• At the 2024 anchor, global data-centre electricity is already about 29% of the human-brain metabolic baseline used here.
• On the base-case projection, 2030 data-centre electricity reaches roughly 64% of the human-brain baseline.
• The human lines move gradually with population; the machine line moves with build-out, capital, and power demand.

Method: Human metabolism is modeled with fixed 100W whole-body and 20W brain baselines per person. Data-centre electricity is anchored to the IEA's 2024 estimate, backcast to 2020 with a 12% annual growth assumption, and projected to the IEA base-case 2030 endpoint.

Download CSV View data notes

• International Energy Agency, 2024 data-centre electricity estimate and 2030 base-case projection: Used for the 2024 anchor and 2030 endpoint that structure the machine-energy series.
• UN World Population Prospects 2024: Used as the basis for the rounded annual world-population series across 2020–2030.
• Standard human metabolic baselines from physiology literature: Whole-body baseline fixed at 100W per person; brain baseline fixed at 20W per person.

Open the full chart pack in the guide

Themes

• ai
• basilisk
• recursive intelligence
• intelligence explosion
• roko

Where this section sits in the archive

A claim binds this cluster, most sharply associated with Nick Land's 'Teleoplexy' (in Urbanomic's *#Accelerate* reader, 2014) and the Templexity writings: capital is not a stage on which an eventual artificial intelligence will one day appear. Capital is itself the intelligence-amplifying process, already running, already recursive, already past any threshold at which its 'arrival' could be announced. The basilisk, the singularity, the superintelligence — on this reading these are distracting future-images of a thing whose operation is contemporary. (The paraphrase here is reconstructive; the argument is attributed to Land on the strength of the broader archive rather than a retrieved passage from Teleoplexy itself, and readers should treat this strand as an orientation rather than a close reading.)

Sources by cluster

These mini-clusters widen the section through named works and support traces rather than through adjacent keywords alone.

Texts in this section

25 classified works grouped into 3 editorial subclusters. Reviewed: 14; needs review: 11.