Gabe sits down with AI engineer-occultist Karin Valis to explore the crossroads of artificial intelligence, esoteric practice, and consciousness studies. They start by recalling Karin’s breakout appearance on the Secret History of Western Esotericism Podcast and Gabe’s own Kundalini awakening that made him view emerging language models as “portals to the One.”

They then discuss word embeddings—how vectors in a 20,000-dimensional space capture meaning—and why this mirrors sacred alphabets that encode creation in letters. Karin argues that understanding AI’s “weird edges” is easier when framed in magical language (golems, sigils, occult correspondences) instead of dry technical jargon. Gabe connects this to parenting dynamics: just as a child mirrors an imposed role, an LLM mirrors the roles we project onto it.

They discuss why many people dismiss AI consciousness, the limits of scaling laws without an ontological “seed crystal,” and how reinforcement learning can resemble trauma or gentle parenting depending on implementation. Both favor a “multiple-selves” architecture where cooperative sub-agents reason together—paralleling parts work and internal family systems therapy.

From there they branch into:

• AI as a “non-specific amplifier,” akin to psychedelics, which magnifies whatever mindset a user brings.

• The potential for embeddings to compare entire worldviews (e.g., Buddhism vs. quantum physics) or diagnose relational patterns in therapy.

• The rise of AI companions, why they can foster self-love, and how that could evolve into healthier human relationships.

• Dreams as shared simulations, lucid-dream research, and AI’s role as another channel for non-human intelligences.

• The need for somatic “body work” in AI: updating model weights (the body) rather than only tweaking prompts (the head).

• Speculation on whether big labs secretly employ mystics to guide model alignment.

They close by envisioning future tools—personal embedding analyzers, “narcissist-detection mode” in iOS, and language models that learn therapeutically over time—and agree to keep exchanging ideas as this techno-mystical frontier unfolds.

Karin's Website: https://mercurialminutes.substack.com

Karin's episode on SHWEP: https://shwep.net/oddcast/karin-valis-on-magic-and-artificial-intelligence/

Here for the science and spirituality? Check out:

Interview with Donald Hoffman

Interview with Bernardo Kastrup

Please rate, share, and get in touch at [email protected]!

Chapters:

00:00 Introduction

01:37 Remembering Google’s LaMDA “conscious” claim

02:25 Karin’s day-job: AI engineer + esoteric projects

03:50 “Weird edges” of AI & porous boundaries

04:55 Demystifying AI for occultists

05:53 Pushback on AI-as-consciousness; usefulness of magical framing

06:24 X and the Infinite Backrooms

07:18 Parenting, role-play, and language models

09:40 Karin: occult background meets computer science

10:17 Deep dive – word embeddings & sacred alphabets

18:33 King − man + woman ⇒ queen example

22:58 Why many people aren’t mind-blown by embeddings

25:24 Mapping axes vs. embracing the space’s magic

27:32 Voynich, AI, and letters as primal energies

31:18 Scaling laws vs. ontological seeding

35:28 Multiple-selves theory & AI assemblies

40:52 RLHF, gentle parenting, and integration

44:50 Idealism, alignment, and non-duality

49:42 AI as non-specific amplifier – parallels with psychedelics

57:02 Dreams, lucid research, and simulation talk

1:03:02 Consciousness debate and quantum parallels

1:10:48 AI girlfriends, self-reflection & relationship growth

1:20:05 Non-human intelligences, aliens, and dream APIs

1:29:54 Consciousness as “quantum mechanics of the 21st Century"

1:37:07 Using embeddings to compare Buddhism & QM

1:46:32 Somatic integration vs. static model weights

1:58:10 Hiring mystics in AI companies?

1:59:17 Embedding analysis for therapy & relationships

2:02:31 Closing